Performance impact of showing protein folding

[kingsapo]

Hey all! Just had a quick question. I'm running F@H on some older computers of mine, and wanted to show the simulations being run on-screen. My question is does showing the folding significantly impact performance? I would think the computer's CPU would do the simulations while the GPU shows the protein. But if showing the protein being folded will make my work units take much longer, then I really wouldn't want to do it. I would be ok with a little extra time, but of we're talking extra hours or days for showing the protein then I'll just turn it off. Thanks!

[Jesse_V]

Welcome to the forum kingsapo!

Yes there will be a performance impact showing the simulations. I think it's a generally good idea to turn them off when you aren't actually wanting to look at them. And what client are you using by the way? Do you know that you can easily fold on your GPU?

[Stonecold]

I think the Viewer for the v7 beta client will turn off (or at least slow down) after a certain amount of time for that very reason, so if you leave and the Viewer is still running, it'll stop itself.

[kingsapo]

Jesse_V,

Thanks, I kinda figured there would be an impact, so I'll probably leave it off. As for clients, I'm using two older computers that are barely running XP, as well as some Macs running Tiger. I know about the high-performance GPU clients, but given the fact that my computers are all around 10 years old, I think the older clients would be better for me; right?

BTW, LOVE your signature.


Stonecold, thanks, I just saw your response, I'll also give that beta client a look.

[Stonecold]

Stonecold, thanks, I just saw your response, I'll also give that beta client a look.

The beta client is still early in the beta stage when it comes to the Viewer. It only displays a demo protein and often the image is untexturized and black. If I'm right, it only displays the actual protein being folded if the core writes .xyz files to the work directory, and only older cores still do that.

One thing I use that is very helpful is Jmol, a light Java application that is a molecular visualizer. Jmol is is much more customizable and uses MUCH less CPU time than the native FAH viewer, so it won't impact the progress rate very much. It uses nearly 0% CPU when the molecule is not being manipulated, rotated, magnified, etc. and only uses up to 10% when you are actively using it (so you don't have to remember to exit the application when you're done). I think (but I'm not sure) it also takes advantage of hardware acceleration with the GPU, so that could be one reason it's not very taxing on the CPU. Here is the Jmol website, here is the Folding@home Jmol website, and here is the post talking about Jmol. Use this link to go to the downloads page.

To visualize the project using Jmol, you first have to download the .zip file under the downloads section (above), extract the folder to wherever you want (there is no installation), and run "Jmol.jar" (you have to have Java installed to run .jar programs). In the window that opens up, press "File" and go to "Open URL". In the window that opens, type "http://jmol.sourceforge.net/fah/projects/p[project #].xyz.gz". Remember to replace "[project #]" with the project number you are folding. Hope this helps!

[7im]

Actually, the V7 viewer is a huge improvement over the v6 viewer.

And when the V7 viewer is adapted to view all proteins, without the .xyz file, it will be quite an asset.

But it will always be processor intensive to draw the pictures in motion, as is any graphically oriented application.

[Stonecold]

But it will always be processor intensive to draw the pictures in motion, as is any graphically oriented application.

Right, and Jmol won't do that, but it is better than the v7 Viewer in its current state.

[7im]

The viewer isn't a priority for me. Not until the fahcores get updated to actually support it.

[Stonecold]

The viewer isn't a priority for me. Not until the fahcores get updated to actually support it.

Are the cores all that need to be fixed for the Viewer to work correctly?

[bruce]

Some cores do produce .xyz files and the viewer does visualize them in motion. That does take a lot of processing power but it was done successfully by Sony on the PS3 and the Windows folks wanted the same sort of viewer, forgetting that the PS3 was specifically designed for full-motion games. Folks who run Windows want to be able to fold with both their CPU (typically the SMP client or slot) and their GPU, leaving no processing power to compute the video of the protein. If there's no .xyz file, you'll see the demo protein.

The viewer does require good GPU driver support. If you don't have it, you'll fail to see the colors and texture. The viewer can display somewhat less spectacular images if you specify the --basic option. Apparently that restores the color for those of us with limited driver support.

I (almost) never use the viewer. As far as I'm concerned, it's only useful if I'm trying to recruit somebody new to FAH. Since I have a PS3, I do my recruiting by showing them the PS3 viewer rather than the Windows viewer. Better still, I sometimes show them the video in paper #53 or one of the other youtube videos.

IMHO, the viewer consumes too much of my resources that could be used for the actual folding process. Others are welcome to different opinions.

[kingsapo]

Thanks for the responses guys! It sounds like I'll give JMol a try, and keep my eyes open for when the v7 client goes release. Thanks again!