MastodonTech.de

Dr. Moritz LehmannInteresting take from an <a href="https://mast.hpc.social/tags/Nvidia" class="mention hashtag" rel="nofollow noopener" target="_blank">#Nvidia</a> engineer I met at <a href="https://mast.hpc.social/tags/ISC25" class="mention hashtag" rel="nofollow noopener" target="_blank">#ISC25</a>: "Do you need IEEE-754 compliant <a href="https://mast.hpc.social/tags/FP64" class="mention hashtag" rel="nofollow noopener" target="_blank">#FP64</a>, or do you need digits? Digits we can get you through FP64 emulation."Not sure what to make of that. Things were an absolute mess before the IEEE-754 standard, and I wouldn't want to ever go back to that. No standards means you cannot at all port software between hardware architectures even from within the same vendor. Having to re-architect software for each and every new chip is not gonna happen.

Dr. Moritz LehmannWith all the hot discussions about <a href="https://mast.hpc.social/tags/FP64" class="mention hashtag" rel="nofollow noopener" target="_blank">#FP64</a> right now, what do you say to Nvidia almost entirely axing FP64 on <a href="https://mast.hpc.social/tags/Blackwell" class="mention hashtag" rel="nofollow noopener" target="_blank">#Blackwell</a> Ultra B300?B200 180GB FP64 performance: 37.2 TFlops/s B300 288GB FP64 performance: 1.2 TFlops/s ☠️They're going from the usual datacenter <a href="https://mast.hpc.social/tags/GPU" class="mention hashtag" rel="nofollow noopener" target="_blank">#GPU</a> FP64:FP32 ratio of 1:2 down to cheap 1:64, like on all their gaming/workstation GPUs. Personally I think this is fantastic - it opens the doors wide for competitors to step into the <a href="https://mast.hpc.social/tags/HPC" class="mention hashtag" rel="nofollow noopener" target="_blank">#HPC</a> market.<a href="https://www.nvidia.com/en-us/data-center/gb300-nvl72/" rel="nofollow noopener" translate="no" target="_blank">https://www.nvidia.com/en-us/data-center/gb300-nvl72/</a>

Andrew Jones (hpcnotes)Provocative question to debate at <a href="https://mast.hpc.social/tags/ISC25" class="mention hashtag" rel="nofollow noopener" target="_blank">#ISC25</a> - in talks, meetings, booths, over beers:What is so special about <a href="https://mast.hpc.social/tags/FP64" class="mention hashtag" rel="nofollow noopener" target="_blank">#FP64</a> that science holds it as a gold standard?If <a href="https://mast.hpc.social/tags/HPC" class="mention hashtag" rel="nofollow noopener" target="_blank">#HPC</a> had never got as far as FP64, eg had got stuck at FP32, surely we would still have been able to do science?

Benjamin Carr, Ph.D. 👨🏻‍💻🧬<a href="https://hachyderm.io/tags/JackDongarra" class="mention hashtag" rel="nofollow noopener" target="_blank">#JackDongarra</a> Makes a Stand for Traditional <a href="https://hachyderm.io/tags/HPC" class="mention hashtag" rel="nofollow noopener" target="_blank">#HPC</a>: "US still doesn’t have a clear, long-term plan for what comes next.... U.S. risks falling behind."Challenges to high-performance computing threaten <a href="https://hachyderm.io/tags/US" class="mention hashtag" rel="nofollow noopener" target="_blank">#US</a> <a href="https://hachyderm.io/tags/innovation" class="mention hashtag" rel="nofollow noopener" target="_blank">#innovation</a>The <a href="https://hachyderm.io/tags/AI" class="mention hashtag" rel="nofollow noopener" target="_blank">#AI</a> boom has led chip makers to focus on <a href="https://hachyderm.io/tags/FP16" class="mention hashtag" rel="nofollow noopener" target="_blank">#FP16</a> and <a href="https://hachyderm.io/tags/FP8" class="mention hashtag" rel="nofollow noopener" target="_blank">#FP8</a>, not the <a href="https://hachyderm.io/tags/FP64" class="mention hashtag" rel="nofollow noopener" target="_blank">#FP64</a> used by scientific research. If chip companies stop making the parts that <a href="https://hachyderm.io/tags/scientists" class="mention hashtag" rel="nofollow noopener" target="_blank">#scientists</a> need, then it could become harder to do important research. <a href="https://theconversation.com/challenges-to-high-performance-computing-threaten-us-innovation-255188" rel="nofollow noopener" translate="no" target="_blank">https://theconversation.com/challenges-to-high-performance-computing-threaten-us-innovation-255188</a>

