<div dir="ltr">Regarding benchmarking real world codes on AMD , every year Martyn Guest presents a comprehensive set of benchmark studies to the UK Computing Insights Conference.<div>I suggest a Sunday afternoon with the beverage of your choice is a good time to settle down and take time to read these or watch the presentation.</div><div><br></div><div>2019</div><div><a href="https://www.scd.stfc.ac.uk/SiteAssets/Pages/CIUK-2019-Presentations/Martyn_Guest.pdf">https://www.scd.stfc.ac.uk/SiteAssets/Pages/CIUK-2019-Presentations/Martyn_Guest.pdf</a><br></div><div><br></div><div><br></div><div>2020 Video session</div><div><a href="https://ukri.zoom.us/rec/share/ajvsxdJ8RM1wzpJtnlcypw4OyrZ9J27nqsfAG7eW49Ehq_Z5igat_7gj21Ge8gWu.78Cd9I1DNIjVViPV?startTime=1607008552000">https://ukri.zoom.us/rec/share/ajvsxdJ8RM1wzpJtnlcypw4OyrZ9J27nqsfAG7eW49Ehq_Z5igat_7gj21Ge8gWu.78Cd9I1DNIjVViPV?startTime=1607008552000</a></div><div><br></div><div>Skylake / Cascade Lake / AMD Rome</div><div><br></div><div>The slides for 2020 do exist - as I remember all the slides from all talks are grouped together, but I cannot find them.</div><div>Watch the video - it is an excellent presentation.</div><div><br></div><div><br></div><div><br></div><div><br></div><div><br></div><div><br></div><div><br></div><div><br></div><div><br></div><div><br></div><div><br></div><div><br></div><div><br></div><div><br><div><br></div><div><br></div><div><br></div></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Sat, 19 Jun 2021 at 16:49, Gerald Henriksen <<a href="mailto:ghenriks@gmail.com">ghenriks@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">On Wed, 16 Jun 2021 13:15:40 -0400, you wrote:<br>
<br>
>The answer given, and I'm <br>
>not making this up, is that AMD listens to their users and gives the <br>
>users what they want, and right now they're not hearing any demand for <br>
>AVX512.<br>
><br>
>Personally, I call BS on that one. I can't imagine anyone in the HPC <br>
>community saying "we'd like processors that offer only 1/2 the floating <br>
>point performance of Intel processors".<br>
<br>
I suspect that is marketing speak, which roughly translates to not<br>
that no one has asked for it, but rather requests haven't reached a<br>
threshold where the requests are viewed as significant enough.<br>
<br>
> Sure, AMD can offer more cores, <br>
>but with only AVX2, you'd need twice as many cores as Intel processors, <br>
>all other things being equal.<br>
<br>
But of course all other things aren't equal.<br>
<br>
AVX512 is a mess.<br>
<br>
Look at the Wikipedia page(*) and note that AVX512 means different<br>
things depending on the processor implementing it.<br>
<br>
So what does the poor software developer target?<br>
<br>
Or that it can for heat reasons cause CPU frequency reductions,<br>
meaning real world performance may not match theoritical - thus easier<br>
to just go with GPU's.<br>
<br>
The result is that most of the world is quite happily (at least for<br>
now) ignoring AVX512 and going with GPU's as necessary - particularly<br>
given the convenient libraries that Nvidia offers.<br>
<br>
> I compared a server with dual AMD EPYC >7H12 processors (128)<br>
> quad Intel Xeon 8268 >processors (96 cores).<br>
<br>
> From what I've heard, the AMD processors run much hotter than the Intel <br>
>processors, too, so I imagine a FLOPS/Watt comparison would be even less <br>
>favorable to AMD.<br>
<br>
Spec sheets would indicate AMD runs hotter, but then again you<br>
benchmarked twice as many Intel processors.<br>
<br>
So, per spec sheets for you processors above:<br>
<br>
AMD - 280W - 2 processors means system 560W<br>
Intel - 205W - 4 processors means system 820W<br>
<br>
(and then you also need to factor in purchase price).<br>
<br>
>An argument can be made that for calculations that lend themselves to <br>
>vectorization should be done on GPUs, instead of the main processors but <br>
>the last time I checked, GPU jobs are still memory is limited, and <br>
>moving data in and out of GPU memory can still take time, so I can see <br>
>situations where for large amounts of data using CPUs would be preferred <br>
>over GPUs.<br>
<br>
AMD's latest chips support PCI 4 while Intel is still stuck on PCI 3,<br>
which may or may not mean a difference.<br>
<br>
But what despite all of the above and the other replies, it is AMD who<br>
has been winning the HPC contracts of late, not Intel.<br>
<br>
* - <a href="https://en.wikipedia.org/wiki/Advanced_Vector_Extensions" rel="noreferrer" target="_blank">https://en.wikipedia.org/wiki/Advanced_Vector_Extensions</a><br>
_______________________________________________<br>
Beowulf mailing list, <a href="mailto:Beowulf@beowulf.org" target="_blank">Beowulf@beowulf.org</a> sponsored by Penguin Computing<br>
To change your subscription (digest mode or unsubscribe) visit <a href="https://beowulf.org/cgi-bin/mailman/listinfo/beowulf" rel="noreferrer" target="_blank">https://beowulf.org/cgi-bin/mailman/listinfo/beowulf</a><br>
</blockquote></div>