Re: Is Intel exceptionally unsuccessful as an architecture designer?

On Sun, 22 Sep 2024 2:13:42 +0000, Lawrence D'Oliveiro wrote:

On Fri, 20 Sep 2024 06:52:07 -0400, Paul A. Clayton wrote:
>

From what I understand, GPUs also typically have
memory controllers optimized for throughput rather than latency, with
larger queue depth.

>
Fine. If they aren’t designed for low latency, then you can’t call them
“interactive”, can you? Since that requires quick response. They seem
more oriented towards batch operation in the background.

Think about it like this::
A GPU can perform 1T-10T calculations per second running at 1GHz.
Try doing that with a CPU.
They are entirely different on the spectrum of design and architecture.
Things that work to make CPUs faster do not make GPUs faster--and for
the most part--vice versa.