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 -- Ignore unavailable to you. Want to Upgrade?


To: fyo who wrote (66145)7/19/1999 8:52:00 PM
From: Ali Chen  Read Replies (1) | Respond to of 1578130
 
<Any idea of how much the higher latency would hurt performance?>
There are some... Especially if SRAM latency remains
the same. :)
See also:
Message 10565357



To: fyo who wrote (66145)7/19/1999 9:03:00 PM
From: Tenchusatsu  Read Replies (1) | Respond to of 1578130
 
<There have been some rumors that AMD would pull a 'DDR'-SRAM to make it more feasible to do 'full-speed' L2 cache. Any idea of how much the higher latency would hurt performance?>

Higher latency? I didn't know there will be any just by going to DDR SRAM. I thought that by moving to DDR SRAM, latency stays the same (in terms of absolute time), but burst transfer times are cut in half. This allows for a slight performance improvement for only a slight increase in cost. (Theoretically speaking, of course.)

This is different than Xeon's expensive CSRAM, where both latency and burst transfer times are shorter than Pentium III's BSRAM.

Tenchusatsu



To: fyo who wrote (66145)7/19/1999 9:27:00 PM
From: Scumbria  Respond to of 1578130
 
Fyo,

Any idea of how much the higher latency would hurt performance?

It would be crazy to sacrifice latency to achieve bandwidth. CPU's generally burst data only 32 or 64 bytes at a time. You don't need a lot of bandwidth to keep up with that. Latency is far more significant for a single processor system.

Scumbria