• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

GPU Folding Guide

Overall a slight decline maybe 10-20% production decline from the SMPs. If it was at say 3500 from a quad core I'm getting 3k, plus the GPU of 4-5k.

 
guess my 7900 GTO's wont do the trick here.
 
Sunin said:
A new Nvidia core (v1.07) has been posted at http://www.stanford.edu/~friedrim/FahCore_11.exe. Scott has added support for a new environment variable, NV_FAH_CPU_AFFINITY, that allows the user to directly set the processor affinity mask or set it to zero to disable all affinity settings. The CPU affinity is set with individual bits (it's a mask, not a CPU selector):
setting NV_FAH_CPU_AFFINITY to
1 ==> CPU 0
2 ==> CPU 1
3 ==> CPU 0 and 1
4 ==> CPU 2
8 ==> CPU 3
15 ==> CPUs 0-3
etc

Click to expand...

Ok why in the heck am I not seeing where to put the above? I have 612 beta8 and fahcore 1.07...

Where do i put the NV_FAH_CPU etc thing at?

Ok its not that hard... I figured it out after some research:

Go to start and run type in:

sysdm.cpl

Under the advanced tab click the environmental variables button

Click new

Under Variable name put NV_FAH_CPU_AFFINITY

Under Variable value put the corresponding # you desire... I chose 15.

walla you have now setup the environmental variable hit ok.
 
Have the CUDA drivers been updated for the 8800 series yet? I've read through the guide but havent really been givin a solid response to that.
 
Godmachine said:
Have the CUDA drivers been updated for the 8800 series yet? I've read through the guide but havent really been givin a solid response to that.
Click to expand...

Yes it did so you can grab the 177.35 driver for your 8800 ;)

 
Thanks for the guide. I've got the GPU client running on my 8800 GTS, but now I want to know: can I run a CPU client at the same time? I installed the regular F@H client but it says only one client can run at once. I have a dual core CPU (Core 2 Duo @ 3.1 GHz).

Any advice?

Thanks!
 
Yes download the SMP client, set the GPU to higher priority and wall you should be good to go... if you want to play it safer or if the SMP has a huge impact on the GPU you can always download the console client and load it up!

Either one will play nice if you set it to idle while the GPU is at higher priority.

 
What are the differences between the 1.06 and 1.07 cores? Should I update my client?

 
Apollo:

The 1.07 client is a must for Nvidia multi-GPU. It also introduced the ability to use environment variables to specify the core for the GPU to use. It may have other fixes and tweaks as well. I'm running it on 2 separate single card boxen and it's at least no worse than 1.06 was.

It will be pushed to the "masses" eventually. I'd run it just so you'll have it there as an excuse to go multi-GPU... :D

 
Ok, some advice from you guys is needed. I've got a toaster here, dual core 2ghz T7250 with 2gb ram. It has the QUADRO NVS 140M video card. Yesterday I setup the gpu client and the smp client, with XP being the OS. Now I understand that XP has different results with the gpu client needing one whole core to itself, and I see that happening in the task manager. But my results have been less than expected. Usually on a 2ghz dual core I can finish an smp unit in 24 hours, but with the gpu running I am going to be taking much longer. The gpu only crunched out 9000/25000 since yesterday afternoon while the smp client was running. I've disabled the smp client for now and hopefully the results are better. Anyone have any suggestions on what I can do to improve this setup? I will have over 40 more of these machines to fold with and would like to maximize the points from them!
 
Pogo, create a new thread so we don't clutter up the Guide. ;)
I've got a T8100/3MB folding - SMP + Quadro NVS 320M/256 - GPU2
XP SP3

Turns in about 3k PPD with the GPU way OCed (850, 1840 core/shaders) or about the same with the SMP points higher and GPU2 points lower with GPU at 650/1840

Fahmon is currently reporting GPU2 @ 2043 PPD and SMP @ 1006 PPD (650/1840 GPU OC)

I'm sure we can get things working on your setup.
 
Is it me or is the Stanford folding site down all the time?

Oh and one question, I saw some mention of this in the thread but nothing concrete... How well does the GPU client step back and use less resources when you're actually doing something with your system? Compared to the CPU client perhaps... Though I guess quad/multi-cores probably complicate that question...
 
