Great timing, supercomputer upgrade lead to successful forecast of eruption
news.illinois.edu
news.illinois.edu
Of course, it could be pure hindsight, but the tour guide said that the 2018 eruption was anticipated by observable signs, and that our group was safe because there would be many signs before the next eruption.
I wish I knew about this model at the time so I could have asked a question or two.
Obviously the bureaucracy and management of such a system would be difficult, but it would be an inexpensive way to give these projects additional compute capacity.
I've been on a quest to find some sort of workload or task runner to take advantage of it and have been coming up empty handed.
The issue is profitability. The capex has been spent... now we just need to find people willing to either risk opex or find profitable workloads.
ETC has a 6.5 day deposit time on exchanges, which makes it pretty hard to mine as small miners won't want to wait to get paid. This should come down as the hashrate (and security of the network) increases, however... without much network use (aka: utility)... the only thing to do with it will be to mine and dump.
Thus, the price of all the alts will drop, profitability will go down and cards will turn off. I'm already seeing signs of people turning off their cards in the GPU Miners forums...
We saw a preview of this behavior when Grin came out...
Could this workload be approximated with neural net models and ran on tpus?
Do they actually? Or do they just accelerate matrix multiplications of a very specific size?
I am not really a domain expert for scientific computing, but for example in finite element analysis, the tetrahedral stiffness matrix is 12x12. Presumably the matrices in fluid flow simulations, climate modeling, etc are also modestly sized, and the challenge has more to do with just how many total multiplications there are.
It is not at all clear to me that an accelerated 128x128 multiplication is helpful in these contexts.
So far TPUs have not really proved their worth for general purpose simulations across a wide range of fields. They weren't built for that, and we have other systems that can run the workloads faster, although it's unclear what wouuld happen if you took a really great team and had them work on a hard simulation problem (for example- my previous area was molecular dynamics, and TPUs can do great work on n-body simulations, but not so great on other parts of the force field.
If you have sparse mixed problems then CPUs are still the most cost effective. If you have dense matrix problems GPUs are the most cost effective. If you have problems that don't fit on other systems and you have a good team to optimize to the hardware, TPUs are an option and could be cost effective in principle.
If so, you can represent them as a Nx12x12 "tensor" for some large N (presumably proportional to the number of elements?), and I'm reasonably sure that's within the realm of what TPUs accelerate well.
There are a few steps.
1) discretize the problem, breaking the item into a mesh. Maybe triangles/ tetrahedrons, but maybe also cubes, as the element math is easier.
2) for each element apply a geometric transform to the element stiffness matrix to convert it from a unit stiffness matrix to the global coordinates. Note that this step accounts for the shape of the unit as well.
3) assemble the stiffness matrix by iterating over all the degrees of freedom of all the nodes and adding the contribution from each element stiffness matrix. This will result in a (nodes * dof) Square symmetric matrix, that’s generally sparse and tends to the diagonal. If you’re doing dynamic, the damping and mass matrices need to be assembled as well.
4) Solve the equation using some factorization method, either LU or similar for static, or some Eigen solution for dynamic.
If you’re doing nonlinear/ plastic then repeat generating the stiffness matrix at each iteration.
It’s been years since I did this, but at the time (‘97) I went to a lecture that asserted to that point, since the first cray, the hw improvements and the sw improvements were both about a 10^6 speedup for FEA problems.