![]() |
![]() |
||||||||
| Updated June 15, 1999. | |||||||||
ResultsI have had my iMac A clock-chipped to 300 MHz since the beginning of November, 1998. The first week it ran unreliably for 45 minutes to two hours at a time while I experimented with different cooling solutions. On 11/8/98 I installed my modified K6 heatsink & Pentium fan, and it has been running very reliably since. I have had it running all day, every day for the last several weeks, including several 6- or 7-day stretches without restarts, without problems. I really have only had a handful of crashes since clock-chipping: once in a great while Quicken 98 hangs the system when launching, a couple (and I do mean only two or three) when using Sherlock, and a few instances where the only way I could establish a reliable TCP/IP connection was by restarting. (My iMac has had that problem since I got it, first with FreePPP and now, to a lesser extent, with ARA 3.1.2. Limiting the modem with Apple's 33.6 connection script has helped tremendously). I have upgraded to 8.5.1, now 8.6, and have applied the modem updates and all ROM/firmware updates up to and including the Firmware Update 1.2, all without incident. Apple's System Profiler still identifies my iMac as an iMac, albeit at 300 MHz.
I have noted two quirks since the modification: TechTool Pro 2.1.1 reports that the PRAM itself tests fine, and other PRAM functions (mouse speed, General Controls settings, disk cache size & startup disk) are remembered.
Performance is fast. Very fast. I have to admit that in day-to-day use of the iMac, I don't really see much of a difference from before -- I don't use it for page-layout, lots of Photoshop work or complex number-crunching (that's what my 7300/200 at work is for, and WOW can I tell the difference there -- slow!). The benchmark results are impressive: processor and FPU scores are about 29% better, and even the disk score improved slightly when using the System Info benchmark-standard 128k disk cache. I'm running 8.5 on my iMac, and when I set my disk cache to the "recommended" size (5120k in my case, with 160MB of RAM), my disk score improves tremendously. My RC5-64 "Bovine" client now cracks up to about 977 kkeys/sec, from 755 kkeys/sec maximum before (29.4% faster). See below for some benchmark results using System Info from Norton Utilities 4.0. The hard drive was optimized before each run; all test parameters were standard. Please note that all G3's below are beige G3's. I hope to change that soon.
You can see from the above benchmarks that an iMac "A" at 300 MHz is not quite as fast as a beige G3/300, due to the latter's 1 MB of Level 2 cache (the iMac has 512k) and its Rage Pro graphics chip (the iMac "A" has a Rage IIc chip, which doesn't perform quite as well). Still, the gains over an unmodified iMac are significant. The beige G3/300 and my iMac were both running OS 8.5; thus the wide variation in the disk scores for each (OS 8.5 significantly improves disk cache performace when it is set to the "recommended" size). The wide variation in disk scores between the unmodified A model and unmodified B model above is due to OS 8.1 on the A and 8.5 on the B. Click here to download a stuffed & binhexed file containing my iMac System Info files (24k), as well as those I made for a beige G3/300, and those made by Yoshi of an iMac B. If anyone has the time to make a file for the Rev C model, please e-mail me a copy! I have just received confirmation from a reader that clock-chipping the Rev. B models works the same way as the A models. Yoshi's iMac B is running at 266 MHz now, with no extra cooling employed. Thanks, Yoshi! I have no confirmed information on the Rev. C (266 MHz) or Rev. D (333 MHz) models, except to mention once aagain that Lars in Norway has taken the processor card from a Rev. D, overclocked it to 400 MHz, and stuck it into a Rev. A, with no apparent trouble. If you have clock-chipped your iMac, Rev. A, B, C or D, please e-mail me at danb@thelittlemacshop.com and let me know how fast you got it to go, whether you needed extra cooling, and what you did to get that extra cooling. I'd like to get a rough tally of the number of people who have managed to do this, what speeds they have achieved, and what they did about the heat problem. I'll post whatever information I learn. If you have an iMac, or any other PowerPC-based Macintosh, be sure to install Motorola's LibMotoSh library (unless you're running System 7.5.5). It can speed up some floating-point operations tremendously. Here's the link for the download page, and here's a link to a FAQ page. The benchmarks above were all done with this library installed, but, for some reason, Norton's System Info doesn't seem to register a difference. Other peoples' results:
|
|||||||||
Free Stats Tracking |
|||||||||
|
|
|||||||||
| All of the guidelines and photographs on these iMac modification pages are the original work of Dan Buettner. Please feel free to print, copy or save this material for your own reference or for the reference of others. Also feel free to provide links to these pages. If you are going to re-post or otherwise reproduce a small portion of this material in any way that is not for profit, please give credit where credit is due. You may not reproduce all of this material, or profit from the reproduction of any or all of this material, without my express permission. Contact me at danb@thelittlemacshop.com. | |||||||||