Mac ATI Video Card OverclockingReturn to News Page
Thunderbolt Dual Drive Dock

Reader Feedback on ATIccellerator II Overclockhrefing Utility
Posted: 8/30/2004
Latest SW Version: 1.0.6b
Reports Updated: 1/22/2008

This page has reader reports on using ATIccelerator II Overclocking utility. See the linked homepage for warnings/disclaimer before using. As with any overclocking there's some risk and "your mileage may vary" but it's an easier way to experiment than the old (Mac) method of tweaking clock frequencies in the card firmware/reflashing the card, etc.

This page has desktop and Powerbook/iBook owner reports on using it to overclock their ATI cards and on-board graphics chips. Some reports included benchmark tests.

    Version History:
  • 1.0.6b - Fixed memory frequency for the 1.5 GHz Mac mini's Radeon 9200
  • 1.0.6a - Support for the iMac G5 (iSight)
  • 1.0.6 -Now requires Mac OS X 10.3.9 or later (up from 10.3.0)
    - New tab showing information about cards: amount of memory, memory type, and bus width - Support for the iBook G4's Mobility Radeon 9550 - Support for the latest PowerBook G4's Mobility Radeon 9700 - Fixed the G5's non-Pro 9600 memory frequency detection - Fixed support for a Rage 128 Pro variant - No longer load and crash for the unsupported Rage Pro/Mobility cards

    Requirements: Mac OS X 10.3.9 or later and ATI graphics card not older than the Rage 128"

If you send a report for posting here (news at xlr8yourmac.com), please note what version you're using. (Also include system/graphics card/chip details and OS version used. Thanks.)

Reader Reports on ATIccellerator II
(I welcome feedback on this utility from readers - please include mac model/ATI model/OS version and ATIccellerator II version you're using. thanks.)

(added 1/22/2008)
"I used ATIccellerator II v.1.0.6b to overclock my stock ATI Rage128Pro, ROM Revision: 113-63001-110, on my Powermac G4 "AGP Graphics" 450 (upgraded with Sonnet Encore to 1 GHz). I achieved these levels before I started seeing artifacts:

  • Processor: 255.00 MHz (+18.6%)
  • Memory: 146.00 MHz (+13.94%)

My system is stable at the above settings under all apps I could test. OS X GUI, DVD player, iPhoto and web browsing seem noticeably faster. Very nice application, easy to use, am very impressed by results. I am running OS X 10.4.11.
Regards, Pat"

(added 7/10/2007)
"I'm interested in your opinions on my experience. Unlike most people I used ATIccelerator II to down clock my graphics card. I recently started experiencing artifacts while playing 3D games and was getting frustrated trying to figure out the problem (while also looking for scarce and expensive replacement cards). Then I wondered what would happen if I changed my card's operating frequencies. I knew there was a tool available, so I found the ATIccelerator program and tried it out. And happily, I was able to eliminate the artifacts by tweaking my card's operating frequencies. The best completely stable condition (+2.25% CPU / -7.5% memory) leads me to belive that my card has developed a memory issue or some form of cooling issue. (Most of my games worked okay at -5% for memory, however, CIV IV needed it a bit lower.)

FYI my system specs are:
Type: 2004(June) Dual 2.5GHz G5, 3GB RAM
OS: Mac OS 10.4.10
ATIccellerator II vers: 1.0.6b
Video Card: ATI Radeon 9800XT (OEM card w/256MB VRAM)
Clocks (original): 400.50/360.00 MHz (core/memory)
Clocks (stable): 409.50/333.00 MHz
Thanks, James G."

I'd also check the card for dust buildup on the heatsink fins, fan operation, etc. - you may be surprised what you find (dust clogging the heatsink fins, etc). There's a recent article here also on the subject - see Overheating X1900 XT due to Dust Clogging Heatsink Fins.

(added 10/9/2006) "...I downloaded and tried ATIccellerator to great success, and though it's probably not as useful now as it once was, I thought it'd be helpful to report my findings (since much info on the 1.42 GHz G4 iBook seems pretty scarce).

Here are my System Specs:
Type: 14" iBook G4
Proc: 1.42 GHz G4
RAM: 512 MB
OS: Mac OS 10.4.8
ATIccellerator II vers: 1.0.6b
Video Card: ATI Mobility Radeon 9550
Clocks (origional): 209.25/182.25 MHz (core/memory)
VRAM: 32 MB DDR 64bit
Clocks (stable): 398.25/256.5 MHz
Clocks (stable but show low-moderate artifacts): 375/300
(approximately, distortion varies with each overclock)

I found these numbers incredibly high, and as you can imagine it pretty much breathed new life into my machine, especially when playing Halo and WoW. I can get 20-30 FPS outside of major cities, up to 40+ FPS in interior spaces (cities/BGs still lag about the same, probably due to my limited amount of RAM). (I am running WoW at 16 bit however, which in itself can increase the FPS by about 25-50%). Generally I see about 50% increase in frame rates across the board. I have played games for several hours at 350/250, and it is 100% stable, though the card temp can hit 80C pretty easily, and the fans are on constantly. I wouldn't dare to try this on battery power :P. Another use for ATIccellerator: I can use it to underclock my card to gain 20-30 mins of life. Anyways, I hope this dated info is useful.
-Jonathan "

IIRC in the past sometimes reported speeds were not accurate (although you noted better performance). That's a steep overclock esp. for a notebook. (Hope it stays reliable over time.)

Problems with 1.06b after 10.4.8 Update (flashed 9800 pro card) (First report from 10.4.8 user, however later report above didn't mention any problems in 10.4.8.)

(added 10/2/2006)
"10.4.8 breaks aticcellerator (using 1.06b?-Mike)
My screen went totally wonky after updating. I removed aticcelerator from my startup items in safe mode, rebooted, all is well. To confirm my hypothesis, I opened aticcelerator and had the same problem.
(I asked for info on his Mac/ati graphics card)
I am using ATIcellerator 1.06b (latest as of 10/2/2006). I wasn't able to find the prefs file. (to see if the app would run in 10.4.8 without overclocking the card/not using previous OC settings) I'm using a flashed PC Radeon 9800xt running 412/365 in a Digital Audio dual 533mhz. I know, the card is a little overkill, but I got it cheap. I didn't have any problems in 10.4.7. After upgrading to 10.4.8 I could boot in safe mode, but if I opened the ATIcellerator preference pane the screen would go crazy just like when I booted in normal mode.
-Joshua "

I wrote the author of ATICellerator II about this and he replied.

"This is surprising, but I'll test tomorrow. (Time to go to bed here!)
-Thomas P. "

A later report above from an IBook G4 owner is using 1.06b with 10.4.8.

Another ATI X800 Report:

(added 5/23/2006)
"I've just overclocked one in my 1.6GHz G5 to:
Processor: 560.25MHz (+18.57%)
Memory: 587.25MHz (+17.57%)
Totally solid in any app; warcraft 3, Unreal tournament 2004, Call of duty, and all the ati X800 demos.
The fan has been upgraded to an Arctic Cooling 4 and (the G5) has 1GB of ram.

My installation guide here:
Simon C."

ATICellerator II Author comments on Updates for X1600 Support (4/19/2006) In yesterday's main news page post on Intel CPU Mac ATI X1600 overclocking (using ATItool for WinXP), I mentioned I'd written the author of to ask if an update was planned for X1600 support. Here's his reply:

" Intel support has been done for some time, but the X1600 isn't supported yet... I lack information on the way it operates. ATIccelerator II will be updated when I finally have it!

(I wonder if the author of ATItool would be willing to share some info-Mike)
I already tried a few weeks ago, he didn't answer. I'll try again one of these days...
-Thomas "

Quake 4 Image 'Snow' w/overclocked 9800 SE card:
A couple weeks ago a G5/9800 SE card owner wrote that he saw display artifacts ("snow") running Quake 4 but not in other games (including Doom3, which is the engine used by Quake4). That reader later said it seemed to be helped by replacing the display cable but after another similar complaint this weekend (another G5/9800SE/Dell 24in display owner) I asked him if the problem was ever seen again and he said yes, although it wasn't as bad as before. The 2nd report on Quake4/'snow' included screenshots (peppered with white spots in the game screen) so it wasn't a cable issue as it shows in the framebuffer. He later wrote that he'd forgotten to mention he had overclocked the card using ATIcellerator II:

" Here's an update/solution for you and your readers. After emailing you last night and continuing to toil away, though it has never effected anything at all before in this manner, I shut off ATIcellerator II out of "I've tried everything else so i might as well try this" frustration and whaddayaknow, that solved the problem.
(I didn't know he had overclocked the card before, but definitely something to remember for those that have. If you're overclocking a card and see any display problems, back off to the stock speeds as a test.-Mike)
Even at the fairly modest overclock I had it set at 418.50 (10.71%) / 366.75 (8.67%), it was a major issue with Quake 4. Oddly, I didn't have this issue at all with Doom 3 (in fact ATIcellerator II helped pep its performance up a touch), but perhaps the engine has been tweaked significantly.

Hopefully, this info will help others fix or outright avoid this problem on their machines. A side note; this shouldn't scare folks away from ATIcellerator II, the program has been fantastic and, at the settings I use, has never caused any problems with any other game or software.
Thanks, Christian "

I asked the first Quake4/9800 SE owner that reported the problem if he had overclocked the card and he said no. (Would be interesting to see if downclocking it would help though.)

9800 Pro SE Card:

(added 12/27/2005)
"I've had this card overclocked for about two months with no issue's. ATIccelerator 2 1.0.6, Processor at 441.0 MHz, Memory at 342.0 MHz. (I asked what the stock clocks were)
I think the stock speeds were GPU 365 and Mem 335

Power Mac G5 (Dual 2GHz/4GB RAM) with 10.4.3

ATI Radeon 9800 Pro: Retail G5/256
Chipset Model: ATY,R350
Bus: AGP
VRAM (Total): 256 MB
Vendor: ATI (0x1002)
Device ID: 0x4e48
Revision ID: 0x0000
ROM Revision: 113-A14402-126"

OEM 9600XT 128MB card:

(added 12/26/2005)
"ATIccelerator II 1.0.6, G5 dual 2.5GHz,
OEM 9600XT 128MB video card, OS 10.4.3
I was able to overclock the processor 19.66% to 479.25mhz and the memory 10.87% to 344.25mhz

Beyond that I would see white specks flicking during various benchmarks and games. Benchmarks are showing about a 13-15% increase in performance but games are showing less of a boost. No issues with fans in the machine so I guess I'll keep it set there for a while.
-Jon L "

First iMac G5 iSight reports on ATIccelerator II 1.0.6a: Here's the first iMac G5 (iSight) owner reports on using v1.0.6a (released 12/9/2005). Remember YMMV and there's some risk from OC'ing.

(added 12/12/2005)
" Just to let you know, I was one of the "beta" testers for the 1.0.6a version. So far, I'm able to get my 20" iMac iSight accelerated to 18.92 percent gpu and 13.46% memory. I'm still testing at higher frequencies. One other person was able to go 24% and 15% respectively. A nice little kick in the pants for the video card :-)

(I asked Ken what the stock core/memory clocks were as I don't own one of the systems-Mike)
Right now I'm running at the following, I'm still experimenting but these values are currently stable:

    (per ATIccelerator reporting)
  • Stock: Processor 499.5MHz, Memory 351.0MHz
  • Bumped: Processor: 607.5MHz (+21.62%), Memory: 405MHz (+15.38%)
-Ken "

I'm surprised at those rates. In the past there's been inaccurately reported (very high) speeds seen with some graphics chips with this utility but the stock clocks mentioned above I'm told are correct. Kurt also reported in his page on Inside Photos of iMac G5 (iSight) and Vram info that his 20in (X600 XT) had 350MHz rated memory.

Below is the first report from a 17in (X600 Pro, not XT) EyeMac owner. (the XT is higher clocked typically and not sure what the vram chip speed in this model is.)

(added 12/12/2005)
" unbelievable! I have the 17" iMac w/iSight (yea that unbelievable cool too:). GPU is overclockable to 546Mhz before it freezes. RAM is overclockable to 36x (360+MHz I assume-Mike) before the screen goes wired. That's in both cases beyond the "XT" specifications of the 20" iMac!
Chimera showed a boost of ~15% after GPU overclock (only) - while every Unreal Engine-based game runs so smooth already - no fps drop whatsoever so far - I don't use it regularly yet.
Greets, Gwm "

Feedback on version 1.0.6 (note - for some ati chips the reported OCing speeds are not accurate - I'd write the author (Thomas) about this.)

(added 11/25/2005)
"Hi Mike
machine: dual 1.087GHz (overclocked 1GHz) MDD "Quadnostril" Powermac
RAM: 1.75GB
OS: Tiger 10.4.3
graphics card: flashed PC Radeon 9800XT 256MB card (ROM acquired from the helpful gents over at the strangedogs forums)
stock speeds: GPU: 378MHz, RAM: 337.50MHz
graphics card cooling: stock (for now)
running at: GPU: 441.00MHz, RAM: 369.00MHz (+16.67%, +9.33%)
ATIcellerator version: 1.0.6
Special note: case has been Tycho-ized (modified) for optimal cooling efficiency, see image URLs:
(Links removed after home page now longer online)

I've been incrementally bumping up the speed a bit at a time, and testing for artifacts (none so far!) using v3355 Unreal Tournament 2004 with the absolute highest detail settings at 1600x1024 (22" Cinema Display)- oh it doesn't stop just there yet, I am putting it through the paces with a 32 bot deathmatch (50 frag limit) on some of the more complex maps. I haven't tested the framerates at all, but there certainly is noticeable improvement going from the stock speed to the current speed. I must add that some of the maps are barely playable, with an estimated 2-3 frames per second in certain areas (with all those bots fragging away too!), however, I've still managed to win some matches (by sheer spamming luck) on "adept" skill level. I picked up Doom 3 recently (disappointing game) and have yet to test out the performance with the overclocking. I would expect to see a few more FPS overall.

