Projects 10012-10085 -> accelerate F@H 100 times?

[Stonecold]

Would it? All 50 WUs could be returned with the same coordinates and still be wrong if the simulation process is flawed. That has happened before with GPU1 and with QMD work units.

I was thinking about the reliability. If the whole simulation itself is flawed, then of course no number of simulations would work.

[Jonazz]

It would be great to get an update surrounding these projects!

[bruce]

It would be great to get an update surrounding these projects!

I think it's safe to assume no such acceleration was possible and the project was abandoned. One important fact: the scientists are MORE interested in solutions that are both accurate and fast than donors are, so they would have jumped at such an opportunity if it actually existed.

[Joe_H]

Besides what bruce has said, the WS these were coming from had problems last year around Thanksgiving and was offline. It came back online for a short while after that, but appears not to have been assigning work for some time now.

[Jonazz]

I think it's safe to assume no such acceleration was possible and the project was abandoned. One important fact: the scientists are MORE interested in solutions that are both accurate and fast than donors are, so they would have jumped at such an opportunity if it actually existed.

Besides what bruce has said, the WS these were coming from had problems last year around Thanksgiving and was offline. It came back online for a short while after that, but appears not to have been assigning work for some time now.

I should have added that these project still appear on the psummary page. I assumed they were still ongoing.

[csvanefalk]

Are these projects now officially dead? Are there any other ongoing ones which seek to improve FaH performance?

[Joe_H]

I would assume so, the servers for these projects at ND have been offline for over 2 years. Don't know if it led to published results or not, have not done a search to see if the principles published any papers on this besides the one mentioned in the project summary. Though that link is dead, but the paper's location might have moved in 3+ years.

As for projects that improve performance, I can't point to any specific ones. But they did release Core_A7 which in testing was run on a clone of a project that had been previously processed using Core_A4. Being coded with a later revision of Gromacs and using SSE2 and AVX instructions where available as compared to SSE for the older core, it does provide a significant improvement in throughput. Not 100x, but at least 20-30% on an older laptop that supported SSE2 on its CPU that I tested on during its initial beta release.