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: Scumbria who wrote (28608)	2/15/2001 2:11:58 PM
From: Win Smith	Read Replies (2) \| Respond to of 275872

Without getting over my head, my view from the outside is that processor+ eDRAM, connected by LDT to an nVIDIA Xbox style integrated northbridge memory/graphics controller looks like a good next step PC architecture. Most of the pieces seem to be falling in place for that. If they need more stuff to fill out the northbridge chip, there's always eDRAM and the southbridge to integrate. If LDT (HyperTransport is it now?) is really son of SCI, then there are many, many alternative configurations that should be doable quite easily. Most would be overkill fo mass market PCs, though. -Win.

To: Scumbria who wrote (28608)	2/15/2001 2:22:16 PM
From: Joe NYC	Read Replies (1) \| Respond to of 275872

Scumbria, Onboard memory controllers or graphics are a big, bad no-no. I think AMD has learned from the mistakes of others. The major problem of Cyrix MediaGX, MXi and Timna is that they brought on board too many things. The lesson IMO is not that nothing can be moved to the CPU, but that you have to be careful what to bring over. I don't think you would dispute that bringing L2 on-chip was necessary to maintain performance scaling. To make further advances here is of course to increase the size of L2, but onboard memory controller is the next step. There is another trend, to reduce costs, and building block approach that LDT promises may be the right approach to deliver high performance CPU + flexibility in building the systems. There are 2 approaches that seems to run counter to this. One is as I mentioned Mamba, another is the NVidia approach of (and I am guessing here) to bring massive bandwidth to the system, to be controlled more or less by the graphics processor, and leave some spare bandwidth to the CPU. Joe

To: Scumbria who wrote (28608)	2/15/2001 4:27:35 PM
From: Joe NYC	Read Replies (1) \| Respond to of 275872

Scumbria, Onboard memory controllers or graphics are a big, bad no-no. Another point on this matter: You can reasonable claim that there is a potential for delays of the CPU because of delays in memory controller. But the Northbridge manufacturere faces the same delays. And you have additional delays caused by annual increases of the FSB speed (not necessarily synchronized with increases of memory speed, causing 2x the number of transitions). Let's just review the recent history of Tbird: Apparently, Tbird was ready as early as April or May. It kept getting delayed because there was no chipset for it. It took until about July to see the first KT133, and until August until the motherboard makers started selling the motherboards. Now, if you took the very first Tbird manufactured in May, you could plug it today to a KT133A motherboard, and run it at 133 MHz FSB just fine (increasing performance by a speed grade). But we had to wait until February of this year, almost a year since the time Tbird was ready, to actually get it in the hands of first hobbyists, and it may take until a year after Tbird introduction to get it into the hadns of Compaq's customers. These delays are far worse and far more harmful to AMD (in 100s of millions of lost revenue) than the risk of moving away from this waiting game for chipset and motherboard vendors to catch up. Joe