Overclock.net › Forums › Overclockers Care › Overclock.net Folding@Home Team › MultiGPU F@H Client
New Posts  All Forums:Forum Nav:

MultiGPU F@H Client

post #1 of 4
Thread Starter 
Hiya guys...I've been putting my x1900s to good work lately using the multiGPU F@H client. Some work units seem to go by faster than others. The last one and the one thats currently being worked are taking hours to complete. No biggey, just not as fast I would have thought.

But I noticed that when the client is running, my CPU usage is at 100%. If the client is GPU based, why would the CPU be running to the max the entire time??

Also, is it possible to run a CPU and a GPU client together without getting any errors? If so, which ones?
ElRigTheRig
(13 items)
 
  
CPUMotherboardGraphicsRAM
Q6600 G0 @ 2.4 GHz Asus Maximus Formula Asus EN8800GT 512MB 4GB G.Skill DDR2-1000 
Hard DriveOptical DriveOSMonitor
1.858TB (3x JBOD) HP 16x DVD Burner Vista Ultimate x64 SP2 24" FPD2485 Gateway LCD 
KeyboardPowerCaseMouse
Razer Lycosa Ultra X3 1000w Antec P182 Logitech G5 
Mouse Pad
Razer Xact Mat 
  hide details  
Reply
ElRigTheRig
(13 items)
 
  
CPUMotherboardGraphicsRAM
Q6600 G0 @ 2.4 GHz Asus Maximus Formula Asus EN8800GT 512MB 4GB G.Skill DDR2-1000 
Hard DriveOptical DriveOSMonitor
1.858TB (3x JBOD) HP 16x DVD Burner Vista Ultimate x64 SP2 24" FPD2485 Gateway LCD 
KeyboardPowerCaseMouse
Razer Lycosa Ultra X3 1000w Antec P182 Logitech G5 
Mouse Pad
Razer Xact Mat 
  hide details  
Reply
post #2 of 4
Yeah, you can run all clients to fold on. Just download the CPU-folding version and set that up too.
I have no idea why it's using 100% with the GPU client. Maybe it's a bug?
    
CPUMotherboardGraphicsRAM
AMD X2 3800+ Manchester Asus A8N-SLi SE nForce4 BFG 8800GTS OC 640MB @ 661/2Ghz Nanya 1GB DDR400 
Hard DriveOptical DriveOSMonitor
160 GB WD Caviar Acer DVD+-RW 16x Windows XP Pro x64 Acer AL1916WS 19" Wide 
KeyboardPowerCaseMouse
Logitech Explorer Hiper Type-R Modular 580W Thermaltake Soprano Black Logitech 500 G-Optical 
Mouse Pad
£3 Generic :D 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
AMD X2 3800+ Manchester Asus A8N-SLi SE nForce4 BFG 8800GTS OC 640MB @ 661/2Ghz Nanya 1GB DDR400 
Hard DriveOptical DriveOSMonitor
160 GB WD Caviar Acer DVD+-RW 16x Windows XP Pro x64 Acer AL1916WS 19" Wide 
KeyboardPowerCaseMouse
Logitech Explorer Hiper Type-R Modular 580W Thermaltake Soprano Black Logitech 500 G-Optical 
Mouse Pad
£3 Generic :D 
  hide details  
Reply
post #3 of 4
Thread Starter 
Quote:
Originally Posted by JohnRogers View Post
Yeah, you can run all clients to fold on. Just download the CPU-folding version and set that up too.
I have no idea why it's using 100% with the GPU client. Maybe it's a bug?
I'm not sure. Whenever I run two clients at the same time though I get conflict errors or something and one will shut down or just not run.
ElRigTheRig
(13 items)
 
