Advanced Micro Devices - Moderated (AMD) Message Board

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 : Advanced Micro Devices - Moderated (AMD)

AMD 217.53

+1.5%

Nov 28 9:30 AM EST

Public Reply Prvt Reply Mark as Last Read File

Previous 10 Next 10 Previous Next

To: Joe NYC who wrote (70359)	2/5/2002 2:07:35 PM
From: pgerassi	Read Replies (1) of 275872

Dear Joe: I believe that an AXP 2000+ beats a NW 2.2GHz in unbiased tests (where P4 isn't being pushed as Intel derived benchmarks prevail but, older or more even handed benchmarks are chosen). Thus a AXP 3000+ would do the same for a NW 3.3GHz (50% increase for both). When NW gets 133x4 FSB, Barton and Tbred would get 2166 or perhaps 2200 and AXP is more efficient at the same bandwidths (AMP more so). You also assume that Tbred would not get a similar upgrade to 512KB and that is not known either way but, the die size and the new shape of leaked Tbred photos show that adding an additional 256KB of L2 to one edge of the Palomino die rectangle would yield the new aspect ratio die rectangle pictured. This is the strongest evidence that L2 cache is doubled. The reputed die size is another justification that the L2 is increased. That's information with two to one bias toward 512KB L2 and that opposing one is based on an assumed rumor and thus, I give it less weight. Also you seem to assume that TBred would have no IPC improvements at all and that is not what usually happens during a process shrink especially given the aspect ratio change. Just doubling the L1-L2 width and doubling certain TLB and branch prediction table sizes could add 10% to IPC by themselves. No, I think you are being too pessimistic by far for AMD improvements and too optimistic for NW. NW overclocking shows that by clock they may go high but, the performance improvements stop like an engine that revs too much. Most engines in the car industry have red lines at or near their peak output. It appears that NW can over rev and thus is below the performance at slower clocks. This is definitely due to overclocking (PowerNow in reverse) causing throttling. A NW at 2.6GHz does not get performance in demanding applications of 2.6/2.2 of a 2.2GHz part. It looks to be about 66% to 75% of that ratio or more like a 2.5GHz NW. This would be the true overclock of NW and far away from the 3.0GHz you believe achievable by year end. Given these facts, I predict NW will be around 2.5 to 2.7GHz and in very limited quanities at that by year end. Barton will be somewhere in that range if a little slower say 2.33 to 2.53GHz by year end. Of course, unlike many here, I assume that Clawhammer would arrive too late for Christmas builds except in SD shops but be as fast as a 3.7GHz NW some 50% ahead of the performance of the top NW at that time where even the most highly P4 optimized tasks will be lost to it (excepting those like WME 7.0). Of course NW will lose any benchmark needing over 4GB of memory to work well. Pete

Report TOU Violation

Share This Post

Public Reply Prvt Reply Mark as Last Read File

Previous 10 Next 10 Previous Next