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.

Politics : Formerly About Advanced Micro Devices -- Ignore unavailable to you. Want to Upgrade?

To: Petz who wrote (80926)	11/24/1999 1:50:00 AM
From: Process Boy	Respond to of 1574005

Petz - <PB, can you comment on the following aspect of the CuMine design? It seems to me that the enhanced L2 cache is quite a bit of overkill for the P6 core. When combined with a core that includes a few more execution units, such as I expect on the Willamette, the enhanced L2 will really shine.> I'm not a design or microarchitecture expert, so guys like Ten and Scumbria are probably better suited to addressing this. I actually don't know all the reasons behind the Coppermine L1 vs. L2 strategy. I would be as interested as you to hear some debate on this topic. I do rub elbows occasionally with Coppermine design folks. If I find out anything that I can share on this question, I will certainly post it. PB

To: Petz who wrote (80926)	11/24/1999 2:57:00 AM
From: Tenchusatsu	Respond to of 1574005

Petz, <It seems to me that the enhanced L2 cache is quite a bit of overkill for the P6 core. When combined with a core that includes a few more execution units, such as I expect on the Willamette, the enhanced L2 will really shine.> The main reason for enhancing the L2 cache on Coppermine is that it is relatively easy to do compared to the alternatives. At MPR, Intel revealed they had considered enlarging the L1 cache (huh, huh, he said enlarge), but enhancing the L2 cache was the decision made. Working on the L1 would have brought too many other variables into play, such as frequency scalability and other micro-architectural concerns. <Because of the big L1, AMD may only get 10% improvement in SPECint when they have a full speed L2.> Performance scalability is the big issue here, not necessarily immediate improvements in performance. It's apparent that an off-chip L2 cache is not going to scale in frequency as well as the Athlon core. Moving the L2 on-chip gives Athlon the scalability it needs for the near and far-term future. By the way, look for "scalability" to become an increasingly-used buzzword. Tenchusatsu