SI
SI
discoversearch

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: Tony Viola who wrote (35441)8/3/1998 12:28:00 PM
From: Elmer  Read Replies (2) | Respond to of 1573941
 
Re: "It sounds like the K7 is following the Alpha roadmap, or lead, as far as architecture and implementation at the server level goes. "

So where will this leave AMD on the desktop? This description of the K7 doesn't make for a cost effective PC solution.

EP



To: Tony Viola who wrote (35441)8/3/1998 9:36:00 PM
From: Adrian Wu  Read Replies (2) | Respond to of 1573941
 
Tony: The K7 will not be just a single chip design but has a tremendous amount of flexibility. Like the Alpha, it can incorporate an L2 cache on a backside bus at half-speed or full speed. In that configuration, it makes use of the cache on the motherboard as an L3. To lower cost, it can omit the L2 from the chip packaging entirely and make use of the cache on the MB as an L2. Most of the cost of the PII is associated with the L2 cache, hence the Celeron. This will not be problem with the K7. The problem with the P6 bus protocol is that it cannot support an L2 cache on the "frontside" bus. The K7 is designed to eventually succeed the K6 as a mainstream processor, much like the PII replaced the Pentium. It should not be any more expensive to manufacture than the Celeron. AMD can also produce a variety with different size and speed of the L2 on the backside bus to satisfy customer needs. This design can also migrate to copper interconnect technologies by the year 2000, enabling them to achieve speeds of 1GHz. Initial K7s will be 0.25 micron, migrating rapidly to 0.18.
Chipsets for the K7 are already available even before the processor itself is ready. DEC is already demonstrating their 21164 implementations. All you need is a new BIOS for the chipset to work with the K7. In terms of economy of scale, please explain why the VIA MVP-3 chipsets sell for $25 when the 440BX chipsets sell for $65?
My guess is that the K7 will eventually be available in a variety of flavors. There will be a plain K7 without L2 on the chip packaging (relying on the MB L2), and probably without the cartridge. There will be a K7 with half speed L2, and a K7 with full speed L2. The Alpha 21164 currently has a full speed on chip L2.

Adrian