Andrew Jones (hpcnotes)<a href="https://mast.hpc.social/tags/HPC" class="mention hashtag" rel="nofollow noopener" target="_blank">#HPC</a> and other scientific computing folks should like this image ... it's a floating point 😀 <a href="https://mast.hpc.social/tags/FP64" class="mention hashtag" rel="nofollow noopener" target="_blank">#FP64</a> <a href="https://mast.hpc.social/tags/FLOPS" class="mention hashtag" rel="nofollow noopener" target="_blank">#FLOPS</a>

Benjamin Carr, Ph.D. 👨🏻‍💻🧬<a href="https://hachyderm.io/tags/ElCapitan" class="mention hashtag" rel="nofollow noopener" target="_blank">#ElCapitan</a> Towers Above the <a href="https://hachyderm.io/tags/Top500" class="mention hashtag" rel="nofollow noopener" target="_blank">#Top500</a> in a Big <a href="https://hachyderm.io/tags/HPE" class="mention hashtag" rel="nofollow noopener" target="_blank">#HPE</a> Win using <a href="https://hachyderm.io/tags/AMD" class="mention hashtag" rel="nofollow noopener" target="_blank">#AMD</a> Instinct <a href="https://hachyderm.io/tags/MI300A" class="mention hashtag" rel="nofollow noopener" target="_blank">#MI300A</a> system. With sustained <a href="https://hachyderm.io/tags/HPL" class="mention hashtag" rel="nofollow noopener" target="_blank">#HPL</a> of 1.742EF and peak speed of 2.79EF of <a href="https://hachyderm.io/tags/FP64" class="mention hashtag" rel="nofollow noopener" target="_blank">#FP64</a>, this is a big jump over previous generation systems. El Capitan has a unique architecture as it uses <a href="https://hachyderm.io/tags/APU" class="mention hashtag" rel="nofollow noopener" target="_blank">#APU</a> combining <a href="https://hachyderm.io/tags/CPU" class="mention hashtag" rel="nofollow noopener" target="_blank">#CPU</a> plus <a href="https://hachyderm.io/tags/GPU" class="mention hashtag" rel="nofollow noopener" target="_blank">#GPU</a> onto a package with high-bandwidth memory. Over 44,000 of these MI300A APUs are then packed into the HPE <a href="https://hachyderm.io/tags/Cray" class="mention hashtag" rel="nofollow noopener" target="_blank">#Cray</a> Shasta liquid cooled platform, and connected via the Slingshot interconnect. <a href="https://www.servethehome.com/el-capitan-towers-above-the-top500-in-a-big-hpe-and-amd-win/" rel="nofollow noopener" translate="no" target="_blank">https://www.servethehome.com/el-capitan-towers-above-the-top500-in-a-big-hpe-and-amd-win/</a>

Benjamin Carr, Ph.D. 👨🏻‍💻🧬<a href="https://hachyderm.io/tags/HPE" class="mention hashtag" rel="nofollow noopener" target="_blank">#HPE</a> goes Cray for <a href="https://hachyderm.io/tags/Nvidia" class="mention hashtag" rel="nofollow noopener" target="_blank">#Nvidia</a>'s Blackwell <a href="https://hachyderm.io/tags/GPU" class="mention hashtag" rel="nofollow noopener" target="_blank">#GPU</a>, crams 224 into a single cabinet Announced ahead of <a href="https://hachyderm.io/tags/SuperComputing" class="mention hashtag" rel="nofollow noopener" target="_blank">#SuperComputing</a> conference in Atlanta, Cray's <a href="https://hachyderm.io/tags/EX154n" class="mention hashtag" rel="nofollow noopener" target="_blank">#EX154n</a> platform will support up to 224 Nvidia Blackwell GPUs and 8,064 Grace CPU cores per cabinet. That works out to just over 10 petaFLOPS at <a href="https://hachyderm.io/tags/FP64" class="mention hashtag" rel="nofollow noopener" target="_blank">#FP64</a> for <a href="https://hachyderm.io/tags/HPC" class="mention hashtag" rel="nofollow noopener" target="_blank">#HPC</a> applications or over 4.4 exaFLOPS FP64 for sparse AI and machine learning workloads HPE's new ProLiant servers offer choice of Gaudi, Hopper, or Instinct acceleration <a href="https://www.theregister.com/2024/11/13/hpe_cray_ex/" rel="nofollow noopener" translate="no" target="_blank">https://www.theregister.com/2024/11/13/hpe_cray_ex/</a> <a href="https://hachyderm.io/tags/sc24" class="mention hashtag" rel="nofollow noopener" target="_blank">#sc24</a>

