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: RDM who wrote (47759)	1/29/1999 12:42:00 AM
From: Elmer	Read Replies (2) \| Respond to of 1571691

Re: "Won't a 1 GHz require 2.5 times the bus bandwidth to scale directly from the 400 Mhz case? It seems to me that you may be waiving your hands just a little in stating that 133 Mhz is fast enough for 800 Mhz? No K7 is here yet but it would be outrageous to not consider this problem. The timelines for these speeds are not far off." I didn't say that 133mhz FSB was fast enough for a 800mhz core, although it may be given a large enough L2. Obvisouly as L2 misses drop as a percentage of accesses, the FSB speed becomes less and less a factor. Nonetheless, I think Intel will have a faster FSB to offer by the time 800mhz is available. Re: "My sense is that the K7 will be a $500-1000 chip for the higher performance parts. Thus the pin costs of support chips may not be significant. The 200-400 Mhz speed of the Alpha bus sounds good to me. I think, Intel appologists aside, that it doesn't hurt to have this enormous bug speed. There may be some question as to whether dynamic RAM is available that fast enough to effectively utilize the speed. " I think the idea is that with proper pipelining, Rambus latency can be hidden and the enormous bandwidth can meet the needs of the fastest processors when coupled with very large L2s. AMD will eventually go the Rambus route but Intel has a huge headstart. EP

To: RDM who wrote (47759)	1/29/1999 12:49:00 AM
From: Tenchusatsu	Read Replies (1) \| Respond to of 1571691

<My sense is that the K7 will be a $500-1000 chip for the higher performance parts. Thus the pin costs of support chips may not be significant.> Actually, after having lunch with a couple of co-workers who work in the server motherboard manufacturing department, I get the feeling that the pin cost may be more expensive than people think. Making sure one set of pins runs at a high bus speed is tough enough. Having multiple point-to-point connections could be a big hurdle in high-volume motherboard manufacturability. Remember, the key is high-volume. Even one or two million units per quarter is very high-volume for servers. Tenchusatsu