Japan Tests Silicon for Exascale Computing in 2021 | Hacker News Reader

Japan Tests Silicon for Exascale Computing in 2021

spectrum.ieee.org

63 pointsAliCollins15 comments

For now, the future belongs to the integration of many smaller dies.

https://www.youtube.com/watch?v=G3kGSbWFig4

What are they using it for? Or other ~100 petaflops supercomputers?

"Supercomputers play an important role in the field of computational science, and are used for a wide range of computationally intensive tasks in various fields, including quantum mechanics, weather forecasting, climate research, oil and gas exploration, molecular modeling (computing the structures and properties of chemical compounds, biological macromolecules, polymers, and crystals), and physical simulations (such as simulations of the early moments of the universe, airplane and spacecraft aerodynamics, the detonation of nuclear weapons, and nuclear fusion). Throughout their history, they have been essential in the field of cryptanalysis."

https://en.m.wikipedia.org/wiki/Supercomputer

among other things, to go to Mars

Hi, I work at a supercomputer company.

A short list of things I know our customers use them for:

    - secret spy/tla stuff (no i have no idea what they're used for either)
    - simulating nuclear blasts to know if the bombs we have will work or have issues
    - simulating decay of those bombs (safety first!)
    - simulating weather
    - geological work to find say oil
    - biology simulation for protein folding and understanding things like cancer

Thats all I can think of off the top of my head, but generally, they get used for whenever you have a lot of computation that generates a ton of data and then works on that data.

Lots of good material science also performed.

And don't forget fundamental science, such as cosmology.

Indeed, that list is all I could brain up on the spot is all. >.<

- simulating brains

For a sufficiently broad definition of supercomputer, you'd want to add so many other science and engineering simulation problems: fluid dynamics, controlled and uncontrolled combustion, geophysics and seismology, hydrology, structures and finite-element analysis, crowds and swarms, etc.

Astrophysics is also a close cousin to nuclear weapons research, simulating high energies and pressures, a mixture of nuclear, chemical, and physical interactions, turbulent fluid dynamics, shock waves, etc.

In some sense, we can assume the high end of supercomputing is always applicable to increasing the (spatial or temporal) resolution of any existing grid-based model or the scope/extent of nearly any model (e.g. simulate a larger area/volume/complex structure or system).

Another easy way to soak up computational resources is in coupling of models, e.g. combine your weather or simulation with pollutant tracking, ocean current simulation, gas transfer between the ocean and atmosphere, and so on.

I was watching the documentary HyperNormalization and at 1:12:00 it was mentioned Larry Fink built one of the first "supercomputers" (at least a large system for its time). It was used for risk analysis it was built in 1986. The company was BlackRock and the computer was called Aladdin.

https://youtu.be/-fny99f8amM?t=4320

Also particle physics simulation and data crunching.

https://www.ci.uchicago.edu/research/cern

Rendering electron apps... but seriously, as well as the list of uses someone else pointed out, it's the kind of thing that the more you have the more uses you will find. What's it's called again where demand is static (but not maxed out) until supply increases and then demand increases to compensate and soak up additional supply?

One of the things that amazes me is that it is still not possible to do an ab initio quantum mechanics calculation of the rotational spectrum of even a diatomic molecule with sufficient accuracy to allow identification from measured spectra. Two atoms, and we can't precisely calculate the spectrum. There's lots of room for more computing power.

> ab initio quantum mechanics calculation of the rotational spectrum

That's not a sequence of words my brain was prepared to parse :)

I work on a project related to combustion and fuel efficiency and we need to run thousands to fluid mechanics simulations/FEA on the cluster computing facility in our university. Sometimes they go like weeks to finish, a powerful clsuter machine will definitely speed up everything.

Reply on news.ycombinator.com