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: Scumbria who wrote (85084)	1/5/2000 2:25:00 AM
From: Petz	Read Replies (1) \| Respond to of 1574001

Scumbria, <RDRAM improves bandwidth, but is not very good for latency. Unfortunately, latency is the problem here, and Rambus is not the solution.> What about the integration of the RAMBUS controller on the CPU? If that can reduce latency, why not integrate a DDR ESDRAM controller onto the CPU? Too many pins? I'm sure we could get around that in a year. Petz

To: Scumbria who wrote (85084)	1/5/2000 2:34:00 AM
From: ptanner	Read Replies (2) \| Respond to of 1574001

Scumbria <RDRAM improves bandwidth, but is not very good for latency. Unfortunately, latency is the problem here, and Rambus is not the solution.> For small (piece) memory loads then pre-fetching and effective cache design would seem most effective (as noted in your earlier post-85038). However, when it is a large access, say for a repetitive graphic manipulation then wouldn't a high bandwidth be more significant than latency? The original question was how to reduce memory latency so my mention of RDRAM was off-track (maybe I need gray matter ECC?). As far as a real-world solution to problems with cutting edge processors... simply make it easier to transition from one machine to another, buy a notch down from the top and increase the frequency of purchases. There are applications (professional CAD/graphics/etc) for which it always pays to have the best you can buy but when the best keeps evolving by 20-50% annually then the only realistic solution is faster replacement and hopefully the old machine can find a new home (trickle down worked well at my former employer and the engineers were always happy to get an "old" CAD machine). A former co-worker took this approach with his last home PC (we had both purchased high-end several years before) and figures that for the same annual cost he can have a new machine every year and overall have a higher average performance. PT