All in all I am impressed with the overclockability of this card, and I wonder where the limits are with the stock 9800XT cooling. When I find them, I think I will get one of those heatpipe coolers that have the extra overhang in one of the PCI bays (all 4 of my PCI slots are empty) and let the 120mm fan on top of the case work some cooling magic on it, and then find some higher limits!
-Leo Porkstacker "

Here's an example of what of a clearly incorrectly reported Core clock speed with some chips/systems:

(added 11/25/2005)
The Radeon 9550 in the iBook G4 is an absolute overclocking gem :D Thomas Perrier is a saint for adding support for it.
Stock 210/183
Best 450/235 (core clock not accurate- see below-Mike)
I've played a couple rounds of UT2k4 Onslaught at that speed with no problems yet. 470 + causes hard lockups. Memory artifacts at 250. No benchmarks yet, but will be interesting to see how it fares.
-Jordan C.
(I'd report this to Thomas (author) - as others in the past have said that the (insanely high) core clock speeds reported with -some- chips are not really accurate. I doubt any chip would actually run at anywhere near 2x the stock clock rate reliably.-Mike)
(he later wrote)
Update to my previous note:
UT2k4 Santaduck pseudo-flyby benchmarks, MAX visuals:
stock vs. clocked
19.860 vs. 16.525 +20.2%

stock vs. clocked
8.117 vs. 6.363 + 27.6%

Nice increase, but I'm still not sure whether the core is really overclocking in excess of 100%. Something is definitely overclocking, however.
(another later mail said)
Well, as I suspected, the Core clock numbers are FUBAR. The memory clock seems to work though. See here:
-Jordan "

(added 11/25/2005)
"today I installed the 1.0.6 version and its the same as before: I overclocked the 9700-cpu over 100% - not even an artifact.
(note - some reported OC'ing Core speeds are not accuate, if you're seeing 2x the stock core clock rate being reliable - I'd seriously doubt it's accurate.-Mike)
I also repeated the cinebench tests. 3x in order stock frenquency of the rom and then overclocked with "some" value, always the same results: w/stock frequencys a opengl hw value of around 547CB points, w/overclock a value of around 555CB points.
(Note - The author's page has a bold note that benchmarks like Cinebench, Xbench, etc. should be used as they're not good tests for OC'd graphics chips.-Mike)
And again, the memory shows artifacts on screen when i overclock to much.

The ROM is version 113-A07525-130, it's identified as Radeon 9800 Pro with 128MB on a 256bit bus. One thing i can't say tho is, what was done to the ROM, since not me flashed the 9700 with it but i may be able to ask the guy who did.
Thx and regards, Gwm "

(added 11/23/2005)
"Hey Mike,
I just took the plunge and installed ATIccellerator 1.06 onto my Mac G5 2X2 (2003) running OS 10.4.3. I'm using an Radeon X800 XT ATI Graphics card, and so far I'm able to boost performance by 10% in Processor to 519.75 Mhz and 11% in Memory to 533.50 Mhz. I've been using this setting for the last 24 hours with no artifacts or any other apparent problems. Trying to go to 15% in the same areas causes problems with redrawing and artifacts, so I'm guessing this is about it for speed improvement for the X800 XT. This update is a marked improvement over previous versions. The info part of the program is spot on, showing the make of the graphics card, memory size, bus width (bits) and type of RAM.

Previous versions of ATIccellerator failed to come close on my G5 running OS 10.3x. A decent upgrade, and the price is perfect.
Jonathan K. "

Tiger user reports on version 1.0.5 (see above for later 1.0.6 version update.)

(added 6/30/2005)
"I just recently updated to OS 10.4.1 on my (2003) 2X2 G5 Macintosh, because Universal Audio finally updated the drivers for the UAD-1 PCI card that my audio/video studio and many other A/V studios use on a daily basis.

Everything is running pretty smoothly in OS X Tiger, however I can't seem to accelerate (overclock) my ATI X800 XT Video card anymore, like I could with OS 10.3.9. (What a difference a decimal point makes). I got this response from Thomas that oversees Graphiccelerator (rom file tweaker) and ATIcelerator II (on-the-fly overclocker).

    ME: This Graphiccelerator 1.2.3 doesn't seem to work with OS X.4.1 and my ATIRadeon X800XT-Rom version 109.

    THOMAS: You'll have a problem there: the X800 is the only ATI card whose ROM can't be dumped on a Mac, so using Graphiccelerator is problematic.

    ME: I keep getting a message that your ATI Flasher is "Read Only".

    THOMAS: Did you copy it to a hard drive? The disk image is read only?

    ME: This version doesn't work whether I'm in "Safe" mode or normal. Does Graphiccelerator have to be used or can I use ATIcelerator (II) 1.05a?

    THOMAS: Both can be used, but ATIccelerator II is easier, and safer because you can deactivate it if the card is overclocked too much, while with Graphiccelerator you'd have to flash in a PC to save it.
    -- Thomas

So, it sounds like it would be crazy to use Graphiccelerator if you own (only) a MAC. I tried using ATIcelerator (II) 1.05a with OS 10.4.1, but it's really touchy, and I can't get nearly the acceleration I was getting before. In fact, I just gave up using ATIcelerator all together.
Just thought I'd pass this along in case anyone out there have an answer as to what happened in OS 10.4 to change everything.
If anything, video should have improved with OS 10.4, not gone backwards!
-John K. "

Below is a previous Tiger/PowerBook G4 user report on ATIccellerator II

(added 5/9/2005)
"Just noticed that after installing ATIcellerator II (Using versio 1.0.5a) under Tiger, the display no longer crashes when you open the control panel after system sleep like it used to (on a Powerbook G4 1.25GHz Radeon Mobility 9600 and an iBook 1.07 Radeon Mobility 9200).
Danny S."

(NOTE - reports below were before Tiger/OS X 10.4 was released)

PB G4/9600 Mobility:

(added 4/8/2005)
"Hello, Got some very impressive results using ATIccelerator on my PowerBook.
System Specs:
PowerBook G4 (Al) 1.25GHz, 1GB RAM
Radeon 9600 Mobility w/ 64MB VRAM
OS X 10.3.8, ATIccelerator II 1.0.5a
ATI Displays v4.4.2f5 (Now supports OEM cards without RadeonEnabler!)

Test Setup:
Quake 3 1.3.2 (AltiVec optimized)
1024x768 with all settings maxed out
Built in timedemo (FOUR.DM_68)

For all of these tests I took the average of three runs and rounded the decimal.

Test 1:
Stock Speed (300/200) - no FSAA/no Anisotropic: 106FPS
OC'd +40%/+35% (420/271) - no FSAA/no Anisotropic: 107FPS

The difference is negligible and I'm guessing is the result of the 167MHz bus speed bottleneck. (1024x768 mode is more CPU bound than graphics card. Much higher res. or enabling advanced features (i.e. FSAA, AF) needed to make the graphics chip the bottleneck.-Mike) But that'll change in the next test.

Test 2:
Stock Speed (300/200) - 4x FSAA/8x Anisotropic: 69FPS
OC'd +40%/+35% (420/271) - 4x FSAA/8x Anisotropic: 90FPS (30% Increase!)

Awesome! The first test was obviously limited by something other than the video card, but in the second test the FSAA and Anisotropic filtering really put a strain on the card. I would've done more tests but... I'm lazy, sorry. Aside from those results everything else was fine, I checked it out with a couple of other games and in the OS (using expose, minimizing/maximizing) everything seems normal. Thanks for the great site, thanks to Thomas Perrier for ATIccelerator, and thanks to ATI for finally supporting OEM cards!
-Marc "

I wonder if those clock speeds are accurate - seems like a really steep overclock for the OEM Mac 9600 Mobility, although as others have mentioned in older reports, Apple's clock speeds for the 9600 Mobility were well under what the max specs were on some variations. (Reminds me of a previous PB owner report about huge overclocks/error long notes.)

Radeon 9800 Pro: (retail)

(added 4/8/2005)
"Hi, I've been using ATIcelerator to tweak my Radeon 9800, and I thought I'd share my results. With the stock cooler device and no cooling for the memory, I was able to up the CPU by about 4% (393.75Mhz) and the Memory by about 5% (355.5Mhz) before artifacts would start to appear in Unreal Tournament 2004, Doom 3, and Halo. The stock speeds for the CPU and Memory on my card are 378/337.5 Mhz, respectively.

However, I recently purchased two cooling accessories for my card to see if I could push it a little further. First, I bought some OCZ BGA copper ramsinks - small heatsinks that sit on the RAM modules on the card to cool them off. Secondly, I bought a fan that sits in a PCI slot and blows hot air out the back of the computer. Since I don't have a G5, the in/out airflow on my case was pretty weak. I thought this would help to move the hot air coming off the chip on the card outside the case. Each of these cost about $6, well worth the speed increase they afforded...

With the upgrades installed, I was able to push the card quite a bit further. I gradually upped the clock speed of the memory and CPU independently until artifacts began to appear in UT2k4, Doom 3, and Halo. I was able to increase the memory to 400.5Mhz, when suddenly severe artifacts started to appear in all three games. I got the CPU up to 409.5 Mhz in all three games, but artifacts started appearing then.

Currently, I'm running it at 405/393.75 Mhz, a 7.14% increase in CPU speed, and a 16.67% increase in Memory speed. All three games mentioned run stable indefinitely, I just spent 7 hours straight last night playing Doom 3 on it with no problems.
-"tirefire" "

Dual 2GHz G5/9600 OEM:

(added 4/8/2005)
"G5 2GHz DP
Stock Radeon 9600 (not XT model)
OS X 10.3.8

The first picture is Aticcelerator showing my card at it's current setting. The second shows what happens when you try to OC even further.
(Both PDF files he sent showed the same clock speeds - 573.75MHz for Core and Memory (way too high to run) so I asked for more info - see his later comments below-Mike)
Stock frequencies for the (Original) Radeon 9600 Pro 64 MB RAM
Processor 364.50 MHz
Memory 294.75 MHz

I did install 1.0.5, but it still says 1.0.4 in the panel... dunno why. I installed 1.0.5 about 15 min. before I wrote to you.

It's been running day and night for about four months with no artifacts or glitching (I am a heavy FPS enthusiast).
I'm using a modified ATISilencer II heat sink.
Would artifact like crazy at any setting above 21% before I added the heat sink.
(I asked what mods he did to the VGA Silencer cooler and FYI - there's a previous VGA Silencer Install Guide on OEM 9800 Pro in G5 here, as well as other cooler install guides.)
The ATISilencer II is a huge heat sink with a low-speed blower that sandwiches the videocard in a combination Heat/RAM sink. This requires an extra PCI slot for the fan vent (includes grille plate). Unfortunately, the metal deck that separates thermal zones in a G5 makes it impossible to install the sink since the Silencer has long pins to accomodate the second half of the heat sink.
I had to remove the back side of the heat/RAM sink (the radeon 9600 has no RAM on the backside of the card) and clip the pins to length so the video card would lay flat. Since the heat sink uses those pins to grasp the card, I had to fashion some spring clips using heavy duty paper clips. The clips hold the heat sink to the video card at locations where there are few-to-no circuits nearby.
Use a y-splitter from the DVD-RW for the fan's power and the included thermal grease and you're off and running!
I got mine from newegg.com for $20 (sale). The Silencer II is specifically compatible with the 9600 Pro, SE, and XT
Incidentally, OC'ing the card to it's current level induces instant garbage and lock-ups when the heat sink is not present.
(I assume he means with the stock heatsink, as the chip probably won't last long without any heatsink.-Mike)

Now, it's rock solid; response is so snappy, you can almost hear the whip crack... :)
If you have any comments, or questions, feel free to contact me!
-Dave P.
(I wrote to ask him about the actual OC frequencies since the screenshots showed much higher freq. than would be reliable-Mike)
Your comment there got me to thinking about how my ATIcelerator intall seemed to be messed up when I upgraded to 1.0.5. from 1.0.4. So I completely uninstalled and reinstalled the new version (mine reads 1.0.5a).
Now, ATIcelerator reads my card correctly. Unfortunately, I can no longer OC the card beyond +15.43/+10.69%. Strangely enough, Quake III actually runs a few fps faster than before. So I am now led to believe that ATIcelerator was OCing the card before, but due to a corrupted file somewhere, it was reading the wrong clock speed. I got suspicious when I could under clock the card to 9MHz and QIII ran the same as before.... :)

Quake III Time Demos:

Setting		Normal		High Quality		Custom (Ultra High Quality)
Resolution 	640X480	|     800X600		|	1600X1200
Average FPS	337.1	|     283.3		|    108
Spike FPS		590		|     400			|	190
Minimum FPS	200		|     150			|	50
Duration		4.4 secs	|	4.2 secs		|	11.6 secs
Filter		Bilinear	|	Trilinear		|	Trilinear

Here's some links to install guides for various graphics card coolers here from the Video card topics page:

The Video topics page also has past tests (at stock clock speeds) of X800 vs 6800 vs 9800 Pro (OEM), 9800 Pro OEM vs retail vs 9600, etc.

