Punchy

Grandpa_01

TheWolf

I'm no Linux uGuru so I wasn't able to find any software that
would show the % being used in the short time I ran the test.
In W7 2x480's I see 95/96% on one card 98/99% on the other, why one it using less I have no clue.
One of the 480's the TPF is always just a few seconds higher than the other no matter the WU.

You can use top to see the % in Linux as you can see Fahcore_a4 is using 10 cores and fahcore_17 is using 1 core on my 980X

Top shows CPU utilization; I believe TheWolf was looking for a tool to show GPU utilization.  I'm afraid I haven't tried GPUs under Linux myself so I'm no help - maybe linuxrouter knows.

Correct Punchy, still looking for something for the GPU in Linux and a way to adjust fan speed.
Edit: "nvidia-smi" will give some GPU info in Linux: here is a Link that may help others.
Monitoring Nvidia GPU Usage in Linux
 

This works, but you will have to reboot 1st before it shows up in the Nvidia settings.
Haven't tested Option "Coolbits" "1" yet.
Edit: Tested "1" doesn't do anything.

jinihammerer

enabling coolbits like bill posted HERE SHOULD work. Only change the option to
Option "Coolbits" "4" 
will get you fan control from the Nvidia control panel.

here is a screeny of what coolbits enabled "used" to look like.

As i said i have not had opportunity to test it but my understanding is it still works.

Coolbits will work for the primary display adapter card (but not for other cards in the computer).  You can use coolbits to adjust the fan, but not the clocks.
 
nvidia-smi only reads the info the driver provides, which for fermi and kepler (consumer cards), does not include clocks.
 
If you want to OC in linux, you need to flash the bios of your card (which is really simple these days, point and click interfaces are available).

jinihammerer So, I finally got to getting set up on my 3770 K machines with the 670 ftw and the gtx 580.
Rather unexpected results.
The gtx 580 is significantly faster then expected and the 670 ftw is significantly slower then expected.

The GTX 580 is currently folding at 1.50 tpf  71314 PPd and the 670 FTW us at 2.36 tpf 42110 PPD.

If I had to guess its reporting the cards wrong (backwards)  the GTX 580 is really the 670 and the 670 is actually the 580.
...

I observed the same mix-up with my GTX 680 and GTX 570.
 
Heiko

Not to mention no easy way to lock GPU to a certain CPU core if you want to CPU fold as well.
Another problem I'm seeing on a BA rig is with dropping cores, can't use 23,22 or 21 cores of a 24 core CPU.
-smp 22 will auto drop to 20 cores, so you loose 4 cores for your BA folding.
CPU core folding do seem to lock to the 1st 20 cores, but even tho it looks like the GPU is using 1 of the
last 4 cores I don't think is the real case since TPF increases from 2:04 to 3:24/3:44 for GPU folding while SMP folding.

Still wouldn't lock the GPU to a certain CPU core, plus I can't use the terminal that way on a EVGA SR2.
See problem here.

I'm confused by what your saying I think, are you saying to not use the OS GUI or not to use the V7 GUI interface?
 
I have disabled everything in bios related to that, that I could find its still there,
as for the floppy its one of the 1st things I always disable in any mobo bios.
I think it may be cause by the SR2 having so many PCIe slots, but I have manually
disable them all via jumper and still hasn't helped.

I read at the FF that you can stop the Mint GUI completely and still fold GPU.
But it 1st has to be loaded, another words you can't boot pure terminal mode and GPU fold.
With the complete load of Mint the Nvidia driver gets loaded and remains loaded even if you kill the GUI were as it doesn't in pure terminal.
 
Yea guess I will have to break down and find the time to make a account to ask the Mint uGuru for help.
 
 
 

Here's the post from Folding forum.
I'm sure if you kill the GUI overhead it will help TPF some even for GPU.
After all you can stop and restart it at any time by this report.
I would have tested but my SR2 problems makes it harder "it can still be done"
I wanted to try to help find a better way, but my SR2 problem henders that.

"Napoleon"I've got a remotely monitored dedicated Mint14 setup on my LAN, and I rarely need the GUI/VNC features so I followed this Ubuntu 12.04 tip: booting to text mode (run level 3) and no X instructions.  The GUI desktop still is only one 'startx' command away, and I noticed that if I do startx and kill X with CTRL-ALT-BACKSPACE, I'm able to start the (tarball) FAHClient from console and GPU compute works - it won't work without the initial startx though.
Apparently the X startup loads some stuff (kernel modules?), and you can kill X and still have functional GPU compute afterwards.  This works for me since the Linux setup is in the adjacent room and I don't reboot it unless updates demand it.  But there's got to be a more elegant way to have the GPU compute enabled from TUI than this.
I'm not particularly Linux savvy, so could someone instruct me how to configure it so that GPU compute works without the extra steps, and I can just run startx from SSH console if I want to VNC to the GUI desktop?
In any case, the A4 work gets a little bit of extra oomph with Xorg and cinnamon processes out of the way.   :e)

If that were the case then "sudo service mdm stop" wouldn't help SMP folding,
so disabling the GUI is a help and should also help GPU folding as well.

But also some tpf is regained by smp once disable even if you allow a core for gpu, so its still good to "sudo service mdm stop" it will still help as both the GPU CPU core & SMP CPU cores since they are being used less now that there is no GUI overhead running in the back ground to use CPU cores. Got to remember this could also be eating away at the CPU core that the GPU is using.

I guess we will have to agree to disagree.
It is only logical for it to improve "if only a small amount" if less resources are being used on all CPU cores.
 
Only rig I have setup at this time is the SR2 with the 660ti.
I'll have to work out my other little problem before I can do any real testing.
With the error speeding by "updates once ever second" it makes it to hard to know if what you input into the terminal is correct or be able to see the results of that command.

