Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Intel Corporation (INTC) -- Ignore unavailable to you. Want to Upgrade?

To: Paul Engel who wrote (153859)	1/5/2002 3:13:41 AM
From: wanna_bmw	Read Replies (4) \| Respond to of 186894

Check out some Prestonia pictures - Complete with HYPERTHREADING benchmarks! whiningdog.net I love that Windows Task Manager with 4 CPU windows. :-) Here are the results. Test 2x1.7GHz Foster 2x1800+ Athlon XP 2x1.8GHz Prestonia (SMT only) 2x1.8GHz Prestonia (SMP + SMT) CPU INT 6240 MIPS 8480 MIPS 6825 MIPS 4655 MIPS CPU FP 4051 MFLOPS 4248 MFLOPS 4318 MFLOPS 7040 MFLOPS MM INT 13423 it/s 16714 it/s 14116 it/s 13001 it/s MM FP 16314 it/s 19520 it/s 17241 it/s 22563 it/s These are some interesting results. With only a single 1.8GHz Prestonia CPU enabled (2 logical threads), it's able to partially outperform 2 Foster CPUs at 1.7GHz, and come close to outperforming 2 Athlon XP 1800+ CPUs. When 4 threads are enabled with both SMT and SMP, however, we see that floating point math gets a huge performance boost (63% in the CPU test and 31% in the multimedia test), but integer performance takes a large hit (more than 30% in the CPU test and almost 8% in multimedia). Keep in mind that this is a synthetic test, and as such it could expose problems that wouldn't show up in other applications, but there definitely seems to be a bug, either with the Sandra application, or the CPU itself. Hopefully, there is a fix on the way, since floating point performance looks very promising, and gives the Athlon XP a run for its money. Some other observations include an option that is not set in the Sandra console labeled "Dynamic MP/MT Load Balancing". This might be one reason for the low integer scores. The author himself postulated that context switching between similar datasets (being that this is a synthetic test) might also have something to do with performance. Another possibility is that this Prestonia sample is old, and since it was released, performance bugs may have been solved. Time will tell, of course, but I am eager to see more. wbmw