Path: ...!eternal-september.org!feeder3.eternal-september.org!news.eternal-september.org!.POSTED!not-for-mail
From: Lawrence D'Oliveiro <ldo@nz.invalid>
Newsgroups: comp.arch
Subject: Re: Is Intel exceptionally unsuccessful as an architecture designer?
Date: Thu, 19 Sep 2024 04:27:16 -0000 (UTC)
Organization: A noiseless patient Spider
Lines: 26
Message-ID: <vcg973$edcv$1@dont-email.me>
References: <memo.20240913205156.19028s@jgd.cix.co.uk>
	<vcd3ds$3o6ae$2@dont-email.me>
	<2935676af968e40e7cad204d40cafdcf@www.novabbs.org>
	<vcd7pr$3op6a$3@dont-email.me> <7wCGO.45461$xO0f.1783@fx48.iad>
	<20240918190414.00005806@yahoo.com>
	<8e1aed9ce25c70cc555731140ae14eb1@www.novabbs.org>
	<vcfln9$836k$1@dont-email.me>
	<4409967ac141a7ee932f2ebd8894ca19@www.novabbs.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Injection-Date: Thu, 19 Sep 2024 06:27:16 +0200 (CEST)
Injection-Info: dont-email.me; posting-host="2c7e13dca6d3b404af04090a5ecd8e97";
	logging-data="472479"; mail-complaints-to="abuse@eternal-september.org";	posting-account="U2FsdGVkX19W+0MNNXm8Ul+MqmwBPMgW"
User-Agent: Pan/0.160 (Toresk; )
Cancel-Lock: sha1:90qj4/vBQO49IlfSjp6mArahhjY=
Bytes: 2365

On Thu, 19 Sep 2024 00:29:09 +0000, MitchAlsup1 wrote:

> On Wed, 18 Sep 2024 22:54:33 +0000, Lawrence D'Oliveiro wrote:
> 
>> On Wed, 18 Sep 2024 16:23:01 +0000, MitchAlsup1 wrote:
>>
>>> On the other hand, and this is where the deprecation of the CPUs come
>>> in, The engines consuming the data are bandwidth machines {GPUs and
>>> Inference engines} which are quite insensitive to latency (they are
>>> not not latency bound machines like CPUs).
>>>
>>> When doing GPUs, a memory access taking 400 cycles would hardly
>>> degrade the overall GPU performance--while it would KILL any typical
>>> CPU architecture.
>>
>> But if it’s supposed to be for “interactive” use, it’s still going to
>> take those 400 memory-cycle times to return a response.
> 
> That is why you use the CPU for human interactions and bandwidth
> engines for the muscle.

But then those bandwidth engines become the interactivity bottleneck, 
don’t they?

Unless you use them only for precomputing stuff in some kind of batch mode 
for later use, rather than doing processing on-demand.