Guys I need some help setting this up...

Just got a 9800GTX to add to my 8800GTS, on a 680i I put the 9800GTX in slot 1 and moved the 8800GTS to slot 3.

I have tried to follow the instructions but when I run my two shortcuts the instances only see the 8800GTS. This is with the 9800GTX hooked up to the monitor and the 8800GTS on a dummy VGA plug. The desktop is extended to monitor 3 (the first one for the 8800GTS) and does not cancel after leaving the display settings. The first shortcut has -gpu 0 and the machine ID is 2, the second shortcut has -gpu 1 and the machine ID is 3. This is of course using 2 directories.

It's strange because when I remove the dummy VGA plug and run both clients they see the 9800GTX instead, I think on either card both instances are working but when I try to display both one has the progress and the other has test protein. One at a time they each show progress.

Any ideas? I hope to get this sorted soon as I can see the GTX is going to be a beast. :D

 
I've posted this elsewhere but the best way to set up multi GPU clients is to .............
First set up Windoze so that cross-fire/SLI is off.
Then make sure each card has a monitor output, either real or dummy.
Now span Windoze across both monitors. You can span across a dummy output but you'll just see nothing.

The ATI uses the same client as the nVidia so the setup is the same.
Download and install the correct drivers for folding.
Download and install .net framework 2.0 if needed.

Now set a single GPU client up.
Config it and get it to run.
Once its running correctly, stop it and download FahCore_11.exe v1.07
Now replace the FahCore_11.exe v1.06 with the FahCore_11.exe v1.07 in the data folder.
Check it runs with the new core.
Make a new shortcut to start it, check it works.
Now add the -gpu 1 switch to the new shortcut and check that you are now folding on your second card.
If that works then make a copy of your client folder and data folder.
Rename them as needed.
Make a new shortcut that points at the new client and data folders and add the -gpu 0 switch.
Now in one of the data folders, delete the work folder and queue.dat file.
Now use notepad to open the client.cfg file.
You need to change the MachineID from 2 to 3.
If another client is useing MachineID 3 then the first unused you are not useing.
Now starting both clients from the shortcuts sould work.

Luck ............. :D
 
Thanks Tiger, here we go.

Your instructions look good, think I was not doing things methodically enough before. Checked the SMP and it's using machine ID 1 so that's OK. I uninstalled the GPU client and deleted the extra folders and shortcuts I'd made. Already have .net 2.0 and running 177.35 drivers on Vista Home Premium 32 bit. Windows is spanned across the monitors, using dummy VGA on the 8800GTS, monitor connected to the 9800GTX. No SLI set up.

Installed 6.12b8, ran and set it up. Folding OK on the 8800GTS.
Went to download FahCore_11.exe v1.07 from http://www.stanford.edu/~friedrim/FahCore_11.exe

Link seems to not be working? Is there a mirror?

 
So I'm wanting to hop back into things here.. A few questions:

Do we need SLI boards to run multi-GPU?
I have a P965 and a P35 board with 2 PCI-E x16 slots each.. Will these be suitable for multi-GPU, or do I need to sink more money into SLI-capable boards?

It looks like we need to have balanced (same) cards in one machine. Has anyone gotten dissimilar cards (diff cores or diff memory amounts) working with mutli-GPU?

TIA

 
If SLI/crossfire is enabled, this stops Windoze detecting the multi GPU cores.
This stops multi GPU folding.
So you only need the PCI-E slots for folding.

Luck ............... :D
 
The core I'm downloading from Stanford with the client is 1.06 and the link posted in this thread to 1.07 isn't working anymore.

Does 1.06 work for multi-GPU now?

EDIT: 1.06 appears to work with multi-GPU. woot!

 
I asked over at the folding forums.
The required additions for multi gpu folding have been incorperated into the standard core. So there is no longer a need to download a different core.

Luck ............ :D
 
Dan-515 said:
Thanks Tiger, here we go.

