[rescue] SGI temps?

[Francisco Javier Mesa-Martinez]

On Mon, 17 Feb 2003, Bjorn Ramqvist wrote:

> On a related discussion, the insane claim that RDRAM is "useless" (quote
> from PeeCee weenies) is that due to its serialized nature, it's intended
> toward different applications uses. SGI, for one thing, couldn't be all
> that unwise to use RDRAMs for texturememory, otherwise it wouldn't be
> there.
> Intel, on the other hand, got this suit-wanking talk-over with Rambus to
> use their proprietary RAM technique, for use as main memory. Without any
> ground-up re-thinking architecturewise, performance would be
> complicated. High bandwidth and high accesstime is completely different
> from low bandwidth and low accesstime.
> I suppose Sony found their use of RDRAMs in the PS2 for other reasons
> than what Intel claimed to do.

I dunno want to be intel defensor by any means... as I already got in hot
waters when I dared to say that the PC (original) architecture when
compared with other architectures of the time wasn't that bad.

But basically I read this message as; "well I dunno why SGI put RDRAM
there, but since it is SGI then they must be correct... however Intel just
don't know what they are doing, so whatever reason they had to use RAMBUS
it must be totally wrong." This is not only wrong, but rather biased.

The reason why intel chose the RAMBUS RDRAM for the original P4 was quite
simple. It was a perfect match for the original P4 bus architecture.
Instead of using a very wide bus (a la SUN) they just used a set of
narrower "fast" buses multiplexed together (this is acutally so that you
can overlap the routing of these narrower buses on parallel layers of the
PCB), thus you can reduce some of the cost of ultrawide bus routings.
RAMBUS memories are quite narrow, and can match the speed of the
sub-buses (both in latency and throughput).

Now, I hate RAMBUS not because of their technology, but rahter because
of the legal crap they are trying to pull with the SDRAM patents.