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: Elmer who wrote (35502)8/4/1998 10:06:00 PM
From: Scumbria  Read Replies (1) | Respond to of 1573648
 
It wouldn't surprise me if you're right. On what do you base this 10% estimate?

Back of the envelope calculation:

L1 hit = 1 core clock
onboard L2 hit = 3 core clocks
off chip L2 hit = 10 core clocks
Dram page hit = 20 core clocks
Dram page miss = 100 core clocks

L1 hit rate = .98
L2 hit rate = .90
Dram page hit rate = .5

onboard L2 latency = 1*.98 + 3*.019 + 20*.005 + 100*.005 = 1.637
off chip L2 latency = 1*.98 + 10*.019 + 20*.005 + 100*.005 = 1.77

The average memory latency decreases by less than 10% with an onboard L2.

Scumbria




To: Elmer who wrote (35502)8/4/1998 10:11:00 PM
From: Jim McMannis  Respond to of 1573648
 
Well, I said 50% on some apps becasue of the Shiloh Jennings article which Petz now says is bogus. I suspect Petz is right...but no one really knows. If it's only 10-15% then like scumbria said the K6-3 isn't as important as the Mendocino. More important is AMD getting the speed grades up so it can make some money on it's current K6-2 production set up.
Jim


HomeMailHotSubjectMarksPeopleMarksKeepersSettings
Terms Of UseContact UsCopyright/IP PolicyPrivacy PolicyAbout UsFAQAdvertise on SI
© 2025 Knight Sac Media.  Data provided by Twelve Data, Alpha Vantage, and CityFALCON News