Overclocked Mac X800 XT:

(added 3/4/2005)
"Hey, just wanted to tell you that I noticed that when using ATiccelleratorII, that it can effect the minimizing of windows to the dock. There gets to be a slight jitter to the smoothness of it. Basically, its not smooth after using it. I was able to over clock safely to 500CPU speed no problem with no other bad effects from doing so. Games all ran just fine. And no other OS problems. Just that annoying jitter in the minimizing, and only in minimizing. Maximizing is fine. Go figure. :)
no bad anything in any game. Just that stutter in the minimizing of the windows, and only on the way down not up. I don't hear the fans run any harder either. Sounds the same to me as before the over clock. I guess I should try to up the memory as well. I just wanted to know if you or anyone else sees that happen on the X800 Cards. Also, how far Do you think I could push it? (impossible to say) The PC side is 500 right? I would only take a guess that maybe 20MHz beyond the 500 would be the max. Do I need to run the CPU and Memory in any specific way? As in if the CPU is 500 the Memory has to be 525, kind of thing. Thanks for the response and any help.
-Paul V.
Dual 1.8GHz, X800 XT, 4GB Ram
OS X.3.8, 20" Cinema Display ADC
(he later wrote)
Just to let you know, I was able to clock it up to 520/520. Everything was working just fine all day. No stuttering when minimizing or any artifacts at all in the games I play. I will try it again once I get a hold of Doom III to see if it helps any. From the looks of the benchmarks, I will need all the extra MHz I can squeeze out of it. "

Not sure how much faster you could run the memory (you may run into more problems with that sooner than OC'ing the GPU). It's really impossible for anyone to say what will be reliable with any particular sample/system. Maybe write the author of ATIccellerator to see what X800 reports he's gotten. And in the end, it may not be worth it.

B&W G3/Rage128 PCI:

(added 11/22/2004)
"I just thought I would add some information regarding this utility. I have seen moderate overclocking of cards in the posts by other people trying this utility and have seen some fairly small overclocking numbers. I have a B&W 800Mhz Powerlogix ZIF G3 overclocked to 900mhz, 576mb of RAM, and a rage 128 at 68mhz clock speed. I have added an extra case fan and cpu fan, which significantyl cooled my case and CPU down.

  • Before: CPU--68mhz Memory--68mhz
  • After: CPU--100.29hz Memory--100.29mhz

That is a 46.07% increase in memory and clock speed.
Overall, no artifacts or problems and the graphics in the system seem much speedier. Diablo II plays much smoother and I register higher scores on graphics in Xbench. I believe the stability at this high level of overclocking is due to the extra full sized case fan, and the extra CPU fan. Thanks.

As they say - your mileage may vary.

Dual MDD G4/1GHz w/OEM 9000 Pro (Failure):

(added 11/12/2004)
"Ouch! This bastard fried my card. I'd appreciate it if you could post this report, so hopefully others will be more cautious than I was. (there's a warning at the top of the page here and at the Aticellerator II home page) I used ATIccelerator II 1.0.4 on OS X 10.3.6.

The machine is a Dual MDD G4 @ 1ghz, with 1GB of RAM. The card was an Apple OEM Radeon 9000 Pro. I took it from 275/250 to 295/275, as one reviewer suggested worked well for him. After about a minute, I started getting heavy screen artifacts in the Finder. The machine had been running for days and was pretty hot, so I decided to shut it down to give it some time to cool off. (better would have been to reset the clocks back to stock the minute you saw problems, but hindsight is always 20/20.-Mike) When I started it again after about ten minutes, the light on my sony monitor flashed amber/green: unacceptable video signal.

I tried an NVRAM reset and reset button on the motherboard, but neither helped. Finally I swapped in an old Radeon 8500 I had lying around so I could boot the machine. Later I tried swapping the 9000 back in, but I got the same bad video signal result. I guess it's really dead.

Needless to say, I don't think I'll be overclocking the 8500. I'm too poor to wreck another card!
-Brandon "

If the card won't work even in another system, I guess it was damaged (ATIccellerator II does not flash the card rom per the site). Failures in general are rare (as generally people reset the clocks back when they see indications of artifacts, etc.) but remember like overclocking anything - there are risks involved and each card/chip may have different limits. BTW - I have a very early sample 9000 retail card that has faster rated memory chips than an OEM 9000 card that shipped in a Mac, but don't know if that's typical as it could have changed by production run. Also note the latest retail 9000 card (128MB ram model) has a lower core clock rate (250MHz vs 275MHz) than the original 64MB 9000 model. (Also always start with small boosts, not large ones and don't assume you can repeat other's results. If you see any problems, reset back to the stock clock speeds.)

MDD Dual 1GHz/OEM Radeon 9000:

(added 9/21/2004)
"ATIcellerator II (ATI Radeon Overclocker) 1.0.4 Update
Powermac G4 Dual 1Ghz MDD
OEM Radeon 9000 (with dual monitors)
1.25GB RAM, MacOS 10.3.5

No matter what I set the CPU and/or RAM to, I would get ugly screen artifacts when playing games such as Unreal Tournament 2004. I think maybe running dual monitors dramatically effects the results of this software. At first I tried a setting of 299.25Mhz/290.25Mhz since I read that someone else with the exact same computer had good results. (never assume you can repeat other's results and never start with large boosts in speed.-Mike) That was not the case for me. UT2004 was unplayable. It was like looking at the game through purple horizontal AND vertical blinds. Both screens would show these ugly artifacts, even though the second screen was black during game play. I lowered the settings several times and all it did was reduce the amount of purple artifacts. I never could get rid off all the artifacts though. So, back to the default clock rate I went. Oh well...
-Jason "

(Note: reports below are from v1.03 or older)

(didn't include system, OS details or utility version used)

(added 9/21/2004)
"I have a Sapphire 9800se 128mb, infineon RAM (hw mod'd ROM chip), with Arctic Cooling's VGA Silencer and RAM heatsinks. i can run steadily at 15% OC. the performance increase is obvious all the time, on all apps, even the finder. As far as I'm concerned the cooling mod and the OC are a success! Oh ya, with the VGA Silencer and RAMsinks, i can OC 50% more than with stock cooler. I advise getting some better cooling to OC a 9800.
-ron r "

MDD Dual G4/1.25GHz:

(added 9/20/2004 from 9/17 email)
"On a stock Powermac MDD 1.25GHz DP w. Radeon 9000/64 MB running 10.3.3 I was able to overclock the card from 275/250 to 295/275 MHz using ATIccelerator 1.03 . It works stable and without any artifacts and lockups. Here are the results with some games that I've tried it with:
Quake 3: demo four - 1680x1050 max details 8xAF
              275/250    - 73.5 fps
              295/275    - 80.1 fps
RtCW: demo0001 - 1680x1050 max details 8xAF
              275/250     - 37.5 fps
              295/275     - 40.1 fps
Halo: timedemo - 1024x640 medium details
               275/250     - 23.2.fps
               295/275     - 23.5 fps

UT2003: Antalus flyby - max framerates are about 10% higher when card is overclocked, but overall score is about the same (botmatch sees no difference, just like in Halo). (too CPU bound)

I would like to point that although when I check "Set custom frequencies at startup", it doesn't work when I restart the computer and it doesn't work when computer is woke up from sleep either. I need to open the preference pane to make the changes active. Other than that, I have no other bugs to report. Maybe application specific profiles would be nice, but anyway this is just 1.03 version of the program..
Miki "

Version 1.0.4 (see above) is supposed to fix some bugs.

B&W G3 w/G3 900MHz CPU Upgrade: (strange report on better performance when underclocked)

(added 9/20/2004)
"Hey Mike,
Thought I would pass this on since it surprised me.
First Smurf stats:
System:     B&W G3(Smurf) with PL G3/900Mhz ZIF (originally a 450Mhz G3)
Memory:     1Gigabyte
I/O:        ATA/133 SIIG card
OS:         MACOSX 10.3.5 with PCI Quatze Extreme enabled
Monitor:    24" FW900 @ 1600x1024x16
            17" Smurf display
Graphics:   ATI flashed 9100 Radeon 64Mb in 66Mhz slot
            ATI 7000 Radeon 32Mb in 33Mhz slot

I fired up ATIccelerator II after reading some of the feedback to see what I could get it up to (Thinking worst case I burn the card and be forced to get the new 9200 :) Well to my surprise the biggest perfomace increase came from decreasing the clocks. ATIccelerator says my Stock Frequecies are 249.75Mhz GPU & VRAM, I played around with it for a bit using chimera 1.23 as a gauge and with all of Scenes there was a drastic improvement at 150.75Mhz GPU & VRAM. ParicleField-Suite OverlayEdition is one of the more noticable Scenes that benifited from the decrease in Mhz. This is 39.64% drop form the Stock Frequency.

It makes me wonder if the ROM I flashed it with frequency was too high. Either that or some benefit to the PCI bus with the lower clock.
Dale "

Strange... strange - v1.04 was released this weekend and fixed some bugs/added 9200 card support. might give that a try.

iBook G3 800/Radeon 7500 Mobility:

(added 9/8/2004)
"Running on: iBook G3 800MHz (overclocked to 900mhz with iCook) (see previous systems page article on iBook/750FX software overclocking) with a Radeon Mobility 7500 32MB

Results: I got it all the way up to 50% (270/270) with just very few minor artifacts, but it is pointless as there is no performance increase in ANY app I've tried (Quake3, UT2k4 demo, Let1kWindowsBloom, XBench, etc.) Actually, it has only given me 1 fps LOWER than at its normal 180/180. I know its working because once I hit 60% I can't see the screen anymore. Obviously a Mobility Radeon 7500 can't use the clock speed given it at all.
John-Michael "

Powerbook G4/1GHz - Radeon 9600 Mobility: (another report of an error where the settings were not actually made)

(added 9/8/2004)
"On a 1Ghz PB G4 15" Al (Radeon 9600 Mobility), I was able to do HUGE overclocks with no effect. Checked the error log, and got the following message.

2004-09-07 17:07:14.084 System Preferences[358] ATIccelerator: error
10000003 writing to the I/O Registry

Apparently 1.0.3 doesn't work on this powerbook.
Cheers, Brett B. "

Beige G3/Radeon 7000 PCI:

(added 9/7/2004)
"Beige g3 running 10.3.5
Radeon 7000
768MB ram
ATIcellerator 1.0.3
Problem: installer causes kernel panic
Regards, Svein-Inge "

PowerBook G4/1.5GHz - Radeon 9700 Mobility (128MB):

(added 9/7/2004)
"I have a 1.5Ghz PowerBook 15" and it has the Mobility Radeon 9700 BTO with 128Mb VRAM. I can confirm what the reader said about having a system crash and a screen full of artifacts after sleeping the system (in earlier reports below) and reawakening it and opening the control panel.

The version I tested was 1.0.3. It also seems that the video processor overclocking also does not work or has little effect on OpenGL performance (Benchmarked using OpenMark) I stopped testing after I had the crash and tried to find out why it was happening, but the memory overclocking seemed to improve performance. If the crashing is fixed I will probably send some figures from some better benchmarks (i.e. Halo timedemo).
- Joe "

Quicksilver G4/867 - Radeon 8500:

(added 9/7/2004)
"Hey, everybody. I've got a Quicksilver 867 with a retail Radeon 8500. Ever since I got the card, I've always felt the performance was less than it should be, so I started using ATIccellerator (currently 1.0.3 on OS X 10.3.5). At first I tried upping both core and memory clock rates the same percentage, and got up to around 10% before I started getting stuck colored pixels or blocks/arrays of discoloration onscreen. This caused me to change my tactic for two reasons.
First, I believe these artifacts to be caused by *texture memory* corruption due to too much heat, so 10% is too much for the memory chips. Second, I looked around online and saw that most stock PC Radeon 8500s have a core speed of 275 MHz, but ATIccellerator reports the Mac card only having a 250 MHz core by default (memory is 275 MHz as it should be).
Through multiple tests, I've now settled on clocking core/memory both to 294.75 MHz, which corresponds to 18.02%/7.38%. This setup is free of artifacts except under the highest of graphics loads. I once tried overclocking just the core, up to 20%. This seemed to work fine until I suddenly got a hard crash (though, strangely, I could still move the cursor).
Anyway, I would suggest to people experimenting with this program to try adjusting the memory speed independently of the core. Try taking the core as high as is stable, and then upping the memory until just before you see artifacts under high stress. This is probably especially true for 8500 owners, due to the factory underclock. I've seen slight performance increases in benchmarking programs and games, but nothing spectacular.
Aaron "

Sawtooth G4/AGP w/Radeon 8500:

(added 9/7/2004)
"Hi Mike, I have a 500MHz G4 Sawtooth with a Radeon 8500 running 10.3.5 and used version 1.0.3. I think my Radeon was one of the last ever made - I bought it from OWC nine months ago.
To test the stability of the various speeds I used the ATI demo screensaver called "ATI Tubes" which I think pushes the cards harder than just about anything else. I found that upping the processor too far created flickering in isolated areas of the screen almost like lightning, whereas RAM problems caused random pixels of the wrong colour, mostly blue.
After rearranging my PCI cards to give the Radeon as much airflow as possible, I was able to push the processor to 315MHz before faults occurred and the RAM to 310.5MHz. I've now got them both set at 306MHz and haven't had a single problem. I assume that having them both at the same speed means there will be no wasted cycles. The ATI Tubes screensaver certainly runs a bit smoother now.
James "

Powerbook G4/667 - Radeon Mobility (16MB): (updated to include 1.0.3 notes)

(added 9/7/2004)
"ATIccelerator II Report
TiBook 667
  Tipo:    display
  Bus:    AGP
  Slot:    ATI
  VRAM (Totale) :    16 MB
  Fabbricante:    ATI (0x1002)
  ID dispositivo:    0x4c59
  ID Revisione:    0x0000
  Revisione ROM:    113-XXXXX-115

I don't wanna report any performance tests but compatibility. I could speed up the GPU and memory up to 15% with no problems for long times and surely I noted that graphics were faster. The problem is that I use the TiBook with an external monitor as a desktop workstation, and when I restart, power on or wake up from sleep mode and I forget to put down the top-screen the laptop goes in 2 screen mode. At this point the graphic card goes in tilt and everything is 1000 times slower. It seems the chip produces two much heat or simply has problem to handle 2 screens with bad parameters. Even an accelleration of 1% hangs the laptop. If I restart in 1 screen mode everythings goes well. The author should add an option to disable accellerator if 2 screens are handled by the G-card.
Problem happens indifferently if system is set to use the 2 screen as desktop extension or screen duplication.
Verified for version 1.0.1 and version 1.0.3 of ATIccelerator II
giulio "

PowerBook G3 (Pismo)/Rage128 Mobility (8MB):

(added 9/7/2004)
"Machine Data
  • PB G3 2000 (aka Pismo) w/G3 900 CPU Upgrade
  • 512 megs ram
  • 15GB Toshiba HD, 8 meg cache

Installed accelerator with no kernal panics and have been running for about 1 1/2 hours without any problems.

Could not get Chimera and OpenMark to run so I down loaded Cinebench (2003) which did run with the following results:

ATIcellerator Processor/Memory Processor/Memory 
Settings 140/124 105/105 

C4D 93 99 
SW-L 191 193  
HW-L 119 124 

Open GL Speed 2.06 1.95 

Hardware Lighting 

Scene 1 65.35sec/1.6fps 63.4/1,6 
Scene 2 78.79sec/1.161fps 72.36/1.2 
  Polygons 77,898 81820 

Software Lighting 

Scene 1 47.95sec/2.1fps 49.03/2.1 
Scene 2 30.97sec/2.9fps 28.57/3.2 
   Polygons 198,177 214,825 

4D shading  

Scene 1 114.98sec/.09fps 106/1.0 
Scene 2 48.63sec/1.9fps 46.16/2.0 

  Polygons 126,209 

Single CPU 318.5sec 293.2sec 

The machine subjectively seamed faster on graphic heavy pages prior to running the tests. I know very little about what all of these numbers mean but in total it seams to indicate a slightly faster response which I would say is about 6 to 8 faster.

On the Pismo settings above 150 for the processor and 126 for memory results in instability to the point of triple sceeens or a black screen with blue vertical lines about 1/4 inch in width. So I backed the setting off to 140/124 respectively.

The preference pane does not hold processor settings on shutdown, restart, or fast user switching.
Regards, Wil N.
he later wrote)
After 3 or 4 restarts the accessing the ATIcellerator preference pane would cause multiple screens and running verticle line so I removed ATIcellerator from Preference Panes and System...extensions....
Prior to this I had been using it at 130 mhz for both the processor and memory settings. All-all I feel minor speed increases not worth the effort and potential harmful effects on machine. "

Dual 2GHz G5/OEM 9600 Pro (64MB):

(added 9/7/2004)
"Hey Mike, I've been using ATIcellerator II 1.02 (see no reason for me to run 1.03) since you first posted it, on my G5 Dual 2GHz Mac with 3.5Gigs of RAM running OS 10.3.5. ATI Radeon 9600/64MB
I'm running at a 25% increase with 369 mhz Memory and 456.75 mhz Processor speeds, over the last few days with no signs of any artifacts. Don't know what all this means, but my refresh seems a little snappier. Doesn't seem to interfere with anything (Music Apps -Logic Pro & Cubase SX), Browsers (Safari, Explorer, & Netscape), GraphicConverter, and my monitor is responding normally using 1280 X1024 with 75hz refresh rate.
Maybe I'm just the lucky one.
John K. "

Dual 2.5GHz G5/9600 XT (128MB):

(added 9/7/2004)
"Hi mike, I'm working with an ATI Radeon 9600XT/128 MB in my G5 Dual 2.5GHz, 1.5GB of RAM
I Overclocking 25.28 percent for the GPU (501.75Mhz) and 15% for memory (357.75Mhz), it's seems work fine on my G5, but I don't know how to check if the card is running faster than before..
-micmic "

Popular ways to test are running a 3D game (Quake3 timedemos are often sensitive to performance changes) although other tests like Cinebench2003 and Xbench some said didn't really show any improvement. (not a big fan of Xbench's graphics tests).

Powerbook G4/1.33GHz - Radeon 9700 Mobility:

(added 9/7/2004)
"I've just used this app and at the moment my mobility 9700 64MB (1.33Ghz Powerbook) has been clocked to 432Mhz for the core (10% higher) and 230Mhz for the memory (13% higher) so far no problems at all.

However, using Xbench, the scores in opengl, quartz etc have not improved, there's only 0.5-2 difference between all the tests when comparing the stock speed to the settings I bumped the core and memory up to. This is using 1.0.3. Using 1.0.2 resulted in an array of lines taking over my desktop and needing a restart. So 1.0.3 doesn't crash my Powerbook, it simply doesn't work it seems.
Matt "

Powerbook G4 1.25GHz - Radeon 9600 Mobility:

(added 9/6/2004)
"Hi Mike, I think your PowerBook G4 readers (at least those with Radeon 9600/9700 Mobility cards) should be aware that there is a bug in v1.0.2 (haven't tested 1.0.3) which hard-crashes your Mac with total screen-artifacting if the ATIccelerator prefpane is opened anytime after wakeup from sleep.

This happened to your reader Galen with a 9700, and I can consistently repeat this with my 9600 (15", 1.25GHz, etc). It might be tied in to some sort of built-in GPU throttling after sleep, and I have notified the developer who couldn't replicate this on his G4 MDD, which is of course a different setup altogether.

In the meantime I shall refrain from using ATIccelerator. It would help if other laptop users could test for this behaviour and send their panic.log file (Library\Logs) to the developer, as this is what he advised me.
Sailesh P.
(I asked if he had tried 1.0.3 but that may not fix this from the comments from the author.)
Just a quick note to say that I've tried out 1.0.3 on a clean (default) user account on my computer and it crashes just the same. BTW, I'm running OS X 10.3.5.

Couldn't find a panic.log file due to the total hard crash, therefore I guess this is pretty useless advice for others. I guess it's up to the developer to work through a fix with some beta testers if poss. BTW, top work with the site! "

Dual 2GHz G5/Radeon 9600:

(added 9/6/2004)
"Hello again Mike, I just wanted to report in with a newly added BFG fan to my Radeon 9600 in my 2GHz G5. I bought this fan thinking I would get a few more MHz out of the card over the stock passive cooler. It was overpriced compared to the other things you can buy-$20 for just a core cooler. I had it at 477MHz core 380MHz memory (disregard my previous post) with ATIcellerator V1.0.2 After installing this fan, I found it to be absolutely pointless; the card could not even run 2 mhz higher. After looking at voltage mods I thought maybe apple has the voltage at something lower than 1.4v (found on good overclockable pc radeons). Also I noticed that there is a substantial amount of heat rising from the processors to the bottom of the card where an optional heatsinks can be purchased.
Hope this helps someone, Noah H. "

Powerbook G4 1.25GHz/Radeon 9600 Mobility:

(added 9/6/2004)
"just tried ATIccelerator II on my 1.25GHz rev. A 15" AlBook... Needless to say, the CPU and bus speeds are limiting what the 9600 Pro (radeon 9600 mobility) in this laptop can do. I cranked the settings for CPU to 400MHz and memory to 275 MHz, with only marginally higher results in CineBench.
Cheers, Martin
(I asked what version he used and if he noticed any problems after waking from sleep.)
It says it's 1.03, so yes, that's what I use. No wake-from-sleep problems.
During experimentation I got, for want of a better word, a snowcrash, but that was when I was really pushing the memory timing. It happened only once, so I guess it's alright.
In conclusion, however, there's no benefit for what I use the machine mainly for, so I will uninstall this later on. _After_ I made a backup ;-) "

B&W G3/Radeon 7000 PCI:

(added 9/6/2004)
"ATIcelerator II is a great find for me. Not for over-clocking, but for underclocking.

I bought a PCI Radeon 7000 (64MB) on eBay to let my variously upgraded Blue G3 (500Mhz G4) use Quartz Extreme. QE works great but in 3D accelerated games and movies there was a lot of annoying green static. I was trying to work myself up to re-selling the card on eBay and finding another one when ATIcellerator came out. By *underclocking* the card's VRAM by 10% (down to 1Mhz), I can eliminate all the static and watch movies in all their glory. Re-clocking the GPU has no effect.

I haven't done any benchmarks since this Mac is pretty useless for games already, but this is a new and helpful benefit, at least to me. BTW version 1.02 shows my old Rage 128 as well, which the previous version did not.

The "Set custom frequencies at startup" checkbox doesn't seem to work, but restarting is pretty rare these days anyway.
-Paul M. "

Just for the record I asked what OS version he's using (10.3.x required, but please include OS version used as driver versions vary.)

Quicksilver G4/800 - Radeon 9800 Pro 128MB (retail):

(added 9/6/2004)
" Mike, Using a Quicksilver 800Mhz x 2, 1+GB memory, Mac OS 10.3.5, Radeon 9800 Pro (retail, recently replaced the stock Geforce2), and ATIcellerator 1.0.2. Software installed ok. I tried increasing both processor and memory by 5, 10 & 15%. Oddly enough, no matter what settings I put into it, cinebench 2003 and Xbench 1.1.3 report no change in performance.

But, playing Halo, there was a noticable increase. At 15%, there was too much noise in Halo, so I backed it off to 10%, which has been running fine all evening. I'll probably run it for a few more days before making it 'permanent'. By the way, great site, keep up the good work!
Ok, a quick update.... With FSAA set to 4x in Halo, there was way too much noise, had to back down to only 5%. But with FSAA off, I was able to go as high at 10%.....??? "

FSAA probably stresses/heats the chip more, so artifacts are more likely to show when OC'ing.

MDD Dual G4/1.4GHz - 9700 Pro (OEM):

(added 9/6/2004)
"I can confirm that it doesn't work in my system. Mine is a Powermac MDD 1.4GHz DP with 1G RAM and ATI 9600 Pro AGP. (he later said it's a 9700 model)

I simply check the fps frame rate of Marine Aquarium screensaver. The un-accelerated rate is about 240-250 fps.
When either the Processor and/or the Memory is accelerated, the rate drops to 200 or so.
FYI, Kenny
(I asked what OS version and ATIccellerator version was used. )
The GPU is listed as RV300.
It is the original Apple OEM 9700 Pro AGP, not modified whatsoever, that came with the Powermac which is running OSX 10.3.5.
I tested the latest version 1.03 again. There is no change in video performance. FPS now remains at around 240-250 whether the frequencies are increased or not. "

Reader FYI - see the Video topics page, graphics card section for my spring 2003 article comparing performance of the OEM 9700 Radeon vs the Geforce 4Ti, Geforce 3, ATI 8500 and ATI 9000 cards in the same system.

Sawtooth G4/AGP - Rage128 Pro:

(added 9/6/2004)
"hi mike.. i've tried the ATIcellarator II (1.0.2) (1.0.3 is out as of the past weekend), and to my worst nightmare, it basically "shrunk" the viewing area on my monitor! (normally that happens if the card refresh rate is set higher than the monitor supports)

here's what my setup is...
sawtooth G4/600 (oc from DA 533Mhz)
768MB RAM/ ATI Rage 128 Pro (ATY,Rage128Pro)
monitor 17" Philips 107T5 (1152x864@75Hz)

my video card was running at 108MHz processor / 64.24Mhz memory before any accelleration was made (i believe this is not the stock setup because i have used the previous version of ATIcellerator before)

this is my recollection from what i have done last night:
i've increased the memory speed to 75Mhz or around that value until i saw vertical lines appearing on the monitor...
decrease it to the maximum possible without having any vertical lines.. and then i've slowly increases the processor speed... until suddenly the monitor goes blank!

in panic, i've quickly resetted the computer and when it boots back again, the image that appears on the monitor appears like this --> (_ ) it only covers about 70% of the monitor screen and i have to play around for 5-10 minutes to get the best setup (which is still very bad)

i attach the screenshot of my monitor for your information
the pictures are factory settings (pict0881.jpg) after the overclock and (pict0882.jpg) is my best attempt to fix the situation..
the image usually fills the whole glass screen before..

hope this become a reminders to others... there's no else to blame other than me :-(

by the way, i've run techtool pro 4 test on the video memory, and it reported everything as usual (nothing broke)
-salah- "

1.0.3 may not help with his problem (no other reports about that problem that I remember though) but with a rage128 pro, I don't think it's worth Overclocking that card. I'd try deleting the display prefs perhaps and clearing the nvram, etc. as a last ditch if nothing helped. I'd also try booting from CD to see if the problem is still happening.

G5 1.6GHz - Radeon 9600 Pro (OEM):

(added 9/6/2004)
"Hey, I just read about the ATIccelerator II, and decided to give it a try (v1.0.2). My machine is a PowerMac G5 1.6 GHz, 1280 MB RAM, with a Radeon 9600 Pro (64 MB... *sigh*), running OS X 10.3.5.

