Folding Forum

Hello,

So I am very new to folding in a practical sense. (My brother had always mentioned it and he has been "folding" for various groups on the odd laptop and spare desktop for years).
I am very fortunate to have quite a powerful system due to my line of work (Composer for film and tv), and one of the first thing I thought was why not give some run-time to folding!

I do have a few questions, the first being is there a way to see how much CPU, GPU, RAM FAH is using? And if so can I control how much?
Are GPUs better at folding than CPUs or am I barking up the wrong tree here!

Anyway, the main reason I ask is I wouldn't mind allowing FAH full use of the system resources now and again.

Here are by system specs, in-case it is of use:

OS: Windows 7 Pro 64Bit SP1
Motherboard: Gigabyte X79-UD5
CPU: Intel i7-3930K @ 3.8GHz (6 core 12 threads)
RAM: 64GB 1600MHz DDR3 Quad Channel (Patriot Intel Masters - 8x8GB)
GPU: GeForce GTX 670 2GB DDR5
SSD1: 256GB Samsung 830 - SATA3
SSD2: 240GB OCZ Agility 3 - SATA3
SSD3: 128GB Samsung 830 - SATA3

Kind regards,
Kristopher

Version 7 of F@H will try to install a SMP (Symmetric Multi Processor) Slot will 12 threads and a GPU slot. The SMP core will be the lowest priority on your system, and 'hide' nicely when you need to use the PC for something else.

The GPU slot has no inherent priority, so it is more likely to be 'in the way'. On machines where it is noticeable, you can pause just one slot if you want to.

How do I see the settings for the usage?
I am actually looking to see if I can give FAH more power

Each Slot either 'works' or does not work. You could raise the priority of the SMP slot, but it will not do more work, it will just make doing other work more painful.

Changing the priority of the GPU slot is even less usefu,l as it is not using CPU resources, only GPU resources.

F@H will use as much memory as it wants, there is no way to give it 'more'

You could use task scheduler etc, if you wanted to get inventive...

You are helping the project most by receiving work units, processing them as fast as possible, and returning them. You should monitor your log file to see if you get any errors and report them on this forum, with your log showing the error. You probably shouldn't mess with your task priorities for F@H. You should make sure you have a passkey. As things are progressing, a good GPU is going to produce the highest number of points most of the time. Make sure you update your GPU drivers to a good version. (most of the time the newest non-beta version) You could get involved in overclocking your hardware but that's a whole new thing.

KristopherFisher wrote:How do I see the settings for the usage?
I am actually looking to see if I can give FAH more power

If you want to see the actual settings, I've a few screenshots that would be of interest to you. Basically, on this desktop running Windows 7, I have one SMP8 client and two GPU clients (GTX660Ti and GTX560Ti). I won't analyze all the components I've posted here.

As others have posted, there is not much, if anything, you can do within the Folding at Home configuration to give more resources to the F@H software on your system, if you take the default installation parameters from the F@H installer. By default, the installation is greedy for resources. You can look for parasitic things running and turn them off to ensure that maximum resources are available to these processes. I'm not a good example of that - I do have a lot of other things running that do consume cycles, like HFM.net. But, as you can see from the images below, the lion's share of the points come from the two GPUs on the system which take very little of the CPU resources.

System:
--CPU: i7 2600K @4.2 GHz (4 cores/8 Threads) with Noctua NH-U12P CPU Cooler
--MOBO: ASUS P8P67 DeluxeB3
--RAM: 4GB ADATA 1600/8 GB Corsair
--Power Supply: 750W Corsair
--HDDs: Seagate Hyb 750GB--WD Caviar Black 1TB
--GPU Slot 0: EVGA 660GTX-Ti FTW - Signature 2 GPU@ nominal 1046 MHz
--GPU Slot 2: MSI GTX560Ti TwinFrozr@911 MHz
--Win7Home64
FAH V7.2.9; 306.97 drivers

This screenshot shows the readout of the "Processes" tab of the Windows Task Monitor, sorted to show only the F@H Relevant tasks.


This screenshot shows the Resource Monitor part of Windows Task Monitor (accessible from the "Performance Tab) filtered more or less for the same items. This does show an averaged CPU utilization for each of the various tasks.


These performance items are relevant only in the context of what F@H work units are being processed. Here's the project progress summary for this computer from HFM.net, which I keep running. It's relevant to look at the amount of time the SMP task has been running to appreciate from TASKMON how many CPU hours have been consumed. [EDIT] For instance, the SMP Work Unit has been active for 16 hours 41 minutes, hence the over 126 hours shown in the TaskMon screen for this core (all of the 8 CPU Hyperthreads are aggregated to give that total).



Finally, here's a screenshot of my F@H Control GUI window showing also what's running.

Welcome to foldingforum.org, KristopherFisher.

The technical details are shown above are certainly not wrong, but they're probably too detailed to give you a simple overview. Let me boil it down to a couple of sentences.

The SMP slot uses all of the CPU resources that are NOT being used by something else. By design, it runs at the ultimately low priority, it should have almost no impact on anything else your computer might be doing.

GPUs are better at folding than CPUs. Nevertheless, the GPU drivers know nothing about priority so it can have a major impact on anything else that uses the GPU. From what I know about film and tv software, it is highly dependent on the GPU. As far as specific software that would be used by a composer, I have no idea.

My recommendation would be to run SMP folding 24x7 and run GPU folding whenever you're not using the computer. Use that as a starting point and increase or decrease from there.

The %CPU setting isn't useful for either SMP or GPU folding. To reduce CPU utilization, reduce the number of cores allocated to SMP from the default of 12 (aka "all") by even numbers although if you do get as low as 3, that does work.

Unfortunately, GPU folding is pretty much either ON or OFF. When you need to suspend GPU folding (if possible) anticipate that and use the Finish setting rather than the Pause setting so the current result will be returned quickly. If it's not possible, the results will still be accepted when you resume folding and anything you contribute will be appreciated.