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: kash johal who wrote (51354)	3/1/1999 3:47:00 PM
From: Tenchusatsu	Respond to of 1571401

Like I said before, point-to-point is different, not necessarily better or worse. Intel is looking into point-to-point topologies (e.g. NGIO) in order to reduce the pin count and the number of wires/traces. Compaq will take the point-to-point bus of the Alpha even further in the 21364 and integrate an RDRAM controller right onto the CPU. AMD is doing neither. At least not yet. Even for a dual-CPU system, two point-to-point connections is more complex than a single dual-processor bus. In Intel's case, chipsets like the 440BX can be soldered onto a single or a dual CPU motherboard without change. This leads to relatively inexpensive dual-processor workstations. AMD, however, will either need a separate chipset to support dual-processor K7 systems, or they'll have to stick with one dual-processor chipset and disable one of the P2P ports for single-processor motherboards. This isn't necessarily a bad thing, though. Perhaps AMD feels confident that a dual-K7 system can blow away a dual-Xeon system, so the extra complexity may be justified. Or like I suggested above, AMD might have other K7-derivatives in the works, like one which integrates a memory controller. But in its current form, I really can't see the advantages behind multiple point-to-point connections. Even Intel is going to handle the limitations of the multiprocessor bus speed with large L2 caches. If the K7 can support up to 8 MB of L2 cache (a limit that will probably never be reached in practice), then putting the K7 on a multiprocessor bus will not hurt its performance that much. Tenchusatsu