Managed 486 MHz GPU speed (33.33% overclock), and 378 MHz memory speed (28.24% overclock). I actually could go just a little higher, but I wanted some margin to speeds that gave artifacts.

Increase in FPS (tried Halo and UT2004) seem really good. Not quite 1:1, but pretty close.
Great little utility this is, good work Thomas!
-Anders "

General Comment:

(added 9/6/2004)
"I'd just like to comment as to a possible reason why some people maybe experiencing errors in regards with clock frequencies. In the PC realm, the Radeon 9600XT and 9800XT have a feature called OverDrive and when enabled, will overclock the cards for the user. However, OverDrive will also scale back clock frequencies if things get too hot. The normal, Pro and Mobile versions of the 9600 and 9800 all have OverDrive functionality built into the core but appear to be disabled.

The Mac versions of these cards have the support in hardware since the cards use the same chips as PC versions. However, I have seen no indication that OverDrive is enabled, at least for overclocking. As a safe guard, ATI may use the OverDrive feature to dynamically scale back clock speeds on the Mac cards.

I don't know if that feature is enabled in the Mac rom and from some of the reports (noting increased performance, or overclocking artifacts from too high a clock speed) I don't think it is on the Mac. (I'll try writing a contact/programmer at ATI to ask though.)

iBook G4/933 (14in) - Radeon 9200 Mobility:

(added 9/3/2004)
"I just thought I'd throw my two cents in on this interesting little utility. I have a 14" iBook G4 933, stock configuration except that I've beefed up the RAM to 640 megs, so it's a Radeon Mobility 9200 card. I'm using version 1.0.2 of the ATIcellerator II.

I decided to use Cinebench 2003 (available here) for benchmarking, as I don't have any of the usual FPS games available (this is supposed to be my "work" machine). The scores listed are averages of three run-throughs using the "Open GL SW" and "Open GL HW" tests; the "C4D" (processor rendering only, I guess) score was identical at 102 CB at all acceleration settings.

"Stock" config (reported as Processor 183 MHz, Memory 190 MHz) 
* Open GL SW: 231.7 CB 
* Open GL HW: 309.3 CB 

Clocked to Processor 201 MHz, Memory 201 MHz: 
* Open GL SW: 232.3 CB 
* Open GL HW: 316.3 CB 

Clocked to Processor 217 MHz, Memory 217 MHz: 
* Open GL SW: 233 CB 
* Open GL HW: 317.7 CB 

With the processor and memory both clocked to 237 MHz, I got a slight increase to 235 and 322 in one run-through of CineBench; a second attempt at running the Hardware Open GL test resulted in the screen wigging out and the computer freezing (I had to do a hard reboot, and the screen kept showing artifacts until I reset the frequencies back to normal even after the reboot), so I guess I had surpassed the available limits.

While it looks like the gizmo actually works, I only really got any improvement in this particular benchmark when doing the initial acceleration from 183/190 to 201/201; the subsequent bumps didn't seem to help. I'm slightly paranoid about this machine, so I doubt that I'll mess with it too much more. As always, thanks for your site and all the work that you do.
Regards, Brent O. "

iBook G4/800 - Radeon 9200 Mobility:

(added 9/3/2004)
Here's my data for my 12" iBook G4 800 with a 32MB 9200 Mobility, 768MB RAM, and 10.3.5. I was using the 1.0.2 version of the software. Room temp was about 72 F. Benchmarks were done using Openmark in windowed mode, all applications closed except System Preferences with the ATIcellerator pane open. I just had my logic board replaced last week, so I pretty much ran it up until I hit artifacts where I had to restart my computer (they covered the screen). Please don't include my name in case I need to end up sending the iBook back in for another LB replacement, however after nearly 24h of use I haven't had any problems. You might want to re-format this next part so it looks better on the site. (I asked if he had a spreadsheet to take a screenshot of this data in a table format - too swamped here currently to reformat it into an HTML table)

Processor Clock:  MHz(%), Memory Clock: MHz(%), Openmark FPS 12Tri, 
Openmark FPS 1200Tri, Openmark FPS 7500Tri, Fan Operation
183.38(+0%), 190.13 (+0%), 525, 283, 91, Normal (off)
194.63 (+6.13%), 199.13 (+4.73%), 555, 299, 91, Normal (off)
201.38 (+9.82%), 205.88 (+8.28%), 570, 310, 91, Fan turned on at slow 
203.63 (+11.04%), 214.88 (+13.02%), 595, 321, 92, Fan speed increased 
and is much more audible
214.88 (+17.88%), 232.88 (+22.49%), 629, 331, 92, Fan on full
219.38 (+19.63%), 241.88 (+27.22%), 647, 330, 92, Fan on full

221.63/241.88 created artifacts and required a restart. After letting the computer sit for a while, I tried to up the frequencies to, and then past these marks however it appears that this is the top cap for the card.

I've been running at 221.63 (+20.86%)/239.63 (+26.04%) for the past day and it runs fine. I chose these final settings because they have the most minimal fan usage of the upper frequencies, and they are the closest together ratio-wise of them.
Let me know if there's any more info you need.
(name withheld by request) "

Dual 2GHz G5/Radeon 9800 Pro 128MB (OEM)

(added 9/3/2004)
"Hi Mike,
Rev. A Dual 2.0GHz G5 w/9800 Pro 128MB (OEM) and 3GB RAM:
I increased processor to 14.10% = 400.50 MHz, and memory to 10.42% = 357.75 MHz. I mainly play RTCW, and this has had no problems.
Play is smooth as before and I have seen no artifacts like other 9800 users have reported. Maybe the difference for me is that as soon as I replaced the stock ATI heatsink/fan with the Arctic Cooler, which has a 2-speed large slow speed fan and a long heatsink encased in plastic that is designed to blow air over the heatsink and out the back of the computer (unlike the stock heatsink, which just blows the hot air back into the computer). It is so large it fills the adjacent PCI slot space. I paid $20 for it last fall but earlier this year have seen it for $10. It is a great cooler and has 2-speeds, so I leave it on the lowest which, the manufacturer claims cools better than the stock ATI cooler (and I believe it). So, that may be why I have been successful at running 14/10% with no problems. I really don't care to go much higher, but am glad to squeeze extra performance out of a great video card.
BloodHound "

The "dentist drill" noise of my OEM 9800 Pro's fan was more than I could take, so last year I posted a guide to Verax's G03 (Quiet) Cooler Install on 9800 Pro. Much more expensive than the artic cooler though. (I posted info on that cooler last year in the news page, as well as other alternatives, but many need mods to install in the G5.)

G5 1.8GHz/Radeon 9800 Special Edition (256MB)

(added 9/3/2004)
"I can't remember which version it was, but I downloaded it on Monday...
Running 10.3.5 on a G5 1.8GHz (Single CPU), 1GB RAM, Radeon 9800 Pro SE (256MB)

Was able to (mostly) successfully OC to 10% (both RAM and GPU) without any major artifacts. I ran the Halo timedemo, which showed roughly a 10% increase (which I would expect) on avg FPS (21 to 23.5 approx).

At 15% level, artifacts were clearly present on games.
At 5% level, games work fine, and no visible artifacts.
Ryan L. "

iBook G4/800 - Radeon 9200 Mobility - UNDERclocking Report (from a reader that asked about saving battery power by reducing clock speeds, although other components like the LCD backlight is a much higher drain.)

(added 9/3/2004)
"I just tried underclocking with interesting results: (ibook g4 800 with mobility 9200)
I can underclock memory by 50%(94 mhz) and chip by 25%(135 mhz) before getting artifacts in Mac os X GUI. The slowdown works as expos and the spinning cube are much slower. normal operation like web browsing email and word processing, file management seems as fast as normal. DVD PLAYBACK is smooth with very minor artifact (hard to notice) so maybe its only a matter of tweaking the underclocking to get it perfect .

this is on dual screen setup with the NVRAM hack on single screen I seem to be able to lower mem speed by an additional 15% to -65% and chip speed by 15% to -40%

the most I could get (but could not always reproduce) without artifacts was cpu -50% mem-65%

the second screen in mirror mode did not like the tweaking and displayed strange lines before I rebooted (resetting speed in the panel did not solve the problem nor did a disconnect-reconnect of the video cable, a reboot was needed)
the main screen on the ibook was fine of course.
Alexandre "

PowerBook G4/17in - Radeon 8600 Mobility:

(added 9/3/2004)
"Hi Mike, I tried the the ATIcellerator II 1.02 on a pb 17" 1.33 w radeon 9600M, but I didnt see any improvement in xbench. I speeded both values +30% in the prefs pane...
See the tests for u self...
Mr O Rossini "

Please don't sent xbench file attachments (just summarize/include the opengl/graphics scores in the mail). And if possible I'd use something other than Xbench (for reasons I'm too tired to repeat :-)

Powerbook G4/400 - Rage128 Mobility:

(added 9/3/2004)
"ATIcellerator II v1.0.2 runs just fine on my TiBook (400 MHz). Performance boost at 140 MHz GPU/121.5 MHz memory ranges from 5% to 25% depending on benchmark. Speeds above 121.5 MHz for VRAM resulted in shimmering pixels somewhere on the screen.

I was able to achieve GPU speeds as high as 243 MHz (twice VRAM speed), but performance was best in the 140 MHz range. Weird but true. More on this on Low End Mac soon.

I just tried it on my eMac 1.25 GHz -- no dice. Every time I open it, it messes up the screen. If I hit the default speed button, I can go on working.
Dan K.
http://lowendmac.com "

PowerBook G4 15in/1.33GHz - Radeon 9700 Mobility:

(added 9/2/2004)
"Hi Mike,
Take a look at this forward of a message I wrote to the developer of ATIccelerator II. Basically, at least for me, it doesn't work quite right. I'm not writing to beat up on the author or anything (hey, at least the price is right on this software!), but I think you might like to note those XBench scores and that "limit" I seem to be hitting on your reports page.
(I asked if he was using v1.0.2 )
Yes, 1.0.2. Also, I just (couple minutes ago) opened the prefs pane and it scrambled my screen! Could see blocky thing that represented cursor, nothing else. Had to reboot. No overclock settings were applied (0.00% on both) when I opened it at that time. Only thing I had done to the machine was let it sleep for 8-9 hours overnight and when opened, the screen scrambled. After rebooting, prefs pane opens fine.
(Also seeing that same console error as mentioned in reports
2004-09-01 23:49:22.235 System Preferences[447] ATIccelerator: error 
10000003 writing to the I/O Registry]

Begin forwarded message:
From: Galen Z
To: thomas perrier
Subject: Radeon 9700 Mobility Report

Hi Thomas, I've been using ATIccelerator II under my PowerBook 15" 1.33 GHz, OS 10.3.5, with Radeon 9700 mobility and I think I've found a bug. It seems that there's a very low "limit" as to how much acceleration will happen. My card is displayed as "Built-in" and "ATY,RV360M11" within the ATIccelerator II PrefPane.

Stock performance on XBench is ~160 for Quartz and ~95 for OpenGL. Increasing processor and/or memory can increase this to ~180 and ~108 - a small but decent improvement. However, no matter how high (even over 1 GHz!) I make the processor, I cannot improve that number. There are no artifacts, no defects, no freezes, nothing. I can restart, fix permissions, open/close the PrefPane, etc. There is some sort of "limit" here. Equally, I cannot get more a small gain from memory performance, no matter how insanely high (hundreds of percent increases!) I set it.

Also, there seems to be a GUI bug with % numbers randomly becoming negative at higher MHz/% values, especially when you click on the other performance adjuster (i.e. memory goes negative when processor speed clicked, vice versa). I don't know if this is visual or indicative of something else, but even when extremely negative, XBench is not affected. The MHz being displayed does not change. The value shown seems nearly random with no apparent relationship to anything else. (No, I haven't looked that hard, but then again, it's late and I don't have your source code). And there are two other things: on my system, I use "0.00 %" as the default format for percentage, but your program shows "0,00 %" until I adjust the percentage up/down, but whenever I hit zero again, the comma returns. Finally, there is one additional percentage bug: 1125.00 MHz maps to "+40.23 %" (totally wrong!) yet 1122.75 MHz accurately maps to "+186.78 %".

So basically, I think this program isn't working properly on my machine. Any suggestions? If you have a beta version you'd like me to try out, please let me know and I'll give it a whirl!
Galen Z. "

G4/PCI (Yikes)/Radeon 7000 PCI:

(added 9/2/2004)
"This is a follow-up to my report of yesterday:
Yikes with an OWC Zif G4 upgrade to 600 MHz (system bus still 100 Mz) and a pair of PCI Radeon 7000s.
I'm running Quartz Extreme via PCI Extreme.

Xbench Video tests:

Slot 11 PCI Card:  Accelerated 20%/Stock

Quartz Graphics Test   46.06/45.80
OpenGL Graphics Test   81.56/85.38
User Interface Test           64.43/63.03     


Cinebench 2003 Tests:

Slot J11 PCI Card:  Accelerated 20%/Stock

Rendering (Single   CPU): 59/59 CB-CPU
Shading (CINEMA 4D)                : 72/72 CB-GFX
Shading (OpenGL Software Lighting) : 192/192 CB-GFX
Shading (OpenGL Hardware Lighting) : 105/107 CB-GFX
OpenGL Speedup: 2.67/2.66

Slot J12 PCI Card: Accelerated 20%/Stock

Rendering (Single   CPU): 59/59 CB-CPU
Shading (CINEMA 4D)                : 75/74 CB-GFX
Shading (OpenGL Software Lighting) : 111/111 CB-GFX
Shading (OpenGL Hardware Lighting) : 75/75 CB-GFX
OpenGL Speedup: 1.49/1.49


I interpret this result - no significant speedup - to likely mean that the cards are already limited by the PCI bus constraints and acceleration of the card cannot be seen past this bottleneck.
Regards, Larry S "

iBook G3 900/Radeon 7500 Mobility:

(added 9/2/2004)
"iBook G3 900MHz/Radeon 7500 Mobility
It says 180MHz for memory and GPU speeds, if i downclocked it to around 30MHz, i got a black distorted screen image, the 1.0.2 update did fix a lot of problems, this is a good app. "

Dual 1.8GHz G5/9800 Pro 128MB (retail):

(added 9/2/2004)
"I tested it on my retail 9800 pro 128mb in a dual 1.8GHz G5 (revb) increased 10% on both and had artifacts on the cinema display while playing UT.
increased 5% on both, still artifacts.
Less is pointless to me.
Put it back to factory settings, all works as before.
(I asked for more info)
ATIcellerator 1.0.2, and 2 monitors (both apple) connected. I might also add that one PCI is filled with the radeon 7000 "

MDD G4 Dual 1.25GHz/Radeon 8500:

(added 9/2/2004)
"Hello mike, I just tested version 1.0.2 an my G4/1.25 Dual, 2 GB RAM, Radeon 8500/64MB. I passed several tests with xbench:

No overclocking: 

    Quartz Graphics Test    196.27    
        Line    148.15    3.77 Klines/sec [50% alpha]
        Rectangle    170.15    11.97 Krects/sec [50% alpha]
        Circle    177.73    4.10 Kcircles/sec [50% alpha]
        Bezier    173.51    1.89 Kbeziers/sec [50% alpha]
        Text    685.45    11.17 Kchars/sec
    OpenGL Graphics Test    124.37    
        Spinning Squares    124.37    87.03 frames/sec

10% overclocking of cpu and memory:

Quartz Graphics Test    196.59    
        Line    148.66    3.78 Klines/sec [50% alpha]
        Rectangle    171.09    12.04 Krects/sec [50% alpha]
        Circle    177.63    4.09 Kcircles/sec [50% alpha]
        Bezier    173.07    1.88 Kbeziers/sec [50% alpha]
        Text    687.12    11.20 Kchars/sec
    OpenGL Graphics Test    123.53    
        Spinning Squares    123.53    86.45 frames/sec

Test with version 1.0.1:

10% overclocking of cpu and memory:

    Quartz Graphics Test    185.20    
        Line    148.42    3.78 Klines/sec [50% alpha]
        Rectangle    169.56    11.93 Krects/sec [50% alpha]
        Circle    175.95    4.06 Kcircles/sec [50% alpha]
        Bezier    172.94    1.88 Kbeziers/sec [50% alpha]
        Text    345.27    5.63 Kchars/sec
    OpenGL Graphics Test    129.19    
        Spinning Squares    129.19    90.41 frames/sec

It is difficult to say anything about these numbers...in 1.0.1 Open GL seems to be faster...Quarz Graphics are slower
In 1.0.2 Quartz Graphics (Text plus 100%??) are higher than Open GL, but both results are hardly identical to each other... So it seems that there is no need to overclock!?
Where do these enormous deviations come from??
juergen "

Any test can have some small % of run/run variation, but as I've said many times before, xbench has shown run/run variations more than I'd like at times. I'd prefer some other test (even quake3 timedemos are preferable).

Dual 2GHz G5/Radeon 9800 128MB (retail)

(added 9/2/2004)
Radeon 9800 Retail 128MB, Dual 2 (Rev A.), ATIcellerator 1.0.1, 10.3.5
"Hey Mike! I tried some Halo timedemos and here are my results: (First is stock GPU and second at 10% over-clocked GPU and VRAM)
 Date / Time: 8/31/04 7:48:41 (0ms) 
 2000MHz, 2048MB 
 OSX\Applications\Halo\Halo -windowed -ip ... Frames=4700 
 Total Time=96.70s 
 Average frame rate=48.60fps 
 Below 5fps= 9% (time) 0% (frames) (9.657s spent in 11 frames) 
 Below 10fps= 10% (time) 0% (frames) 
 Below 15fps= 11% (time) 0% (frames) 
 Below 20fps= 14% (time) 1% (frames) 
 Below 25fps= 21% (time) 4% (frames) 
 Below 30fps= 25% (time) 7% (frames) 
 Below 40fps= 44% (time) 20% (frames) 
 Below 50fps= 57% (time) 32% (frames) 
 Below 60fps= 66% (time) 42% (frames) 
 ###Sound Options### 
 Hardware Acceleration= No 
 Sound Quality= High 
 Environmental Sound= No 
 Sound Variety= High 
 ###Video Options### 
 Resolution= 1280 x 960 
 Refresh rate= 0 Hz 
 Framerate throttle= No Vsync 
 Specular= No 
 Shadows= No 
 Decals= Yes 
 Particles= Off 
 Texture Quality= High 

 Date / Time: 8/31/04 7:51:31 (0ms) 
 2000MHz, 2048MB 
 OSX\Applications\Halo\Halo -windowed -ip ... Frames=4700 
 Total Time=84.03s 
 Average frame rate=55.94fps 
 Below 5fps= 4% (time) 0% (frames) (3.451s spent in 7 frames) 
 Below 10fps= 4% (time) 0% (frames) 
 Below 15fps= 5% (time) 0% (frames) 
 Below 20fps= 7% (time) 1% (frames) 
 Below 25fps= 13% (time) 3% (frames) 
 Below 30fps= 17% (time) 5% (frames) 
 Below 40fps= 33% (time) 15% (frames) 
 Below 50fps= 48% (time) 27% (frames) 
 Below 60fps= 61% (time) 39% (frames) 
 ###Sound Options### 
 Hardware Acceleration= No 
 Sound Quality= High 
 Environmental Sound= No 
 Sound Variety= High 
 ###Video Options### 
 Resolution= 1280 x 960 
 Refresh rate= 0 Hz 
 Framerate throttle= No Vsync 
 Specular= No 
 Shadows= No 
 Decals= Yes 
 Particles= Off 
 Texture Quality= High  

So a 10% increase in GPU and video RAM gave a 15% increase in frame rate. UNFORTUNATELY, video artifacts and glitches were constantly present. I'll pass on this.
Hector "

G4/450 AGP - OEM Rage128:

(added 9/2/2004)
"Mike, I've got an old Apple G4/450 AGP Graphite with 640 MB of RAM, OS X 10.3.5 and an OEM ATI Rage 128 video card, ROM version 110. The first version of ATIcellerator refused to recognize this card but 1.0.2 does.

The stock frequencies are 108 MHz for the GPU, 64.07 MHz for the VRAM. The best I can manage to boost it to before the card locks up the system (it will either freeze or go to a black screen) is 70.47 MHz (+10%) for VRAM and 116.69 MHz for the GPU (+8.05%). If you click on "Stock Frequencies" after boosting the speeds, the screen is out of sync until you do a whole-screen refresh, like calling a screen saver.

As the machine is not a modern machine I decided it wasn't worth the trouble to do any benchmarks. It does seem a little snappier, though and DVD playback seems smoother.
Pete C. "

Note: Reports below used v1.0.1 unless updated to note otherwise - some problems with certain models may be fixed by v1.0.2 (see link above).

PB G3 Lombard/RagePro LT:

(added 9/1/2004)
"This does not seem to work in a Lombard - not that I actually expected to squeeze any additional performance out of it. The installer runs, but then Kernel Panics as it is finishing. On reboot, a dialogue says the ATIccelerator II extension is not properly installed and to run the installer again. Did so and got another KP.

I sent a note to Mr. Perrier and we corresponded a bit. He does not think his utility will work with the old Rage Pro in the Lombards. The KP crashes were so hard that the panic.log had no record of them.
Jim M.
(he later wrote)
Just tried the current version (1.0.3) and I still get KPs. I guess I'll have to wait 'till I get handed down something better. "

Just for the record I asked Jim what OS version he was using. (10.3.x is listed as required - he later said he was running 10.3.5.) However for older graphics chips like this, I don't think it's worthwhile really.

eMac/Radeon 9200 32MB: (I've asked the eMac/9200 owners to try v1.0.2 which adds support for the 9200 graphics chip. See update note above)

(added 9/1/2004 - using v1.0.1)
"On the newest eMacs (Raedon 9200, I believe), opening the prefpane makes the screen go nuts - it is all very jumbled and completely illegible, so I have no idea what goes on afterwards, and I am forced to restart. Not force restart (I know my way around Mac OS X to restart even with a jumbled screen).
1.25 ghz eMac (the newest ones with USB 2.0), 10.3.5 "

Another eMac owner also reported problems (see next report) - hopefully v1.0.2 will fix that.

eMac G4/1.25GHz w/Radeon 9200 (32MB) (See his later comments that v1.0.2 helped)

(added 9/1/2004 - using v1.0.1)
"Radeon 9200 on emac G4 1.25
Aticcelerator II freeze my computer every frequencies I set, even the stock...
When I open the panel, it didn't recognize the frequencies of the card and i have to press return to set the stock frequencies, after that i can use my computer for 5 or 6 minutes and then totally freeze!! If you give me some help, i would be glad!!!!
Andrea T
(he later tried v1.0.2)
I've tried the new version (1.0.2) on my ibook G4 with ati radeon mobility 9200 (32 MB), works without problem. I've tried a max acceleration of 13 % on both memory and GPU. No glitches on sim city 4 (i've tried only this game) or Finder and the game seems faster. I haven't tried yet on my emac, when I can I will inform you immediately.
Andrea T. "

PowerMac G5/Radeon 9800 Special Edition (256MB)

(added 9/1/2004)
"Hi Mike
It certainly works. I clocked up my Radeon 9800 Special Edition by 10% and ran the Halo timedemos. As you can see, in Halo at least, +10% on the card = about 10% increase in framerate. However at those speeds there was some artifacting (Red pixels in black).

Here is the +10% timedemo - 416.25 core, 371.25mem

Date / Time: 31/8/04 7:19:42 (0ms)
2000MHz, 1536MB
Galapagos\Applications\Games\Halo\Halo Frames=4700
Total Time=137.63s
Average frame rate=34.15fps
Below  5fps= 2% (time)  0% (frames) (3.913s spent in 8 frames)
Below 10fps= 3% (time)  0% (frames)
Below 15fps= 8% (time)  2% (frames)
Below 20fps= 18% (time)  7% (frames)
Below 25fps= 33% (time)  17% (frames)
Below 30fps= 44% (time)  26% (frames)
Below 40fps= 71% (time)  54% (frames)
Below 50fps= 87% (time)  73% (frames)
Below 60fps= 92% (time)  81% (frames)
###Sound Options###
Hardware Acceleration= No
Sound Quality= High
Environmental Sound= No
Sound Variety= High
###Video Options###
Resolution= 1280 x 800
Refresh rate= 0 Hz
Framerate throttle= No Vsync
Specular= Yes
Shadows= Yes
Decals= Yes
Particles= Off
Texture Quality= High

And here is stock Radeon speeds - 380core, 337.5mem

Date / Time: 31/8/04 7:15:53 (0ms)
2000MHz, 1536MB
Galapagos\Applications\Games\Halo\Halo Frames=4700
Total Time=151.47s
Average frame rate=31.03fps
Below  5fps= 5% (time)  0% (frames) (8.729s spent in 11 frames)
Below 10fps= 6% (time)  0% (frames)
Below 15fps= 12% (time)  3% (frames)
Below 20fps= 23% (time)  8% (frames)
Below 25fps= 40% (time)  21% (frames)
Below 30fps= 50% (time)  29% (frames)
Below 40fps= 79% (time)  62% (frames)
Below 50fps= 89% (time)  76% (frames)
Below 60fps= 94% (time)  85% (frames)
###Sound Options###
Hardware Acceleration= No
Sound Quality= High
Environmental Sound= No
Sound Variety= High
###Video Options###
Resolution= 1280 x 800
Refresh rate= 0 Hz
Framerate throttle= No Vsync
Specular= Yes
Shadows= Yes
Decals= Yes
Particles= Off
Texture Quality= High

G4/PCI (Yikes)/Radeon 7000 PCI:

(added 9/1/2004)
"I have a Yikes with an OWC Zif G4 upgrade to 600 MHz (system bus still 100 Mz) and a pair of PCI Radeon 7000s. I'm running Quartz Extreme via PCI Extreme.

I upclocked both PCI cards by 19.32% (processor and memory) without artifacts or any evident problem.

In normal usage motion seems - possibly - smoother, though that could be simply wishful thinking.

Console log shows no errors.

I'm not a gamer, so I don't really know how to check objective improvements and there seems to be some disagreement as to what would correspond to a good general measure of this.

