Folding Forum

Hey everyone, long time user of v6 but after upgrading to v7 I'm completely lost, though I must say it's nice to see that you all have incorporated both CPU and GPU into one client.

I've re-installed v7, and I'm currently running v7 in uniprocessor mode to keep noise to a minimum.
Q: I would like to add my GPU for some more folding horsepower, but is there any way to adjust the % usage of my GPU?

I would run both my CPU and GPU at 100% if the noise from my PWM fans wasn't so distracting! Ideally I want 25% usage for my CPU and my GPU, maybe 50% if fan noise cooperates.

CPU: i5-2500K, stock speed
GPU: MSI 560ti 448-core Twin Frozr III PE/OC

A control for GPU percentage is on the requested features list for the V7 client, but currently most/all GPU fahcores do not support this. The setting to do this is currently inactive and suppressed from showing in the V7 client until processing cores are updated to allow setting a % usage. Indirectly you can slow down the GPU client by adjusting the CPU percentage associated with the folding slot. This works more or less well depending on which core is in use by "starving" the GPU of data and instructions to process.

Welcome to foldingforum.org, spartanerik.

The simplest way to adjust the SMP slot is to adjust the number of threads. With an i5, rather than running smp:4, you can set it to smp:3 or smp:2 or Uniprocessor. If you really want 25%, don't run SMP, run Uni.

Adjusting the GPU is a work-in-progress. The adjustments actually take place in the FahCore, not in the client, but during the testing of V7 we discovered that although V6 donors were adjusting FAH_GPU_IDLE, it only worked with some of the GPU FahCores. Stanford's development department is working on fixing those FahCores, but I can't promise that they'll work yet. If you already have the environment variable set, you'll still have the same capabilities that you had in V6. and the situation should improve whenever the new FahCore is ready to be distributed.

Ah good to know, I'll check back on the forum regularly for updates. Would be very glad to see this implemented in the near future; thanks for the welcome!

One of the best ways to reduce the fan noise is to improve air circulation. More fans and bigger, slower turning fans are are almost always quieter, and they're generally easy to add.

bruce wrote:One of the best ways to reduce the fan noise is to improve air circulation. More fans and bigger, slower turning fans are are almost always quieter, and they're generally easy to add.

Airflow isn't a problem, the side panel to my case (NZXT Vulcan) is mesh, and I've got 2x 120mm exhaust fans on top, and a 120mm intake fan in the front. However, I am still using the stock cooler for my i5-2500K which explains the noise (and poor cooling for the CPU when under load).

Once the cooler I'm looking at goes on sale on Newegg I'll purchase that as a replacement.

Mesh side panels can look great, but may not be any help in cooling. Some of the best cooling cases have all solid panels. What is more important is proper airflow over the heat sinks and exhausting the hot air from the case. Those top mounted exhaust fans can help, but only if the air has gone past the CPU and GPU coolers first. Otherwise you can be just pulling air in one place and sending it out without ever actually providing much cooling.

Well I plan on getting the Enermax ETD-T60-TB, a down-flow style cooler. Traditional tower coolers are a bit too tall for my build, and the rear of my case only has space for an 80mm/92mm fan which isn't all that effective.

The 120mm fan on the Enermax will pull in air from the mesh side panel. This can also be helped once I mount a 200mm intake fan onto the mesh (has rubber grommets for screws); in this mATX build it's a bit difficult to get focused directional airflow, so I'm just having heat exhaust upwards while cold air runs in from the front and the side

To the original question - is there a way to adjust GPU usage in V7? The answer is yes, much of the time. But, it is dependent on the WU and maybe the core.

My problem child is my laptop - it has a GTX560M GPU, which does quite well with folding the Core 15 WUs that are served. But, being a laptop, there is only so much internal or external cooling that can be done, even with an aggressive laptop cooler base to sit the laptop on. Doing nothing results in the GPU temp hitting 93 or 94, much higher that I want on the laptop.

If you are running Windows, there is an environment variable "FAH_GPU_IDLE" that I have set at 30. If you look at the Afterburner display above, you see that it forces the core to rest for a while. The exact value of the parameter is something you may have to play with. As I have it set now, for a lot of the Core 15 WUs, the trough at the bottom of the GPU usage curve in Afterburner is wide enough to let the GPU cool and keep the temp in the low 70s. You can see the GPU temp fluttering, allowing enough of a cool cycle to keep the temp where I want it. That's exactly what my Afterburner plot looks like now, when I'm running a Proj 8034 WU under v7.1.52 on the laptop.

Since my problem was thermal control, I find MSI afterburner very useful to show you what's going on and what's the resulting temp. Afterburner plot of the GPU usage show you how it is periodically idling the GPU. I had to play with the value of FAH_GPU_IDLE to find one that allowed enough cooling on the down cycles to allow for the max temp to be appropriately limited. It does work, albeit reducing your ppd count (as Heinlein would say, TANSTAAFL).