Silicon Investor (SI) -- The First Internet Community

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) -- Ignore unavailable to you. Want to Upgrade?

To: wanna_bmw who wrote (61149)	10/30/2001 12:03:12 PM
From: combjelly	Read Replies (1) \| Respond to of 275872

"EV7, while a highly scalable server processor, is not inherently different than EV68 if you use it in uniprocessor mode." That is not correct. The high bandwidth, low latency memory bus makes a huge difference in uniprocessor mode. That is, for anything that doesn't fit in the cache. I agree that for those programs, EV7 doesn't make any difference. But the micro-architectural changes in the Hammer would help some (albeit not a lot) with even those programs. And the bandwidth/latency issues are important. If you look at Hans DeVries paper where he compares the throughput between a (as it turns out, hypothetical) Mustang vs. a Willamette, you see how the IPC drops a lot as you progress through the memory hierarchy. Assuming that the slides that AMD presented at MPF are accurate, the Hammer will flatten that IPC degradation a lot. And that will go a long way towards optimizing the access to memory, which is becoming an even more serious bottleneck than it is now. Until the underlying technology behind memory technology changes, this is something that will be an issue. Sorry if you mis-understood, I am not trying to underscore AMD's scalable server plans. I am saying that AMD has laid the groundwork for moving SMP and NUMA into the mainstream.