3 replies [Last post]
Joined: 01/22/2020
Groups: None

I have a CPU usage up to 40% even in idle mode (no script working). I switched off music and sound unchecked all checkboxes in View options but still see about 30% CPU usage when "1794: CRISPR-Cas Transposase Part I: Electron Density" is opened. I tried to move the protein model off the view area, but nothing changes. What could be the reason?

Joined: 12/27/2012
Groups: Beta Folders
Welcome to Foldit...

I'm not sure why, but this is more or less normal behavior in Foldit. Just sitting idle, a Foldit client can use 1 logical processor plus some graphics (GPU). Minimizing the client drops the CPU usage down to a reasonable value, and eliminates the GPU.

The standard advice is to minimize the Foldit window when you're running a recipe, or when you're not actively looking at the protein.

Moving the protein off the screen has been standard advice for a long time, but I'm not sure it helps. Still, when running a recipe, I move the protein off the screen, hide the GUI (shift-G or "Hide GUI" in view options menu), close the behavior pane and any others, and minimize the window. I leave the recipe output window open, but otherwise, it's a clean screen.

My default view option is cartoon thin with EnzDes coloring, and "don't show" checked for sidechains. I have a pet theory that score-related coloring is slower, but zero evidence on that. View options like "pulse when working" and "show sidechains with clashes or exposeds" are off, similar reasoning.

The guess is that Foldit is always obsessively checking the score, even when nothing is happening. Sometimes, when you add a band, you'll get a "good bonding" popup message, even though nothing has changed. In general, all the banding recipe functions seem quite slow. Based on some comments here and there, the band functions manage to trigger some additional score checking, even though none of them can change the shape of the protein on their own. (You need to wiggle for that.)

(See number of processors on the wiki for the "logical processor" thing.)

(edit: better link)

Joined: 01/22/2020
Groups: None
I have 4 logical processors

I have 4 logical processors and they all were busy with some work even though nothing useful was happening. It seems 40% of CPU in idle mode is too much.

Joined: 12/27/2012
Groups: Beta Folders
40% does seem high

I'd expect an idle Foldit would use no more than 25% of your available CPU with four logical processors. (Or 50% with two processors, 12.5% with eight....)

Most of the mysterious Foldit churn happens on one program "thread", which is confined to one logical processor at a given point in time.

For a long time, the rule was one logical processor per Foldit client. After an update in the past couple of years, however, Foldit now uses more than one logical processor when running a recipe. The usage is typically around 1.25 to 1.5 logical processors per client, but it can fluctuate quite a bit for unknown reasons.

I use Microsoft/Sysinternals Process Explorer, which lets you look at how much CPU specific program threads are using. I have several clients running recipes on an eight-processor system, and they seem to be using around 17% of CPU each, so roughly 1.3 logical processors. Looking at one client, three threads are doing most of the work, with one at about 10% CPU, a second near 5%, and a third near 2%. (Sometimes just one thread is using most of the CPU, but usually at least two are active these days, out of 14 or 15 total threads.)

Both Process Explorer and the stock Task Manager have a sortable "CPU" column that let you see who's on top. That lets you be sure it's actually Foldit hogging the CPU. If it is Foldit, we can discuss further, otherwise it's time for some cleanup.

For example, a while back, I noticed (in Process Explorer) that OneDrive was using as much CPU as a Foldit client. Strange, since I don't use OneDrive. I found that you can uninstall one drive, and lots of other things Microsoft throws in, such as Candy Crush Saga, and a weather app. Not sure how much it helped, but I haven't missed them yet.


Developed by: UW Center for Game Science, UW Institute for Protein Design, Northeastern University, Vanderbilt University Meiler Lab, UC Davis
Supported by: DARPA, NSF, NIH, HHMI, Amazon, Microsoft, Adobe, RosettaCommons