bob wrote: ↑Thu Oct 03, 2019 4:08 am
For the fastest machines around, the top two are at Oak Ridge Lab and Lawrence Livermore Labs. The NEXT two are in China...
These are the machines we know about. I'm pretty sure the NSA won't tell us where the faster machines are, and what they're already able to do.
Ever wondered what a 'Department of Energy' actually does?
The U.S. will join probably 2021 with Aurora (Intel) and Frontier (AMD).
I guess most of them will use some kind of heterogeneous computing, a CPU/GPU mix.
--
Srdja
If true it looks like Moore's law is being surpassed when it comes to supercomputers.
...
The actual Moore's law is about doubling transistor count every two years on a chip,
no word about super computers.
Considering the new types of cpus, AMD is already 'cheating' in this context, they
put up to 8 chiplets with 8 cores each on a cpu.
Since IBMs Roadrunner from 2008 (1 Petaflop) or so, HPC uses a heterogeneous computing
approach with accelerators beside CPUs, these are harder to program, but deliver more
instruction throughput.
bob wrote: ↑Thu Oct 03, 2019 4:08 am
For the fastest machines around, the top two are at Oak Ridge Lab and Lawrence Livermore Labs. The NEXT two are in China...
That is when you count flops, the top two use a bunch of NVidia GV100 racks, maybe nice for a NN engine. The new Chinese Tianhe-3 will be operational next year and is going to surpass the IBM Summit by a factor of 5 to 6.
One bird in the hand is better than 2 in the bush.
dannyb wrote: ↑Sat Oct 05, 2019 11:33 am
Could a chess engine running on a quantum computer be developed? I've read that probably the Grover's algorithm would be used.
Given Google 's history of faking scientific milestones (look AlphaZero), and engaging with pseudo-science, I have little surprise that most experts are extremely sceptical about Google' s latest PR campaign.
Astatos wrote: ↑Sat Oct 26, 2019 4:53 pm
Given Google 's history of faking scientific milestones (look AlphaZero), and engaging with pseudo-science, I have little surprise that most experts are extremely sceptical about Google' s latest PR campaign.
Perhaps I missed something but what about AlphaZero was "faked"?