If someone would describe reasonably easy tests I am willing to run before-and-after on this system and report results.
Regards, Larry S.
(see above for his later 9/2/2004 report/tests) "

Dual 1GHz (MDD)/Radeon 9600 (from G5)

(added 9/1/2004)
"ATIccelerator II worx great on my (taped and soldered) Radeon 9600 (OEM G5) inside a 1 Gig Dual G4 MDD.
it is completely stable at 459 MHz Processor (+26%) and 380 MHz Memory (+29%). a few percent more resulted in either lockups or those typical artifacts. i didnt measure framerates, but the difference is clearly visible.
(xbench doesnt seem to be very happy with the G5 card in the G4 in general. scores quite low compared to the stock 9000, but UT2003 tells me something else...:-)
but ATIccelerator rocks!
i offer a tenpack to the first overclocker who fries his videocard with that :-D
-chris godi m. "

Dual 2GHz G5/OEM 9600 64MB: (updated 9/2/2004)

(added 9/1/2004)
"Preliminary testing...
I make no claims for safety or reliability in any ways =)

System G5 Dual 2.0Ghz 1.5GB RAM

Stock Test Quake3 1.3.2 Max settings
8x6     1024    1280    1600
205     181     130     83

1600x1200 Tests (processor/memory)
84.9    400/294
90      400/326.5
91      400/330.75

93.3    418/324
93.6    425/324
94      432/324

93.8    425/330
95.7    432/330
104.5   477/355.5

Interestingly at low rez, its completely CPU bound.
211fps 800x800 477/355.5

Is there any way to see how hot the thing is getting? I'm a bit hesitant to go any higher. I did some searching and found some people successful at some pretty high rates. 477/355 seems like a massive overclock 30%/20% CPU/Memory, but is achieving a 26% overall increase in frame rates, which is nice.

If you want any other tests done let me know, maybe I'll figure out how to bench with UT later. (See the Game section of the FAQ for how to test games like UT - also UT2004 FPS db entry page links to a utility for benchmarking the game)
--Jacob R.
(I asked what graphics card he had)
Stock OEM 9600 Pro 64MB

Later I did a bit more testing and found that at around 350Mhz on the memory I started getting visual glitches (dark fuzzy spots and some bright green pixels). Backing off to 344Mhz seems to have cleared it up.

1600x1200 max settings
104.5FPS 477/355 glitches
102.0FPS 488/344 clean
100.1FPS 450/344 clean
83  FPS  350/294 stock

It really seems they have to scale together to give any real meaningful efficiency. An 8%+ increase from 450 to 488 barely yields 2% more FPS.

I think 350/344 is a nice setting, compared to the stock 364/294 I'm getting about a 20% improvement in speed. That's a nice free upgrade for when I need it. "

iBook G3/800 - Radeon 7500 Mobility:

(added 9/1/2004)
"Radeon mobility 7500 in iBook G3 800Mhz
ATIccelerator II incorrectly reports 180MHz as stock speed, instead of 360MHz. I managed 540MHz on both memory and GPU. In Openmark, with 12 triangles, the results are:
  • 360MHz about 400FPS
  • 540MHz about 630FPS

With 7500 triangles however, the numbers are 69 and 73. In Xbench the increase is about 10%, highest in User Interface. Nothing to sneeze at, but not as impressive as the clockspeed increase might indicate. It seems simple things get more of a boost than complex things...
I ran 10.3.0 for the test. I got only to 440MHz on 10.3.3. Since a lot of games do strange things with later versions of MacOS X on this iBook G3, I plan to keep it at 10.3.0 for a while. (the 10.3.4 update feedback page here (linked on Misc/OS topics page) had some reader reports on lower opengl performance in 10.3.4, but the 10.3.5 feedback page had notes on better performance with its later drivers)
For superstitious reasons, I kept the memory frequency the same as the GPU.

It might be interesting to see if a lower clock frequency would increase battery life, but I think the difference will be minimal.
cheers, Michiel "

For reports from notebook owners, any comments on increased cooling fan activity would be good to note also.

G4/Cube w/Radeon 8500:

(added 9/1/2004)
"I am using it with a cube 550Mhz and a Radeon 8500.
(see Systems page, Cube section for past article on mods to fit larger cards like the 8500 in a Cube. PL's Cube housing I think also allows this.)

GPU at 279,00 Mhz that means 11.7 %
Memory at 283,50 Mhz - 3.28%

Overclocking the memory more than 5 %, produces problems. Overclocking the GPU more than 14 % causes crashing the machine.

I tried to downclock as low as possible and the result is: At aprox. 30 Mhz the machine crashed too. Maybe its a hint for cube owners to downclock their machine for cooling reasons.
This is a really good software
-mike "

Sawtooth G4 AGP/Radeon 8500 & iMac G3/Rage128

(added 9/1/2004)
"I tried ATIcelerator II on two Macs, a G4 (Sawtooth) 500 with a Radeon 8500 and a G3 iMac with a Rage 128 Ultra. The Radeon 8500 was already overclocked to 300 MHz on both the GPU and VRAM and has been stable at that speed for over 18 months with no artefacting or crashes. The Rage 128 Ultra is just the OEM unit Apple soldered into Mother's iMac.

With the Radeon 8500 and ATIcelerator installed but no additional overclocking I got noticable artefacting in Celestia, the most graphically intensive app I use. The fact that simply installing ATIcelerator destabilised my card makes me believe 300 MHz is the practical limit for my particular Radeon 8500. Any attempt to overclock with ATIcelerator, even 1-2 MHz, resulted in immediate and severe artefacting and a full system lockup if I did not bump the speed back down within 5-10 seconds.

I had no luck with the Rage 128 either, though I didn't expect much. ATIcelerator reported the GPU and VRAM speed as 65 MHz, not terribly exciting but this isn't exactly high end stuff. Curiously, the overclock factor at this stock speed was reported as 3700% (yes, three thousand, seven hundred percent). Clearly, ATIcelerator has some difficulty with this OEM chip. I gave it a go anyway. At 75 MHz I got noticable artefacting when I closed and relaunched System Preferences. Pressing the "Stock Frequencies" button was followed by pressing the reset button on the side of the iMac. Oh well, no harm done.
Ronald P. "

PB G3 2000 (pismo)/Rage128 Mobility: (other Pismo reports below)

(added 9/1/2004)
"Hey Mike,
I tested this out on my Pismo's Rage 128 Mobility. I discovered that the only setting that you can only overclock the Memory right now on the Rage 128 Mobility with this utility. This would explain why other users didn't notice much of a speedup, I don't know exactly how beneficial overclocked GPU memory is when the Core stays the same? Might help in 3d gaming. I tested this easily by underclocking the core to 9mhz (lowest it will go), then finding Xbench scores to be identical to stock 105mhz. I managed to be able to overclock the memory to 124mhz (+18.1%), anything higher and artifacts would appear on the screen (when doing the spinning squares test in xbench).
At around 128mhz, the screen would lock up with big colored verical bars....which a reset fixed. I believe that when we can overclock the Core of the Rage 128 Mobility, the results will be a lot more noticable with things like 3d gaming.
Ryan "

G4/AGP w/Radeon 8500:

(added 9/1/2004)
"I have a GigaDesign 1ghz (clocked at 1.3) AGP graphics (sawtooth I assume) with a flashed PC Sapphire 8500. The card was rated at 275/275 and has been running at 250/275 with the default Mac rom. Using this utility I can clock the core to about 310 and the ram to 320 before all hell breaks loose and it locks up. I have it running at 300/300 stable right now and don't feel like crashing my system any more to eek out another 5mhz...
Bart "

OEM 9000 Pro/MDD Dual G4 1.2GHz:

(added 8/31/2004)
"Hi there Mike. Just used ATIccelerator II to overclock the OEM ATI Radeon 9000 Pro (64MB) on my dual G4/1.25GHz MDD. ATIccelerator correctly identified the card's stock core/memory speed (274.5MHz/249.75MHz).

Before overclocking, I checked out the card to see which memory chips are on it. I was a bit disappointed to find out that it's only using 4ns Samsung chips (K4D64163HF-TC40) that are only officially rated to run at 250MHz. (I have a very early sample Radeon 9000 card with faster rated memory chips on it, but an OEM card from a past mac I had showed 250MHz rated memory) Faster memory chips would have given me more overclocking headroom but I guess that's too much to ask for an Apple OEM graphics card. ;-)

Still got a pretty good overclock out of the card though. It's currently running at a core speed of 299.25MHz (+9.02%) and I'm surprised to see the memory holding up quite well at 290.25MHz (+16.22%). Played Quake 3 for about an hour earlier without detecting any visual artifacts or "tearing". (Tearing can happen if the FPS rate exceeds the monitor refresh rate setting when Vsync is Off.)

Using Quake 3's timedemo (Demo Four.DM_68) at 1280x1024, with everything set to max, trilinear filtering, r_smp set to 1, average of three runs:

  • Stock (274.5MHz/249.75MHz): 87.2fps
  • Overclocked (299.25MHz/290.25MHz): 99fps (+13.5%)

Not bad. :-)
Keep up the good work. More power!
~Henjie "

Radeon Mobility 9600/PowerBook G4/1.25GHz

(added 8/31/2004)
"I've been trying it for a day now on a stock 1.25ghz 15in pb with Radeon Mobility 9600. It is completely reliable and without artifacts at 37% increase to both core and clock speed. Artifacts begin to appear under heavy stress at about 40% core, and under normal usage 45% core and clock is possible.

This is a huge overclock, however one should remember that Apple have clocked their 9600 mobility well under ATI's spec. These chips therefore have huge headroom above apple's levels.
Daniel S.
(he later wrote)
As a follow up, and after some further testing, here is a screenshot from my powerbook g4 1.25 ghz showing core and memory speeds.

The system is running stably. Xbench and Cinebench show negligible improvement, however the author of aticellerator suggests that these are too cpu bound, so I tried his suggestion which was the chimera benchmark, and saw an increase from around 70fps to around 95fps with the overclock.
That's one fast mobile graphics processor... "

That high an overclock is hard to believe. Sometimes there are errors when trying to apply settings (so no changes are done - console log will show that - see one of the past reports) But if he saw artifacts, then it must be overclocking it.
I asked if he noticed the cooling fans kicking on faster and if he ran any real-world tests (even a quick quake3 timedemo)

9800 Pro/MDD Dual G4/1.42GHz:

(added 8/31/2004)
"Hi Mike,
I've been playing with the ATIccelerator II for the last hour or so. I have a MDD Dual 1.42, 1GB RAM, ATI9800 PRO. My results have shown that my maximum clock rates are 6.55% on the processor and 9.33% on the memory. Anything above that for either and I start getting pixel artifacts.

Bombing Run Anubis 1600x1200 maximum

		before		after
min		10.94		15.11
average	46.47		50.42	
max		82.08		83.15

Asbestos Flyby 1600x1200 maximum

		before		after
min		40.68		50.79
average	102.65		103.32
max		336.98		336.85

I did quite a bit of other tests as well but my results mimic what you see here. The increase is greatest at the minimum frame rate and about 8% on average. I have not done any endurance tests and question whether it would pass those at not. For my money the increase in speed with stock cooling is not worth the potential loss of reliability.
Thank you, Louis S "

Radeon 9600 (non-XT)/G5 System: (Also see his later report dated 9/6/2004 above)

(added 8/31/2004)
"Hey thanks for noting the ATIccelerator update Through testing I've gotten my Radeon 9600 (in my G5) from the stock 365mhz core and 275mhz memory to a whopping 488mhz core and 380mhz memory! A higher core setting causes full system lock up, and I don't feel like playing with the memory right now

I started a base for my testing in Quake 3 with the high quality setting on Boli's config 1024x768

    Before=202.3 fps
  • After=263.7 fps!

Maybe I'll try this on my old iBook with a radeon
Thanks again, Noah "

Radeon 9600XT/Quicksilver G4/1.2GHz

(added 8/31/2004)
"hey mike, i had tested my newly installed radeon 9600XT (today's G5 bto) (assuming he did the tape mod on the card) in a quicksilver powermac g4 at 1.2GHz (PL 7457 upgrade).
while the card works great at 400 MHz GPU, 310 MHz VRAM (factory default), i set the GPU to 520 MHz and VRAM to 420 MHz and had run cinebench 2003. (that sounds way too high an OC to work)
in hw-lightning is no performance-gain! (I asked he check the console log for errors - see this previous note that sometimes the settings changes do not take due to an error)
it reached 867 cb-points at all possible clocks, even i underclocked the card to 300 MHz (GPU).
the same in xbench, there's no performance-gain.
regards, patrick "

Radeon 7000 PCI/B&W G3/400

(added 8/31/2004)
(PC Radeon 7000VE Flashed to Mac)
"I had originally used ATIccelerator I to overclock this card to 260/260 (from 178/178). I have added a fan to the stock heatsink on the card and have been running this way for about 2 months with no problems. I used ATIcceleratorII to reach the maximum speed (i.e. no lockups) of 232.62mhz/232.62mhz. Here are the XBench graphics scores on a BW G3/400mhz in 64bit PCI Slot (64Bit PCI slot (which is 33MHz) or the 66MHz PCI slot?) with Grackle Probe enabled:

Results 57.59  
    System Info  
        Xbench Version 1.1.3 
        System Version 10.3.5 (7M34) 
        Physical RAM 1024 MB 
        Model PowerMac1,1 
        Processor PowerPC G3 @ 399 MHz 
            Version Arthur v2.2 
            L1 Cache 32K (instruction), 32K (data) 
            L2 Cache 1024K @ 200 MHz 
            Bus Frequency 100 MHz 
        Video Card ATY,RV100 
        Drive Type TestHD 
    Quartz Graphics Test 56.21  
        Line 46.54 1.18 Klines/sec [50% alpha] 
        Rectangle 49.33 3.47 Krects/sec [50% alpha] 
        Circle 53.73 1.24 Kcircles/sec [50% alpha] 
        Bezier 55.08 598.54 beziers/sec [50% alpha] 
        Text 95.86 1.56 Kchars/sec 
    OpenGL Graphics Test 48.44  
        Spinning Squares 48.44 33.90 frames/sec 
    User Interface Test 73.24  
        Elements 73.24 23.56 refresh/sec 

The amount of headroom in this card is astonishing, but true! I wonder if the card being a PC Radeon 7000 VE has anything to do with it.
-Marcus L. "

I asked what the before scores were (the above is just OC'd info) and if he could run something more real-world.

