It's not like the DoE are the only ones with expensive computers: The NSA has their underground lair.
There's a commercial need for expensive distributed processing now, with Google/AWS/Azure spending a lot on video cards. The ML craze may make government spending on supercomputers less important, especially if people prefer to distribute calculations on many interchangeable computers.
I wouldn't mind a few NSF or DoD grants for novel computing architectures, but I'm not sure I buy that having massive supercomputers for economic forecasting is "strategically important for nations".
these things get run in phases. phase 1 they insist on novel architectures, rethinking of basic premises, required involvement by academics.
by the end of phase 3 when they actually procure machines its mostly 'just give me one of what you're already shipping, and it better run MPI well'
doe exascale was supposed to be fundamentally different. because by the time you got there all the incremental improvements in power and latency management weren't nearly enough anymore...my guess is that they'll just make big gpu clusters and call it a day
its probably not that much money in the scheme of things, but the sad (or good) depending on your perspective is that HPC got so far out of the mainstream that when corporations finally got around to worrying about scaling, they just did their own thing. so I guess they can thank darpa/doe for infiniband? not really?
Comments
It's not like the DoE are the only ones with expensive computers: The NSA has their underground lair.
There's a commercial need for expensive distributed processing now, with Google/AWS/Azure spending a lot on video cards. The ML craze may make government spending on supercomputers less important, especially if people prefer to distribute calculations on many interchangeable computers.
I wouldn't mind a few NSF or DoD grants for novel computing architectures, but I'm not sure I buy that having massive supercomputers for economic forecasting is "strategically important for nations".
these things get run in phases. phase 1 they insist on novel architectures, rethinking of basic premises, required involvement by academics.
by the end of phase 3 when they actually procure machines its mostly 'just give me one of what you're already shipping, and it better run MPI well'
doe exascale was supposed to be fundamentally different. because by the time you got there all the incremental improvements in power and latency management weren't nearly enough anymore...my guess is that they'll just make big gpu clusters and call it a day
its probably not that much money in the scheme of things, but the sad (or good) depending on your perspective is that HPC got so far out of the mainstream that when corporations finally got around to worrying about scaling, they just did their own thing. so I guess they can thank darpa/doe for infiniband? not really?
and they're still running fortran/mpi