How efficient is Core 21 up 'till now?

[ProDigit]

Is it more efficient than current desktop software?
I was reminded of how back in the day, DirectX improved gaming performance by large margins, on older hardware, through the more efficient software being able to access the hardware better; and MS kept improving DirectX for many years.
Fah core, probably has gone through several revisions.
I was just wondering how efficient it currently is, as it looks to me more like a frame to frame rendering of an image, vs fluid motion video our hardware should be able to produce.

I don't know the details behind the coding, but it seems strange to me, that modern hardware takes such a long time to simulate a thing.

I would think by now, with the right coding, it should be able to fold tens to hundreds of work units in a matter of minutes...

[bruce]

I think you're confusing running the Core_21 for atomic position calculations on the GPU with displaying the images of the those atoms. When the GPU i calculating the positions it's not using the Video Engine components of the GPU which display images, nor it is using DirectX.

FAH does provided a "Viewer" program that produces nice imanges which look nice if you're trying to impress your friends, but it contributes nothing to the science (or the points). All it does is slow down the position calculations by using some of shaders to render the images.

There's a huge difference between calculating the positions of each atom every few ns and rendering them into an image -- adding a background, adding colors, introducing a visual size for each atom, projecting the 3D positions of those atoms onto a 2D image the size of your screen, resolving visual overlap based on distance from the eye ...etc. All of that is unnecessary unless you're going to view them after the positions are calculated.

The FAHViewe program isn't used by those who are in it for the science and/or the points because it wastes resources.

[bruce]

The best way to evaluate the efficienjcy of FAHCore_21 is with a GPU monitoring program (e.g. - like GPU-Z). I think you'll find that the Video engine is running at 0% load, the GPU load is in the 90% range, the memory controller the Bus interfaces are running at some percentages moving data and supplying new work instructions. Essentially, the calculations are running at close to the computational capacity of the shaders.

Hover your mouse over the PerfCap Reason and it will tell you why your GPU can't fold any faster.