I built a PC for gaming, but would like to donate time to F@H when I'm not using it. Basically run F@H all night and day since I work 8-5 and just game in the evenings. Might as well put it to good use since it's running anyway wasting electricity right? I have a Phenom 2 x2 and 2 Radeon HD 5750 cards.
I downloaded the GPU3 systray client but I'm not sure if I have it set up right to make full use of the hardware. It looks like I'm completing a little more than 2 work units a day for 500ish points each, so I'm guessing that only one GPU is being used.
Should I run the GPUs in Crossfire or not? Do I need to have a monitor hooked up to each GPU? What is the best client to use and do I need to run multiple clients (for each GPU)?
If you have any tips/suggestions for configuring things to run flat out folding when I'm at work/sleeping I would appreciate it.
Maximizing hardware usage
Moderators: Site Moderators, FAHC Science Team
-
Zagen30
- Posts: 823
- Joined: Tue Mar 25, 2008 12:45 am
- Hardware configuration: Core i7 3770K @3.5 GHz (not folding), 8 GB DDR3 @2133 MHz, 2xGTX 780 @1215 MHz, Windows 7 Pro 64-bit running 7.3.6 w/ 1xSMP, 2xGPU
4P E5-4650 @3.1 GHz, 64 GB DDR3 @1333MHz, Ubuntu Desktop 13.10 64-bit
Re: Maximizing hardware usage
Welcome to the forum (and the project), jsypolt.
With that setup, I'd suggest trying out the beta release of the v7 client. It's a pretty stable beta, and was designed to be easily accessible to newcomers, in particular by making it easier to fold on multiple pieces of hardware.
Crossfire does not benefit F@h; the Crossfire connection (and SLI, too) has been stated to be too slow for F@h to effectively use it to have multiple GPUs work on one Work Unit. In v7, you'd want to set up 2 GPU slots so that both cards are contributing.
One of the main reasons I suggest running v7 is that AMD cards produce a lot more points with it than v6. Normally client version is independent from performance; the client itself doesn't do the processing, but instead is responsible for managing the scientific cores, which are the software packages that actually run all the calculations. AMD cards are the exception, though- v7 is required to get the newer AMD core (written in OpenCL) that makes much better use of 5000 series and newer AMD cards.
You shouldn't need to have a monitor hooked up to both, but I'm really only familiar with running Nvidia cards, so I may be wrong.
Do you want to fold on your CPU? If so, there are some issues to take into account. I believe the OpenCL AMD core uses a fair amount of CPU time, and running an SMP (multi-core) slot may not be able to make the deadlines on a dual-core processor with one of the cores being hit up by two GPU slots. The best setup may be two GPU slots and one uniprocessor slot, leaving the other core free to handle the GPUs.
With that setup, I'd suggest trying out the beta release of the v7 client. It's a pretty stable beta, and was designed to be easily accessible to newcomers, in particular by making it easier to fold on multiple pieces of hardware.
Crossfire does not benefit F@h; the Crossfire connection (and SLI, too) has been stated to be too slow for F@h to effectively use it to have multiple GPUs work on one Work Unit. In v7, you'd want to set up 2 GPU slots so that both cards are contributing.
One of the main reasons I suggest running v7 is that AMD cards produce a lot more points with it than v6. Normally client version is independent from performance; the client itself doesn't do the processing, but instead is responsible for managing the scientific cores, which are the software packages that actually run all the calculations. AMD cards are the exception, though- v7 is required to get the newer AMD core (written in OpenCL) that makes much better use of 5000 series and newer AMD cards.
You shouldn't need to have a monitor hooked up to both, but I'm really only familiar with running Nvidia cards, so I may be wrong.
Do you want to fold on your CPU? If so, there are some issues to take into account. I believe the OpenCL AMD core uses a fair amount of CPU time, and running an SMP (multi-core) slot may not be able to make the deadlines on a dual-core processor with one of the cores being hit up by two GPU slots. The best setup may be two GPU slots and one uniprocessor slot, leaving the other core free to handle the GPUs.
Re: Maximizing hardware usage
Thanks, I'll download v7 tonight. I discovered that I have a 3rd unlock-able core in my CPU so I'll give that a try as well.
-
Zagen30
- Posts: 823
- Joined: Tue Mar 25, 2008 12:45 am
- Hardware configuration: Core i7 3770K @3.5 GHz (not folding), 8 GB DDR3 @2133 MHz, 2xGTX 780 @1215 MHz, Windows 7 Pro 64-bit running 7.3.6 w/ 1xSMP, 2xGPU
4P E5-4650 @3.1 GHz, 64 GB DDR3 @1333MHz, Ubuntu Desktop 13.10 64-bit
Re: Maximizing hardware usage
You should probably back up your hard drive if you're trying to unlock more AMD cores. I have a friend who tried to unlock the 4th core on his X3 and corrupted his Windows installation (or at least he said that's what happened).
Note that getting the 3rd core to work at all may not mean that it's folding stable. I'd suggest running StressCPU2 before you start folding on it; it runs the same code as F@h at about the same efficiency. If it is folding-stable, you could likely SMP fold on 2 cores and make the deadlines, or see if SMP on all 3 plus the GPUs is faster.
Note that getting the 3rd core to work at all may not mean that it's folding stable. I'd suggest running StressCPU2 before you start folding on it; it runs the same code as F@h at about the same efficiency. If it is folding-stable, you could likely SMP fold on 2 cores and make the deadlines, or see if SMP on all 3 plus the GPUs is faster.
Re: Maximizing hardware usage
Use this at your own risk, but from what I know AMD doesn't/didn't manufacture only triple cores, the ones they marketed as triples were quads with one core disabled. Therefore goods news, I think that if you have three you have four, it's just a matter of whether it's unlockable as well. Time for a trip into the BIOS. 
-
codysluder
- Posts: 1024
- Joined: Sun Dec 02, 2007 12:43 pm
Re: Maximizing hardware usage
I don't think its that simple. It's a matter of early yield. If a chip was designed with four cores but one of them is defective, it is perfectly logical to disable it. The chip will meet all of the requirements of an X3. Enabling it may or may not be a good idea. There's simply no way to tell short of extensive testing.ArVee wrote:Therefore goods news, I think that if you have three you have four, it's just a matter of whether it's unlockable as well.
Re: Maximizing hardware usage
Thanks for the tips guys. My BIOS does show 4 cores, but one is unstable as Windows crashes right away when it starts up with them all enabled. So it looks like there are 3 "good" cores. I'll run the stress tool first though to be sure.