Radeon 7000 PCI/B&W G3/900 (includes notes on console errors)

(added 8/31/2004)
"Hi Mike,
I tested it with ATI 7000 PCI, B&W G3 900 (PL G3/900 ZIF, and card in 66MHz PCI slot I assume).
First to Xbench OpenGL Values from others:
Sorry, forget those values !
An G5 User reported me: Xbench OpenGL NVIDIA 5200 / ATI 7000 PCI are only 30% different.
BUT, using the great Chimera OpenGL Demo /Bench (www.giofx.net/osx/index.htm, on their page there is also OpenGL MARK) the difference is much more realistic:
80 FPS NVDIA to 1-3 FPS ATI 7000 PCI.
(I'd rather see real-world tests, even if just a simple Quake3 timedemo, which is usually sensitive to changes in performance (CPU or video card) unless you're at the max limits of the card perhaps, but an increase in clock speed of gpu/memory should still show some small gain, depending on the overclock.)
My ATI 7000 PCI has same FPS (1-3) so its the card, not the CPU. I checked to tool.
I testet it with OpenMARK (max. Window) and get out a bit more FPS (3-5% speedup, setting MEM + GPU + 5%).
Some users told be they getting +50% GPU/MEM running. I don't think so.
Some problem is in the accell, because one time it changes nothing let set -20/+10 Values. (don't understand that last sentence) Chirmera shows always same FPS.
I looked in the console log, and voila:
2004-08-30 20:35:50.956 System Preferences[512] ATIccelerator: error 
10000003 writing to the I/O Registry
2004-08-30 20:35:51.452 System Preferences[512] ATIccelerator: error 
10000003 writing to the I/O Registry

Each time i changed or change % new entry whith that error appear. Happened only once. Not reproducable. But maybe if running high GPU usage things it is not possible to change it on the fly.
Changing running OpenGL Mark has produced no such error.
mitch (ATI NVIDIA GET RAM tool now with 2 card support) "

Beige G3/Radeon PCI and MDD G4/Radeon 9000:

(added 8/31/2004)
"works o.k. So far on a PM G3/300 (beige, with Radeon Mac Ed.) and a G4 MDD FW800 (Radeon 9000) with 10% overclocking each. Haven't tried more speed yet.
Regards, Holger E. "

iBook G4 933/Radeon 9200 Mobility

"When i first ran ATIccelerator II, the program incorrectly identified my Radeon 9200 Mobility graphics chip stock clock speeds. I have an iBook G4 933 and never downloaded any other tweakers.

The program had a setting of 380 mhz for memory and GPU when i opened the control panel! These are the settings for a Radeon 8500 destop chip. I had to click "stock frequencies" to get it to recognize the correct 180 Mhz GPU and memory settings.
I am not sure if there was any damage to the GPU, but there are now 3 pixel artifacts that hover around my mouse now. A restart may clear this up. (it did he later said)
Thanks Scott
(he later wrote)
The artifacts disappeared the moment I restarted. Everything seems fine without any problems. I even played Unreal Tournament 2004 to make sure it was ok :). Get about 20-40 fps with sound turned off. 10-20 with sound on.
The Radeon 9200 Mobility has a rated speed of 270/270Mhz. It is an upclocked 9000 Mobility with vertex shaders added. (you never know what clock speed apple may use however)
Thomas (the author of ATIccelleratorII) already emailed me about my previous problem and knows what went wrong. He was very helpful. Thomas's response for the incorrect clock settings was

    "This will be easy to fix. For some reason, some chips double or divide by 2 GPU or memory settings compared to most others. No chance of damage, as after clicking the button frequencies were halved! Hence the pixels, memory doesn't like being set too low."

I will let you know if anything else comes up.
Thanks, Scott "

G5 1.6GHz/Radeon 9600 (64MB)

"I came across a link to Thomas Perrier's new ATIcceleratorII utility. I downloaded and installed it on my 1.6GHz PowerMac G5 / 64MB Radeon 9600 Pro / 1.25GB RAM / OS X 10.3.5 system, and everything went without a hitch.
So far, I have successfully overclocked the Radeon 9600 Pro's core clock speed up to 418.5MHz (+14.81%) and the VRAM clock speed to 337.5MHz (+14.5%).

Using the Santaduck Toolpak benchmark suite, (www.santaduck.com/ut2k4/santa_toolpak_bench_ut2004.sit) this has resulted in an increase of the average FPS score in Unreal Tournament 2004 (1600x1200x32, Maximum detail, Flyby, DM-Antalus map) of 38.8 FPS to 44.9 FPS (approximately 15% increase in performance).
So far, there doesn't appear to be any side effects (such as image artefacts or crashes) from the overclock, although I am still to do any serious long-term reliability testing.
Cheers, Mat. "

I asked Mat if he'd run a test with the Asbestos map (which in my experience is less CPU bound than some other maps) and post his results using the FPS db page linked above.

B&W G3/Radeon 7000 PCI

"Mike, FYI, I tried the app this morning on the Radeon7K in my B∓W/G4-650, which is now running both the GPU & memory at +10.3% overclock.
2D/on-screen rendering in PS 7, window resizing & suck-in effect do appear a good bit smoother and faster, as is the Fast user switching rotating cube. Will run this way for a while before cranking it up any further :)

MDD G4 1.25GHz/Radeon 8500

"Hello mike, I'm working with an ATI Radeon 8500/64 MB in my G4 Dual MDD 1.25GHz, 2GB of RAM
Overclocking of 15 percent makes strange patterns after a while...
Now I'm using 10 % for GPU and memory.
... but do I realize any acceleration???...
(he later wrote)
I just passed a test with xbench: cpu and memory 10 % overclocked. See both results attached...
My mac works fine and stable this day without any artefacts.
However...working in Photoshop seems to be a little faster...but it is really not worth a mention. "

PB 2000 (Pismo)/Rage128 Mobility reports

(added 8/31/2004)
"Since I am supposed to switch to a PowerBook 17" in a couple of days, I decided to see how much I could pull from my faithful Pismo. So, I went from 109.5 MHz (I think it's the default) up to 126MHz. I was able to push the processor (GPU) frequency over 130MHz, but when trying to do the same with the memory frequency, at some point the display would just display vertical bars of different colors. Hitting the default "Stock Frequencies" in the pane didn't help and I had to restart.

So, now I'm running both the processor and the memory at 126MHz (+20%). I'll see how it goes. I haven't done much testing beside the Aquarium v2 screen saver which seems a little faster but will try to test some games later tonight or tomorrow.
Thanks! -Laurent. "

"I decided to try to use the ATIccelerator II to speed up my pismo 500mhz powerbook (1gb ram, 5400rom 40gb drive). As i'm sure you know it has an 8mb graphics card which doesn't support quartz extreme, (rage128 graphics chip does not support texture sizes that are not a power of 2, so no Quartz Extreme support.) and so os x performance isn't great.
I ran the latest version of xbench with only the quartz tests, open gl test, and a user interface test. I clocked the memory at the same speed as the processor. (depending on the % of boost, there there may be little real gain, but as I've mentioned here many times before - Xbench in my experience often has enough run/run variation even with no changes that I'd not consider it a great benchmark. And the opengl test is just a spinning squares demo) Here are the results:

Before overclocking (105mhz processor and memory speed):

Results	66.19	
	System Info		
		Xbench Version		1.1.3
		System Version		10.3.5 (7M34)
		Physical RAM		1024 MB
		Model		PowerBook3,1
		Processor		PowerPC 750 @ 500 MHz
			L1 Cache		32K (instruction), 32K (data)
			L2 Cache		1024K @ 200 MHz
			Bus Frequency		100 MHz
		Video Card		ATY,RageM3
		Drive Type		HITACHI_DK23FB-40
	Quartz Graphics Test	57.98	
		Line	47.48	1.21 Klines/sec [50% alpha]
		Rectangle	58.37	4.11 Krects/sec [50% alpha]
		Circle	57.85	1.33 Kcircles/sec [50% alpha]
		Bezier	65.63	713.12 beziers/sec [50% alpha]
		Text	64.46	1.05 Kchars/sec
	OpenGL Graphics Test	62.71	
		Spinning Squares	62.71	43.89 frames/sec
	User Interface Test	82.41	
		Elements	82.41	26.51 refresh/sec

This is at 110mhz (both gpu and memory):

Results	65.84	
	Quartz Graphics Test	58.67	
		Line	51.40	1.31 Klines/sec [50% alpha]
		Rectangle	57.79	4.07 Krects/sec [50% alpha]
		Circle	61.94	1.43 Kcircles/sec [50% alpha]
		Bezier	66.40	721.54 beziers/sec [50% alpha]
		Text	57.95	944.51 chars/sec
	OpenGL Graphics Test	61.77	
		Spinning Squares	61.77	43.23 frames/sec
	User Interface Test	81.09	
		Elements	81.09	26.08 refresh/sec

This is at 113.5 mhz:

Results	68.54	
	Quartz Graphics Test	60.89	
		Line	51.50	1.31 Klines/sec [50% alpha]
		Rectangle	57.69	4.06 Krects/sec [50% alpha]
		Circle	62.20	1.43 Kcircles/sec [50% alpha]
		Bezier	66.41	721.58 beziers/sec [50% alpha]
		Text	70.28	1.15 Kchars/sec
	OpenGL Graphics Test	63.87	
		Spinning Squares	63.87	44.70 frames/sec
	User Interface Test	85.54	
		Elements	85.54	27.51 refresh/sec

This is at 115.5mhz:

Results	67.28	
	Quartz Graphics Test	60.10	
		Line	51.65	1.31 Klines/sec [50% alpha]
		Rectangle	58.51	4.12 Krects/sec [50% alpha]
		Circle	57.12	1.32 Kcircles/sec [50% alpha]
		Bezier	66.36	721.04 beziers/sec [50% alpha]
		Text	70.58	1.15 Kchars/sec
	OpenGL Graphics Test	61.88	
		Spinning Squares	61.88	43.30 frames/sec
	User Interface Test	84.83	
		Elements	84.83	27.29 refresh/sec

As you can see there is no real increase in speed that is predictable. What i mean is - it isn't really worth it for this pismo. I really wish i could get better graphics performance because it's fast enough for what i use it for - i use my powermac for any real work. It's just the jerkyness of the UI that annoys me.

Unfortunately this didn't help - i was too scared to push the overclocking any further although i didn't feel it warming up, and didn't notice any artifacts.
Many thanks for a great site.
Eddie S
(he later wrote)
i don't have any games to test on (not really worth it on a pismo!). i might try running itunes visualiser to see if i notice a higher average frame rate but it's not really worth it, and it's a bit of a unscientific method. Perhaps i could have pushed the overclocking further, but i didn't fancy frying the onboard graphics! "

I agree - you're probably not going to get much more out of the OEM Rage128 mobility chip and the difference may not be worth the increased heat/risk of failure.

ATIccellerator II Info
Here's repeat of the original news post (See the linked page for the latest info/updates, supported card models, etc.)

"ATIccelerator II
ATIccelerator II comes with ABSOLUTELY NO WARRANTY. If you destroy your costly 9800 Pro or your now worthless Rage 128 because of a bug in this program or your careless utilisation of it, you're on your own!

ATIccelerator II can change ATI graphics cards frequencies live, on-the-fly, under Mac OS X. It's much more sophisticated and convenient to use than ATIccelerator(I) for the following reasons:

  • no potentially dangerous flashing required
  • no cumbersome three-steps process (dump, modify, then reflash ROM)
  • no need to reboot for every frequency change
  • no OS 9 required (that's right, G5s and other recent OS 9-free Macs are now supported!)

Graphics cards overclocking on the Mac finally comes out of stone age!

  • Mac OS X 10.3.9 or later (v1.0.5 and earlier supported OS X 10.3.0 and later)
  • An ATI Rage128 or later graphics card

A number of cards and chipsets have been successfully tested, from the Rage 128 Mobility to the Radeon 9800 Pro. However not everything from ATI available on the Mac was tested, so drop me a line if something looks suspicious. (Rage pro cards not supported) "

Return to Main News Page

(Ad/Sale items)

= UPGRADES by Mac =
Upgrades just for

= Refurb Mac Pros =
(Click for List)

SSDs up to 4TB
Fast SSDs for Most Macs/PCs

= ThunderBolt =
Drives, Docks & More

Up to 12TB HDDs
HGST, WD, Seagate, Toshiba

= 2.5in HDs & SSDs =
Notebook Hard Drives and DIY drive/case kit bundles

Lifetime warranty RAM Upgrades!

Internal and External Superdrives/Blu-Ray drives

Graphics cards, Displays, Adapters, Cables & more

Interfaces, Cables, Software, Speakers, Headphones & more

Apps, Utilities, OS, VM, Games and more

WiFi and Bluetooth Devices/Adapters/More

= Repair Service =
for iPhone, iPad, Macs

= iPad/iPhone/iPod =
Accessories, Cases, Repairs & More

NuGuard KX Cases
*Extreme* Drop Tested!

XLR8YourMac T-Shirts

= back to www.XLR8YOURMAC.com =

= Other Site Topic Areas =
Mac Mods/Upgrades | CPU Upgrades | Storage | Video | Audio/HT | OS Updates/Network | Recent

Copyright © 1997-2018. All Rights Reserved.
All brand or product names mentioned are properties of their respective companies.

Legal: Site Privacy and terms/conditions of use.