SI
SI
discoversearch

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.
Politics : Formerly About Advanced Micro Devices

 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  
To: Petz who wrote (87966)1/19/2000 4:32:00 AM
From: Petz  Read Replies (1) of 1572238
 
Thread, a better estimate of the benefits of a full speed L2 cache in the Athlon was calculated based on the second, more accurate model of the effect of the L2 cache on benchmark speed. This model was described in the referenced post above. The benchmark is CPUMark99 which is one of the ones that favors Intel processors. Under WIN98, an 800 MHz CuMine outperforms the Athlon 800 by about 4% when using the BX chipset and 7% when using the Camino chipset and 800 MHz RDRAM.

I used a spreadsheet to calculate a linear regression for the model. Based on the model, an 800 MHz Athlon would run CPUMark99 11.3% faster if it had a full speed cache instead of the 1/2.5 speed cache currently available.

This would push the CPUMark99 perforance of the 800 MHz Athlon to be equivalent to a 900 MHz CuMine with the BX chipset, or an 866 MHz CuMine with RDRAM.

In doing this analysis, I made the assumption that a full speed on-die cache would have the same number of (memory) clock cycles of latency as the slower off-die cache does. Scumbria can perhaps lend his opinion as to whether this is conservative or optimistic w.r.t. latency of on-die caches. I also made the assumption that the on-die cache is 512K, which may or may not be true for the first Athlon with on-die L2. (Mustang?)

Petz
Report TOU ViolationShare This Post
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext