eupolicy.social is one of the many independent Mastodon servers you can use to participate in the fediverse.
This Mastodon server is a friendly and respectful discussion space for people working in areas related to EU policy. When you request to create an account, please tell us something about you.

Server stats:

223
active users

#fp64

0 posts0 participants0 posts today
Dr. Moritz Lehmann<p>Interesting take from an <a href="https://mast.hpc.social/tags/Nvidia" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>Nvidia</span></a> engineer I met at <a href="https://mast.hpc.social/tags/ISC25" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>ISC25</span></a>: "Do you need IEEE-754 compliant <a href="https://mast.hpc.social/tags/FP64" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>FP64</span></a>, or do you need digits? Digits we can get you through FP64 emulation."</p><p>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.</p>
Dr. Moritz Lehmann<p>With all the hot discussions about <a href="https://mast.hpc.social/tags/FP64" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>FP64</span></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">#<span>Blackwell</span></a> Ultra B300?</p><p>B200 180GB FP64 performance: 37.2 TFlops/s<br>B300 288GB FP64 performance: 1.2 TFlops/s ☠️</p><p>They're going from the usual datacenter <a href="https://mast.hpc.social/tags/GPU" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>GPU</span></a> FP64:FP32 ratio of 1:2 down to cheap 1:64, like on all their gaming/workstation GPUs.<br>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">#<span>HPC</span></a> market.</p><p><a href="https://www.nvidia.com/en-us/data-center/gb300-nvl72/" rel="nofollow noopener" translate="no" target="_blank"><span class="invisible">https://www.</span><span class="ellipsis">nvidia.com/en-us/data-center/g</span><span class="invisible">b300-nvl72/</span></a></p>
Andrew Jones (hpcnotes)<p>Some great debate already in response to my question on <a href="https://mast.hpc.social/tags/FP64" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>FP64</span></a>.</p><p>But are we just defending FP64 because it is what we have now and what we know?</p><p>Would we defend FP32 if that is all we had?</p><p>If we had pervasive FP128 would we be arguing that a drop to FP64 is not good enough?</p><p><a href="https://mast.hpc.social/tags/ISC25" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>ISC25</span></a> <a href="https://mast.hpc.social/tags/HPC" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>HPC</span></a></p>
Andrew Jones (hpcnotes)<p>Provocative question to debate at <a href="https://mast.hpc.social/tags/ISC25" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>ISC25</span></a> - in talks, meetings, booths, over beers:</p><p>What is so special about <a href="https://mast.hpc.social/tags/FP64" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>FP64</span></a> that science holds it as a gold standard?</p><p>If <a href="https://mast.hpc.social/tags/HPC" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>HPC</span></a> had never got as far as FP64, eg had got stuck at FP32, surely we would still have been able to do science?</p>
Benjamin Carr, Ph.D. 👨🏻‍💻🧬<p><a href="https://hachyderm.io/tags/JackDongarra" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>JackDongarra</span></a> Makes a Stand for Traditional <a href="https://hachyderm.io/tags/HPC" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>HPC</span></a>: "US still doesn’t have a clear, long-term plan for what comes next.... U.S. risks falling behind."</p><p>Challenges to high-performance computing threaten <a href="https://hachyderm.io/tags/US" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>US</span></a> <a href="https://hachyderm.io/tags/innovation" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>innovation</span></a></p><p>The <a href="https://hachyderm.io/tags/AI" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>AI</span></a> boom has led chip makers to focus on <a href="https://hachyderm.io/tags/FP16" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>FP16</span></a> and <a href="https://hachyderm.io/tags/FP8" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>FP8</span></a>, not the <a href="https://hachyderm.io/tags/FP64" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>FP64</span></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">#<span>scientists</span></a> need, then it could become harder to do important research. <br><a href="https://theconversation.com/challenges-to-high-performance-computing-threaten-us-innovation-255188" rel="nofollow noopener" translate="no" target="_blank"><span class="invisible">https://</span><span class="ellipsis">theconversation.com/challenges</span><span class="invisible">-to-high-performance-computing-threaten-us-innovation-255188</span></a></p>
Andrew Jones (hpcnotes)<p><a href="https://mast.hpc.social/tags/HPC" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>HPC</span></a> and other scientific computing folks should like this image ... it's a floating point 😀 </p><p><a href="https://mast.hpc.social/tags/FP64" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>FP64</span></a> <a href="https://mast.hpc.social/tags/FLOPS" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>FLOPS</span></a></p>
Andrew Jones (hpcnotes)<p>Experiment comparing reactions across social media platforms ...</p><p>One or more of <a href="https://mast.hpc.social/tags/Fortran" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>Fortran</span></a> or <a href="https://mast.hpc.social/tags/FP64" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>FP64</span></a> or <a href="https://mast.hpc.social/tags/onprem" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>onprem</span></a> <a href="https://mast.hpc.social/tags/hpc" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>hpc</span></a> will be obsolete by 2030.</p><p>Ignoring this message counts as agreeing :-)</p>