Your instructions look good, think I was not doing things methodically enough before. Checked the SMP and it's using machine ID 1 so that's OK. I uninstalled the GPU client and deleted the extra folders and shortcuts I'd made. Already have .net 2.0 and running 177.35 drivers on Vista Home Premium 32 bit. Windows is spanned across the monitors, using dummy VGA on the 8800GTS, monitor connected to the 9800GTX. No SLI set up.

Installed 6.12b8, ran and set it up. Folding OK on the 8800GTS.
Went to download FahCore_11.exe v1.07 from http://www.stanford.edu/~friedrim/FahCore_11.exe

Link seems to not be working? Is there a mirror?

Click to expand...
Just got back today, been planning to continue without the 1.07 core but now we don't need it so it should be ok.

Installed 6.12b8, ran and set it up. Folding ok on the 8800GTS.
Made a new shortcut to start it, works ok on 8800GTS.
Added the -gpu 1 switch to the new shortcut (in Target), ran it but it's still folding on the 8800GTS.

Tried -gpu 0 and -gpu 2 but no effect, also tried putting it under the client parameters in the config but still no effect.

Is there something I'm mising here?

:confused:

EDIT: Tried putting the VGA dummy on the other port of the 8800GTS, tried connecting the monitor to the 8800GTS and putting the VGA dummy on the 9800GTX, tried using a DVI-VGA adaptor to hook up the monitor, all to no effect. The only time I can get the 9800GTX to work is if do not expand the desktop to the 8800GTS.

EDIT 2: Now using the 9800GTX connected to the DVI port and the 8800GTS connected to the VGA port on my monitor. It's still folding on the 9800GTX only until I extend the desktop onto the 8800GTS and then it will only fold on that card. Using Rivatuner with the desktop extended across to the 8800GTS I see the clocks reported correctly so I don't think this is a driver issue.

 
I'm running Vista x64.... nVidia seems to be stuck at 175.19 for their drivers. No CUDA support listed.

OK OK, I found beta drivers 177.66 and they seem to work alright.

But is it normal that I hear a high pitch sound coming from the PC with the Folding client running? I have good ears....
 
Truckweb said:
I'm running Vista x64.... nVidia seems to be stuck at 175.19 for their drivers. No CUDA support listed.

OK OK, I found beta drivers 177.66 and they seem to work alright.

But is it normal that I hear a high pitch sound coming from the PC with the Folding client running? I have good ears....
Click to expand...

Generally a tell tail sign your drawing too much power from your PSU

 
Truckweb said:
I have an Antec 500W PSU and my video card is a BFG 8800GT.... It's not enough?
Click to expand...

You might be one of the lucky ones that their card is slightly "special"

I have a 8800GT that whines (like dog whistle sound) Moved it to 2 boxen same noise once I start folding.

 
Kendrak said:
You might be one of the lucky ones that their card is slightly "special"

I have a 8800GT that whines (like dog whistle sound) Moved it to 2 boxen same noise once I start folding.
Click to expand...

Is it safe to let it run? By "special" do you mean, defective?
 
Truckweb said:
Is it safe to let it run? By "special" do you mean, defective?
Click to expand...

The one I have is still running (OC to hell of course :p) while I hope an RMA auth comes in. I think they are fine, just noisy. If I done get an RMA on that card it's going to the basement in one of the boxen down there.

 
Can someone link the ATI and Nvidia PPD databases here? I tried do a search but our search button on the forums once again is missing!

 
Sunin said:
Can someone link the ATI and Nvidia PPD databases here? I tried do a search but our search button on the forums once again is missing!

Click to expand...

RS direct download link for the Nvidia PPD spreadsheet.

I haven't seen the ATI one in a spreadsheet yet, but here's the link to legoman666's chart in gif format. He seems to keep this updated with the same file name so far.

Hopefully this was what you were looking for Sunin?

 
Kendrak said:
You might be one of the lucky ones that their card is slightly "special"

I have a 8800GT that whines (like dog whistle sound) Moved it to 2 boxen same noise once I start folding.

Click to expand...

My card whines too. I think it has more to do with the card or capacitors on the card than the PSU.

I've played with it, and the higher I overclock it, the higher the pitch!!

Like at stock its barely noticeable.
At 680 core I hear it... I don't think my wife hears the same noise.
At 700 its a higher pitch!