Uses 100% of a core on all my W7 rigs so far, don't see why that would change in Linux but its possible.
Actually sometimes it uses part of a 2nd core or so it shows here sometimes less than 100% but most of the time 100% or more.  Lock Zeta to a single core for each 17 core that's running and see how much it uses them.
Also if you lock each core 17 to its own CPU core you should see a small improvement in TPF for each GPU.
Also locking the 10 other cores to the smp client will also help improve its TPF. 
Letting them run loose like that will cause a fight for priority and all will have longer TPF.

I did see some improvement in TPF for the GPU in pure terminal mode vs running in the GUI.
First of this log is in GUI 3:44 there about, later in terminal only I get 3:21 TPF.
This is BA folding at the same time "20 cores" that also saw a little improvement but not as much as the GPU did.
Only about 8 to 10 seconds better TPF for the CPU. Still a far cry from 2:04 TPF with GPU folding only.
Next is to see if I can find a way to tie the GPU to a certain core.
Edit or I should say tie core 17 to one CPU core only. 

BTW my time math is not so good so you might want to take those times with a grain of salt. And check them your self.
Edit2: Setting in the GUI but still folding in pure terminal the TPF has now dropped to
3 mins 16 secs so I really don't know what to make of this or my math was wrong above.

08:51:00:WU00:FS01:0x17:Completed 350000 out of 2500000 steps (14%)           
 08:54:44:WU00:FS01:0x17:Completed 375000 out of 2500000 steps (15%)           
 08:58:22:WU00:FS01:0x17:Completed 400000 out of 2500000 steps (16%)           
 09:02:07:WU00:FS01:0x17:Completed 425000 out of 2500000 steps (17%)           
 09:05:51:WU00:FS01:0x17:Completed 450000 out of 2500000 steps (18%)           
 09:09:30:WU00:FS01:0x17:Completed 475000 out of 2500000 steps (19%)           
 09:13:06:WU00:FS01:0x17:Completed 500000 out of 2500000 steps (20%)           
 09:16:45:WU00:FS01:0x17:Completed 525000 out of 2500000 steps (21%)           
 09:20:21:WU00:FS01:0x17:Completed 550000 out of 2500000 steps (22%)           
 09:24:01:WU00:FS01:0x17:Completed 575000 out of 2500000 steps (23%)           
 09:27:33:WU00:FS01:0x17:Completed 600000 out of 2500000 steps (24%)           
 09:31:02:WU00:FS01:0x17:Completed 625000 out of 2500000 steps (25%)           
 09:34:28:WU00:FS01:0x17:Completed 650000 out of 2500000 steps (26%)           
 09:37:58:WU00:FS01:0x17:Completed 675000 out of 2500000 steps (27%)           
 09:41:29:WU00:FS01:0x17:Completed 700000 out of 2500000 steps (28%)           
 09:44:56:WU00:FS01:0x17:Completed 725000 out of 2500000 steps (29%)           
 09:48:21:WU00:FS01:0x17:Completed 750000 out of 2500000 steps (30%)           
 09:51:46:WU00:FS01:0x17:Completed 775000 out of 2500000 steps (31%)           
 09:55:07:WU00:FS01:0x17:Completed 800000 out of 2500000 steps (32%)           
 09:58:29:WU00:FS01:0x17:Completed 825000 out of 2500000 steps (33%)           
 10:01:48:WU00:FS01:0x17:Completed 850000 out of 2500000 steps (34%)           
 10:05:11:WU00:FS01:0x17:Completed 875000 out of 2500000 steps (35%)           
 10:08:29:WU00:FS01:0x17:Completed 900000 out of 2500000 steps (36%)           
 10:11:52:WU00:FS01:0x17:Completed 925000 out of 2500000 steps (37%)           
 10:15:12:WU00:FS01:0x17:Completed 950000 out of 2500000 steps (38%)           
 10:18:28:WU00:FS01:0x17:Completed 975000 out of 2500000 steps (39%)           
 10:21:50:WU00:FS01:0x17:Completed 1000000 out of 2500000 steps (40%)           
 10:25:10:WU00:FS01:0x17:Completed 1025000 out of 2500000 steps (41%)           
 

10:28:28:WU00:FS01:0x17:Completed 1050000 out of 2500000 steps (42%) 
 10:31:45:WU00:FS01:0x17:Completed 1075000 out of 2500000 steps (43%) 
 10:35:05:WU00:FS01:0x17:Completed 1100000 out of 2500000 steps (44%) 
 10:38:23:WU00:FS01:0x17:Completed 1125000 out of 2500000 steps (45%) 
 10:41:39:WU00:FS01:0x17:Completed 1150000 out of 2500000 steps (46%) 
 10:44:58:WU00:FS01:0x17:Completed 1175000 out of 2500000 steps (47%) 
 10:48:16:WU00:FS01:0x17:Completed 1200000 out of 2500000 steps (48%) 
 10:51:34:WU00:FS01:0x17:Completed 1225000 out of 2500000 steps (49%) 
 10:54:49:WU00:FS01:0x17:Completed 1250000 out of 2500000 steps (50%) 
 10:57:58:WU00:FS01:0x17:Completed 1275000 out of 2500000 steps (51%) 
 11:01:05:WU00:FS01:0x17:Completed 1300000 out of 2500000 steps (52%) 
 11:04:19:WU00:FS01:0x17:Completed 1325000 out of 2500000 steps (53%) 
 11:07:29:WU00:FS01:0x17:Completed 1350000 out of 2500000 steps (54%) 
 

process affinity - linux
http://unix.stackexchange.com/questions/73/how-can-i-set-the-processor-affinity-of-a-process-on-linux