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: steve harris who wrote (91611)2/4/2000 8:48:00 PM
From: kash johal  Read Replies (1) of 1579479
 
Steve,

Re:"Would athlons and cu athlons have same problems"

Quite probably.

As you go down the channel length curve the leakage currents become larger and more significant but the transisitors get faster.

AMD with the k6-3 used an architecture where the on chip l2 RAM runs at full speed and they apparently failed to design it with redendancy.

Intel on the other hand with coppermine came up with a very elegant solution which reduced latency but runs the on chip cache at 1/2 speed. So Intel can actually work with slightly larger channel length in the RAM area and therefore minimize such leakage issues. It also makes the cumine much more scalable in MHZ as cache no longer is in critical path.

Now If AMD has just done a simple shrink of the k6-III into 0.18 micron to generate the K6-2+/K6-III+ we might be in a world of hurt. They will require the cache ram to be full speed and so likely will see a scalalbility issue. As they didn't build much if any redundancy into the K6-III that expalins the choice to offer the K6-II+. This should at least ensure high yields even though die size will not be optimum. IMHO they will likely sell the K6_III+ that do yield into the mobile space.

It seems they have not bothered to really redesign and rearchitect the K6-III due to limited product life although there have been rumors that they did add some more 3dnow stuff.

When spitfire and thunderbird come out - if they have gone with the same high speed cache architetcture we could be in a world of hurt.

Alternately they may have learnt from cumine and k6-III problems and gone with a flexible l2 cache approach.

Remember current AThlons allow the off chip L2 cache to run at several divisors - even 7/8 to 3/4 to 1/2 to 1/3 etc etc.

I suspect they will have that flexibility with spitfire and thunderbird on chip.

By going on chip the latency will still be reduced boosting performance.

We will know soon enough of course.

regards,

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

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