If you hear it, try the different clocks thing to see if you hear it change pitch.

I also can hear it, like when I'm doing anything even 3D stuff on the desktop I can hear the card as if its making different noises when I'm scrolling down a webpage or drawing the cards for solitaire, etc...

Anyway, as a side note I've noticed I get higher PPD's with just the GPU folding, than if I have the GPU + a windows client like 5.03. Like GPU alone, I get about 4800PPD, but then when I've got the 5.03 client running... even though I set the affinity to each having its own core the PPD on the GPU drops about 400 points, but then the 5.03 client is only cranking out about 300++ PPD. So its a net loss.
 
chrisf6969 said:
My card whines too. I think it has more to do with the card or capacitors on the card than the PSU.

I've played with it, and the higher I overclock it, the higher the pitch!!

Like at stock its barely noticeable.
At 680 core I hear it... I don't think my wife hears the same noise.
At 700 its a higher pitch!

If you hear it, try the different clocks thing to see if you hear it change pitch.
Click to expand...
You only need to increase the shader clocks. OC the other clocks hardly makes a difference but it increases heat. Unless you will be gaming a lot, just OC the shaders.

Anyway, as a side note I've noticed I get higher PPD's with just the GPU folding, than if I have the GPU + a windows client like 5.03. Like GPU alone, I get about 4800PPD, but then when I've got the 5.03 client running... even though I set the affinity to each having its own core the PPD on the GPU drops about 400 points, but then the 5.03 client is only cranking out about 300++ PPD. So its a net loss.
Click to expand...
If you isolate the GPU process, you won't see that much of a drop (if at all) and then you can run the other clients for a net gain. Try it with the SMP Seesaw utility and see if it makes a difference.

Speaking of affinity, how does one permanently isolate two GPU clients to a single core? I noticed that every time a new WU is DL by the second client, the affinity resets and it moves to another core reducing both its own production and the CPU clients' production. The first GPU client does not reset affinity but stays locked. So, how do we prevent affinity drift on the second GPU client??

 
If you have core 1.07 then it has added support for a new environment variable, NV_FAH_CPU_AFFINITY, that allows the user to directly set the processor affinity mask or set it to zero to disable all affinity settings.
The CPU affinity is set with individual bits (it's a mask, not a CPU selector):
setting NV_FAH_CPU_AFFINITY to
1 ==> CPU 0
2 ==> CPU 1
3 ==> CPU 0 and 1
4 ==> CPU 2
8 ==> CPU 3
15 ==> CPUs 0-3
etc

If this variable is undefined, the default behavior is exhibited. A primer on Windows environment variables is at http://www.chem.gla.ac.uk/~louis/software/faq/q1.html.

But this core has been withdrawn outside public distribution path.
Please do not ask where you should get it, and wait for an official release or update.

Luck ............. :D
 
Thanks for the information Tiger, but alas I'm still running core v1.06. I tried to locate v1.07 but as you mentioned it's no longer available. I guess I'll have to wait for its re-release or v1.08, whenever that will be. :(
 
Yes, 1.08 should be available shortly. We don't know why it's pulled but nobody told us to stop using them so keep them preciously if you have it.

 
Xilikon said:
Yes, 1.08 should be available shortly. We don't know why it's pulled but nobody told us to stop using them so keep them preciously if you have it.

Click to expand...

and PM me your email if you want core 1.07 I have it on a flash drive :)
 
From the Folding forums ...........

"1.07 should be 2-4% faster than 1.06... There were some perf optimizations thrown in for good measure. They're fine, but since you guys have all done such a great job providing volumes and volumes of data, really really obscure science issues have cropped up and they're getting addressed now. But they're not the sort of thing one notices until there are volumes and volumes of data so we end up with a process like this to shake down the code.

And to clarify: 2% or so on high-end cards, 4% on low-end cards. I've seen the thing pull 810 ns/day (7980 PPD) on supervillin on a GTX280 and I expect that number to improve even more down the road..."

Its Looking better PpD wise.
Also keep up the good work of breaking the client ......... :p

Luck ............ :D
 
You must log in or register to reply here.
Back
Top