https://threadreaderapp.com/thread/1575233618170982400.html
2 replies (most recent on top)
But did you see that 18-yr-old Data Science Masters grad from Harvard (Extension School) who is Intel's Chief AI Ethics Designer? Jim Keller, eat your heart out. We don't need you. We got Ria!
Intel Failed Science 🧵 (1/13)
Because I am not one of the legions of post-docs with funding (or even immigration status) tied to the functionality of a supercomputer that is now years late—a system that was to form the core of my early career—I can speak.
Because I am not working at one of the national laboratories with the only functional long-promised exascale supercomputer in the U.S waiting for my delayed chance to use all or part of an oversubscribed machine—I can speak.
Most of you can’t, at least not publicly. And that’s okay, everyone gets it. You have careers, after all.
But it’s just too painful to watch you sit mutely, politely sit by and pretend this is not a deep embarrassment for your lab and the U.S. and its over the top supercomputing competitiveness advocates.
The worst thing you can do right now is pretend that Intel’s series of delays—not just of an exascale system but your scientific progress are acceptable. But what can you do?
I don’t know. But I do know the worst thing Intel could do—did do—this week: host a glittering event that heralded an era of sparkling, Moores Law defying technical progress that for people in the HPC camp found painfully tone deaf.
(Almost as tone deaf as bragging about zettascale ambitions earlier this year when the long-promised exascale machine was still not built and Sapphire Rapids delivery dates were still unknown.)
Yes, Intel is shipping to Argonne now. Great. How much of it? All of it at once? And in time for the lab to run LINPACK and get it placed on the Top 500 list to show the world the U.S. actually can deliver on its promise of continued top supercomputing prowess? Great but late.
This year at the annual Supercomputing Conference, we will all know that while a U.S. supercomputer will likely take top slot(s), the Gordon Bell prize submissions were crunched on China’s homegrown supercomputers, which appear to be more numerous that what the U.S. possesses.
The U.S. should have had two exascale supercomputers on the Top 500 in 2021. And Aurora, which has been plagued by Intel’s delays since 2017, should have been first—well before “Frontier” at Oak Ridge National Lab appeared. But Intel dropped the ball over. And over. And again.
Enough of you reached out to share the off-the-record anger over delays of what was to be the crowning achievement in technical progress in the U.S. anger at postdoc and exascale computing software and application work that has been put on hold until the machine is functional.
So much science put on hold or shifted to other machines. And not one of you is allowed to say anything publicly. That’s a shame. I’ll say it for you so people understand it’s far more than a processor delay.
So go glitter and glam at Intel Innovation. Big supercomputers and grand-scale science aren’t as se-y as the latest desktop CPUs. But life saving/changing science happens on these machines and careers rely on them. This has been a cascading failure too long unstated.