Sujet : Re: Is Intel exceptionally unsuccessful as an architecture designer?
De : paaronclayton (at) *nospam* gmail.com (Paul A. Clayton)
Groupes : comp.archDate : 22. Sep 2024, 21:58:10
Autres entêtes
Organisation : A noiseless patient Spider
Message-ID : <vcq0d3$2bl9r$1@dont-email.me>
References : 1 2 3 4
User-Agent : Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101 Thunderbird/91.0
On 9/17/24 8:44 PM, Lawrence D'Oliveiro wrote:
On Tue, 17 Sep 2024 23:45:50 +0000, MitchAlsup1 wrote:
"the CPUs are simply I/O managers to the Inference Engines and GPUs."
That particular Wheel of Reincarnation will never turn that way.
Why? It comes down to RAM. Those addon processors will never have access
to the sheer quantity of RAM that is available to the CPU. And
motherboard-based CPU RAM is upgradeable, as well, whereas addon cards
tend not to offer this option.
My guess would be that CPU RAM will decrease in upgradability.
More tightly integrated memory facilitates higher bandwidth and
lower latency (and lower system power/energy). Integration also
facilitates optimizations by guaranteeing certain aspects. As
fewer means remain available to improve "value", more expensive
and/or less flexible improvements become more attractive.
(There are also marketing reasons for integration.)
The rate at which memory density increases also seems to have
declined. This would imply that buying more memory later would
be less common.
"Direct" access to any memory is not technically impossible.
I/O TLBs were perhaps an early step in reducing the distinction
between I/O agent and CPU agent. Integrated GPUs and other
accelerators also made agents less distinct.