Navigate: Advice Beginners BIOS Guide CPUs Links Mainboards Memory Network Storage Video/Sound Cards Contact Forum SiteMap Sponsors WebNews Home	.	.
Prices: Mainboards ABIT ASUS Chaintech Shuttle Soyo Tyan CPU Intel P4 2.4C-800 P4 2.6C-800 P4 2.8C-800 P4 3.0-800 P4 3.2-800 AMD AthlonXP XP 1700+ XP 2000+ XP 2400+ XP 2500+ XP 2700+ XP 3000+ XP 3200+ Athlon64 Athlon64 3200+ Athlon64 FX-51 Opteron Opteron 240 Opteron 242 Opteron 244 Opteron 246 Memory Corsair Crucial Kingston Mushkin OCZ Search Prices:

LOSTCIRCUITS	SHORTCUTS: New Performance Dimensions Test Configuration Cachemem Sandra Everest Analysis-1 FarCry The Poodle's Core More Benchmarks Conclusion Give Us Some Feedback to Help Us Improve our Reviews
Intel Pentium4 Prescott / 925XE chipset New Dimensions in Performance
(Review by MS November 21, 2004)

More Benchmark Results

No rule without exceptions. Unfortunately, this turned out to be true in this case as well. Futuremark's 3DMark2001 and 3DMark'05 were immune against the thermally inflated scores, otherwise there would be no competition left on their leaderboards and I would not have posted this article. Interestingly, the fps shown by FRAPS in 3DMark'05 were up to 10 x inflated compared to those shown in 3DMark's own progress bar.

Prescott at 3.73GHz (red) vs. ExtremeEdition (blue). Prescott wins! In 3DMark2001SE, the 1066 MHz host bus buys the Prescott a 1000 point increase in score over the stock 560 version (25231vs 24246).

Other than 3DMarks, there are a few other benchmarks that did not seem to be affected by the issue described:

• Final Fantasy
• Lightwork RenderTest and most other 3D Rendering benchmarks
• Most System Benchmarks (but we only checked a very limited sample)

Conclusion

The board we used here happened to be an ASUS board, however, we have inflated data from other manufacturers as well and the problem is not confined to the use of a Prescott CPU either since we could replicate some of the results with an ExtremeEdition as well. Basically, any 925XE (or 925X) board using the ICS 954123 or similar clock generator will have the same problem, meaning that the benchmark results will suggest that the board is running faster than it actually is and using an independent counter like FRAPS does not solve the problem either - on the contrary.

One thing that needs to be mentioned here is that Intel's reference / production board uses an IDT CV115CVP clock generator and we will have a look at that one in the next few days to see whether we find similar issues.

Adding the heatsink to the clock generator ameliorates the problem but is far from solving it in that it only provides relief for a limited amount of time. Liquid nitrogen does the trick but it is not exactly a household item. In other words, until the issue has been addressed by the mainboard manufacturers or by Intel as the supreme / divine authority, it is going to be hard to accept any benchmarks unless they have been qualified as not to be affected by the clock-in-heat bug.

On the other hand, our findings may provide Zippy the sales guy with just the tool to show off the superior performance of his systems that he is so desperate to sell and THG with another article.

next page: => More =>