Benjamin Carr, Ph.D. 👨🏻‍💻🧬Happy <a href="https://hachyderm.io/tags/Exascale" class="mention hashtag" rel="nofollow noopener" target="_blank">#Exascale</a> day to all who celebrate and the companies that use it as marketingDoug Eadline has a wonderful piece in HPCWire about how not all exa are equivalent, and the push to brand things with lower precision as <a href="https://hachyderm.io/tags/zetascale" class="mention hashtag" rel="nofollow noopener" target="_blank">#zetascale</a> flies in the face of 30 years of agreeing that FLOPS are <a href="https://hachyderm.io/tags/fp64" class="mention hashtag" rel="nofollow noopener" target="_blank">#fp64</a>, so there is one and only one definition of <a href="https://hachyderm.io/tags/exaFLOPS" class="mention hashtag" rel="nofollow noopener" target="_blank">#exaFLOPS</a>, and its bases on the <a href="https://hachyderm.io/tags/HPLinpack" class="mention hashtag" rel="nofollow noopener" target="_blank">#HPLinpack</a> benchmark.Go Read; In This Club, You Must 'Earn the Exa' <a href="https://www.hpcwire.com/2024/10/17/in-this-club-you-must-earn-the-exa/" rel="nofollow noopener" translate="no" target="_blank">https://www.hpcwire.com/2024/10/17/in-this-club-you-must-earn-the-exa/</a>

Andrew Jones (hpcnotes)Experiment comparing reactions across social media platforms ...One or more of <a href="https://mast.hpc.social/tags/Fortran" class="mention hashtag" rel="nofollow noopener" target="_blank">#Fortran</a> or <a href="https://mast.hpc.social/tags/FP64" class="mention hashtag" rel="nofollow noopener" target="_blank">#FP64</a> or <a href="https://mast.hpc.social/tags/onprem" class="mention hashtag" rel="nofollow noopener" target="_blank">#onprem</a> <a href="https://mast.hpc.social/tags/hpc" class="mention hashtag" rel="nofollow noopener" target="_blank">#hpc</a> will be obsolete by 2030.Ignoring this message counts as agreeing :-)

Rechenkraft.net<a href="https://mstdn.social/tags/PYSCFbeta" class="mention hashtag" rel="nofollow noopener" target="_blank">#PYSCFbeta</a>: Bei <a href="https://mstdn.social/tags/GPUGRID" class="mention hashtag" rel="nofollow noopener" target="_blank">#GPUGRID</a> gibt es eine neue <a href="https://mstdn.social/tags/Linux" class="mention hashtag" rel="nofollow noopener" target="_blank">#Linux</a> (beta) App für <a href="https://mstdn.social/tags/NVIDIA" class="mention hashtag" rel="nofollow noopener" target="_blank">#NVIDIA</a> <a href="https://mstdn.social/tags/GPU" class="mention hashtag" rel="nofollow noopener" target="_blank">#GPU</a>, die aufgrund von "quantum chemistry" Berechnungen <a href="https://mstdn.social/tags/FP64" class="mention hashtag" rel="nofollow noopener" target="_blank">#FP64</a> (double precision) nutzt: <a href="https://www.gpugrid.net/forum_thread.php?id=5429#60963" rel="nofollow noopener" translate="no" target="_blank">https://www.gpugrid.net/forum_thread.php?id=5429#60963</a> <a href="https://mstdn.social/tags/BOINC" class="mention hashtag" rel="nofollow noopener" target="_blank">#BOINC</a> <a href="https://mstdn.social/tags/DC" class="mention hashtag" rel="nofollow noopener" target="_blank">#DC</a> <a href="https://mstdn.social/tags/VC" class="mention hashtag" rel="nofollow noopener" target="_blank">#VC</a> <a href="https://mstdn.social/tags/VolunteerComputing" class="mention hashtag" rel="nofollow noopener" target="_blank">#VolunteerComputing</a> <a href="https://mstdn.social/tags/DistributedComputing" class="mention hashtag" rel="nofollow noopener" target="_blank">#DistributedComputing</a> <a href="https://mstdn.social/tags/CitSci" class="mention hashtag" rel="nofollow noopener" target="_blank">#CitSci</a> <a href="https://mstdn.social/tags/CitizenScience" class="mention hashtag" rel="nofollow noopener" target="_blank">#CitizenScience</a> <a href="https://mstdn.social/tags/Rechenkraft" class="mention hashtag" rel="nofollow noopener" target="_blank">#Rechenkraft</a>

Frühere Suchanfragen

Suchoptionen

Verwaltet von:

Serverstatistik:

#fp64