  
CPUMotherboardGraphicsRAM
Q6600 G0 @ 2.4 GHz Asus Maximus Formula Asus EN8800GT 512MB 4GB G.Skill DDR2-1000 
Hard DriveOptical DriveOSMonitor
1.858TB (3x JBOD) HP 16x DVD Burner Vista Ultimate x64 SP2 24" FPD2485 Gateway LCD 
KeyboardPowerCaseMouse
Razer Lycosa Ultra X3 1000w Antec P182 Logitech G5 
Mouse Pad
Razer Xact Mat 
  hide details  
Reply
ElRigTheRig
(13 items)
 
  
CPUMotherboardGraphicsRAM
Q6600 G0 @ 2.4 GHz Asus Maximus Formula Asus EN8800GT 512MB 4GB G.Skill DDR2-1000 
Hard DriveOptical DriveOSMonitor
1.858TB (3x JBOD) HP 16x DVD Burner Vista Ultimate x64 SP2 24" FPD2485 Gateway LCD 
KeyboardPowerCaseMouse
Razer Lycosa Ultra X3 1000w Antec P182 Logitech G5 
Mouse Pad
Razer Xact Mat 
  hide details  
Reply
post #4 of 4
Quote:
Originally Posted by JohnRogers View Post
Yeah, you can run all clients to fold on. Just download the CPU-folding version and set that up too.
I have no idea why it's using 100% with the GPU client. Maybe it's a bug?
The GPU client needs 1 GPU core and 1 CPU core to work. 100% usage on your CPU core is normal and expected.

It is not recommended to run a CPU client on the same core that is being used by your GPU. It will slow both down and most likely result in a negative performance net. You can run a single client core on your second CPU core.

A GPU WU should normally take between 8 and 12 hours.


From stanford forums
Quote:
HOWEVER, if you read the GPU FAQ, http://folding.stanford.edu/FAQ-ATI.html it says you should have one CPU/Core dedicated to the GPU client to keep it fed with data. When you run the SMP client with 4 fahcores, and the GPU client with another fahcore, 5 fahcores sharing 2 CPU cores is not going to perform well. Your PPD would probably go up by not running the GPU client.

BOTH the GPU and SMP clients are designed as high performance clients, and by trying to run too many clients, you are slowing down both of them, which slows down how quickly the completed work units are returned to Stanford. And a slower return is what these high performance clients were designed to overcome.
EDIT:

E|Mike,
I should clarify a few things about your setup. Currently folding is not designed to take full advantage of the processing power you possess. You have a few options on your system. You could run 2 GPU clients on your system that would use both GPUs and both CPUs. Although a GPU does do the heavy lifting as far as calculations, the CPU must constantly be available to feed the GPU. With your GPUs you would get the same performance with a much lower end dual core processor because all it realy has to do is feed the GPUs. Running multiple GPUs on one system is one of the most complex folding configurations to get set up and working smoothly, but it is supported.

If you instead did not run GPU clients, but ran a SMP client, you would then utilize the full potential of your CPU. Unfortunately you would be leaving GPU processing power on the table. In the end the SMP client would probably produce more points per day than two GPU clients. At this point in time, there is no perfect combination for your setup.

You can experiment between 2 GPU clients or 1 SMP client and decide which works best for you and how easy they are to maintain. Stanford is working on new GPU cores that are less dependent on a free CPU, but at this time, it just is not as efficient as it can be.

Hope this explains it a bit better.
Folding Rigs
(13 items)
 
  
CPUMotherboardGraphicsRAM
AMD or Intel Anything with a lot of GPU Slots PPD Cruncher Just enough to fold. 
Hard DriveOptical DriveMonitorPower
Smallest Possible None None Whatever it takes 
Case
Bare Naked 
  hide details  
Reply
Folding Rigs
(13 items)
 
  
CPUMotherboardGraphicsRAM
AMD or Intel Anything with a lot of GPU Slots PPD Cruncher Just enough to fold. 
Hard DriveOptical DriveMonitorPower
Smallest Possible None None Whatever it takes 
Case
Bare Naked 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Overclock.net Folding@Home Team
Overclock.net › Forums › Overclockers Care › Overclock.net Folding@Home Team › MultiGPU F@H Client