3 ms·

Are people still doing science on Intel silicon? Why? What is it exactly that an Intel-based "supercomputer" can do that a hangar full of NVidia GPU's can't?

Are people still doing science on Intel silicon? Why? What is it exactly that an Intel-based "supercomputer" can do that a hangar full of NVidia GPU's can't? Looking to understand here, I thought GPU's had changed the HPC landscape for ever. [EDIT]: Are they workload where the I/O infrastructure required to feed the compute units and that on only Intel can deliver?

zokier 4y ago
> What is it exactly that an Intel-based "supercomputer" can do that a hangar full of NVidia GPU's can't? Run 30 year old fortran code?
atemerev 4y ago
Because not every scientific library in the world is ported to GPGPU (or even can technically be). CPU-heavy computing is a rule, not an exception. E.g. in computational fluid dynamics, there are not too many GPU-ready codes available, and most of those are either beta-quality or already abandoned.
- jhbadger 4y ago
  Agreed. In computational biology it is similar. About 5-10 years ago there was huge excitement over GPUs and there were things like GPU-accelerated BLAST (program for comparing DNA/protein sequences against each other) and the like, but it really didn't revolutionize the field in general. HPCs like Biowulf (the HPC cluster used by the US National Institutes of Health, a pun on "biology" and "Beowulf") is overwhelmingly CPU-based with a few GPUS that you can request.
  - Maursault 4y ago
    > Biowulf (the HPC cluster used by the US National Institutes of Health, a pun on "biology" and "Beowulf") Sorry to be satanic[0], but "Biowulf"[1] is not a pun. Thieves often make this mistake because they take things literally. [0] https://en.wikipedia.org/wiki/Pun https://en.wikipedia.org/wiki/Pun [1] https://en.wikipedia.org/wiki/Portmanteau https://en.wikipedia.org/wiki/Portmanteau
JonChesterfield 4y ago
Aurora was supposed to be intel GPU for most of the performance. And in general supercomputer means very fast network between nodes as otherwise most of the time is spent in communication. And finally it's amdgpu holding the performance crown for hpc, not nvidia.
sampo 4y ago
> What is it exactly that an Intel-based "supercomputer" can do that a hangar full of NVidia GPU's can't? 1. Crunch problems that are parallel but not "embarrassingly parallel", thus requiring a fast network to communicate between nodes. 2. Crunch problems that have not been ported from CPU code the GPU code.

zokier 4y ago

> What is it exactly that an Intel-based "supercomputer" can do that a hangar full of NVidia GPU's can't? Run 30 year old fortran code?

atemerev 4y ago

Because not every scientific library in the world is ported to GPGPU (or even can technically be). CPU-heavy computing is a rule, not an exception. E.g. in computational fluid dynamics, there are not too many GPU-ready codes available, and most of those are either beta-quality or already abandoned.

jhbadger 4y ago
Agreed. In computational biology it is similar. About 5-10 years ago there was huge excitement over GPUs and there were things like GPU-accelerated BLAST (program for comparing DNA/protein sequences against each other) and the like, but it really didn't revolutionize the field in general. HPCs like Biowulf (the HPC cluster used by the US National Institutes of Health, a pun on "biology" and "Beowulf") is overwhelmingly CPU-based with a few GPUS that you can request.
- Maursault 4y ago
  > Biowulf (the HPC cluster used by the US National Institutes of Health, a pun on "biology" and "Beowulf") Sorry to be satanic[0], but "Biowulf"[1] is not a pun. Thieves often make this mistake because they take things literally. [0] https://en.wikipedia.org/wiki/Pun https://en.wikipedia.org/wiki/Pun [1] https://en.wikipedia.org/wiki/Portmanteau https://en.wikipedia.org/wiki/Portmanteau

JonChesterfield 4y ago

Aurora was supposed to be intel GPU for most of the performance. And in general supercomputer means very fast network between nodes as otherwise most of the time is spent in communication. And finally it's amdgpu holding the performance crown for hpc, not nvidia.

sampo 4y ago

> What is it exactly that an Intel-based "supercomputer" can do that a hangar full of NVidia GPU's can't? 1. Crunch problems that are parallel but not "embarrassingly parallel", thus requiring a fast network to communicate between nodes. 2. Crunch problems that have not been ported from CPU code the GPU code.