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: Win Smith who wrote (148088)	11/14/2001 1:03:46 PM
From: Ali Chen	Respond to of 186894

Win, "Only the Sandra memory benchmark showed a (<1%) improvement" Let me try to explain one more time. No Sandra benchmark is supposed to show any improvements due to cache size. The Sandra memory benchmark is designed to bypass caches of any size, and the result must be independent. The CPU/FPU Sandra benchmarks are totally cache-imbedded, with no off-chip traffic, and therefore will show the same result no matter how big the cache is. In real benchmarks like SPEC, about 50% of benchmark run time (on average) is wasted in accesses to memory. Access to memory occurs when a cache miss happen. Let say the cache hit rate is 97% for 256, and only 1% improvement for 512, or 98%. Seems like not a big deal. However, the whole off-chip traffic is caused by those MISSES, and going to 2% miss rate from 3% miss rate is equivalent to the off-chip traffic reduction by 33%, which will reduce the total run time by 15% (assuming 50% memory-to-CPU time budget). - Ali