jfkfif

Sandia is operated by a government contractor (Honeywell)

use spack.io instead

More performant than using HDF5?

I am working on something super similar. If you remember the name of the startup I would appreciate it deeply

Depending on the size of your application you can calculate flops by hand

https://docs.nersc.gov/tools/performance/roofline/

In the desert? It’s the driest place in the world

this good enough for you?

https://www.amd.com/en/newsroom/press-releases/2024-5-21-amd...

It is not a reasonable assumption to compare your local cluster to the largest clusters within DOE or their equivalents in Europe/Japan. These machines regularly run at >90% utilization and you will not be given an allocation if you can’t prove that you’ll actually use the machine.

I do see the phenomenon you describe on smaller university clusters, but these are not power users who know how to leverage HPC to the highest capacity. People in DOE spend their careers working to use as much as these machines as efficiently as possible.

DOE already mandates this

osti.gov

Econ has a big working paper culture

PFAS is able to be removed via reverse osmosis which direct potable reuse facilities utilize, so not really a problem

nvidia absolutely gives deals to national labs and universities. See Crossroads @ LANL, Isambard in the UK, Perlmutter @ LBL. While AMD is being deployed at LLNL and ORNL, Nvidia isn’t done with their HPC game. Maybe not at the leadership level, but we’ll see how Oak Ridge and LANL decide their next round of procurements

Academic departments with low budgets and cheap electricity who can make due with old CPUs

the problem is multinode runs that communicate through the network

HPC, or scientific computing more generally, are slow when it comes to adopting modern software development practices. Most researchers used to only care about results above all, rather than reliability or reproducibility for other users. DOE initiatives [0] in the past couple years have done a lot to change this, though there are significant hurdles from a security and logistical standpoint [1]. For instance, to uncover issues in a PR, you might need to run the code at >64 nodes on a shared system.

[0] https://ecp-ci.gitlab.io/docs/admin/jacamar/introduction.htm...

[1] https://arxiv.org/abs/2303.17034

What’s the company?

LLNL has ported most if not all of their codes to C++, but I think LANL is still full on Fortran

please email us at hpccenter@stanford.edu, and ask to be put in touch with someone at PSAAP. I would be happy to help!

I really like Legion as an alternative for HPC applications, among many others (https://legion.stanford.edu/)