Installed second video card, but only one is recognized

[melephant]

Bad news... my small video card has stopped folding and all the settings are the same. The big card folds fine. I have no clue why. I haven't tried reinstalling ver.7 yet. Any recomendations before I do that. How can I run memtestg80 on the second card without pulling the first card?

[codysluder]

I don't see any discussion of the Power/Heat Dissipation issues. Are you sure your power supply is adequate to power both GPUs and your fans are adequate to keep both of them cool? Does each one work separately or do you have a bad GPU?

[melephant]

Both cards run less than 66 C which is 150 F. The whole computer and monitor uses 505 W and I have a 850 W PSU.

[MtM]

Bad news... my small video card has stopped folding and all the settings are the same. The big card folds fine. I have no clue why. I haven't tried reinstalling ver.7 yet. Any recomendations before I do that. How can I run memtestg80 on the second card without pulling the first card?

Run MemtestCL, choose platform ( if asked ) choose index, same openCL index you set in the client for the card you suspect.

As to the psu issue, run memtestCL on both cards simultaneously ( for a prolonged time ). Run a cpu burn in/stress test at the same time to push your psu to the max. If either of them fail when being stressed at the same time but not when being stressed individually, upgrading your psu might be in order

As to the Cuda 2.0 issue, not sure ( actually I am sure I been told but I forgot I think ) but doesn't the client report the cuda hw capabilities not the cuda driver version? Mine reports cuda 1.1 and my driver isn't that ancient ( 285.62 CUDA 4.1.1 ). But my g92 hw capabilities are 1.1.

[melephant]

My issue is solved and it is so logical. I found that with one gpu running and one not running you can click on add, select gpu then set the appropriate indicies. The trick was to open slots with out selecting anything before clicking add because it opens with nothing selected. If you select anything there it will not work. That is something so simple a child should be able to figure it out. I am home free and if anything should stop folding again I will know exactly how to get everything back in order. I can't wait to see how many points I'll get tommorrow. I'm saving my money for the next generation of Intel cpu's and Nvidia gpu's to present themselves with adequate drivers. I love Version 7 and I feel my frustration may have ceased. It wasn't the code after all. I hope someone can benefit from my errors. I wish someone could tell what happened to all the NVidia folders. There seems to me half as many active now. How can that be?

[PantherX]

...Is there a setting in which I can configure the forum to send an email to me when a reply has posted? ...

When viewing a thread, you scroll to the bottom of the page and there's an option called "Subscribe topic" and you click it. Once done, whenever a new post is made in that thread, you will be sent an email.

[Scottie_314159]

I know this is an old thread, but the issues is still relevant...

I have successfully followed fredfoobar and melephant's instructions on setting gpu ID and cuda-index ID, this allows me to get my Fermi and non-fermi card folding.

However when I restart, I am still getting the cuda-index resetting to -1 as mentioned by melephant.

Are there any new developments in this area? any new fixes? or methods to ensure settings are saved between restarts?

Thanks
Scott

[Scottie_314159]

SOLVED:

taking a look at the config.xml file in C:\ProgramData\FAHClient (for windows 7) you should be able to adjust the various indexes as required. For example the settings were not being saved for the Femi card between restarts, so i added the extra lines to set the cuda-index and opencl-index. (just copy and paste to ensure you get the syntax right and then adjust values as needed.)

... by the way, this is me: http://folding.extremeoverclocking.com/user_summary.php?s=&u=518943

[PantherX]

...taking a look at the config.xml file in C:\ProgramData\FAHClient (for windows 7) you should be able to adjust the various indexes as required. For example the settings were not being saved for the Femi card between restarts, so i added the extra lines to set the cuda-index and opencl-index. (just copy and paste to ensure you get the syntax right and then adjust values as needed.)...

Could you please provide the config that V7 would generate (which would fail) and yours (which works) so we can see the difference. Remember to remove your passkey before posting it here.

[Scottie_314159]

as requested....

Again, the problem was fahcontrol remembering the settings for my device IDs....

This is the config file that wasn't working for me... notice slot ID 0 wouldn't save the cuda-index and opencl-index, i would have to change the indexes manually each time I restarted fahcontrol...

<config>
<!-- Folding Slot Configuration -->
<gpu v='true'/>

<!-- Network -->
<proxy v=':8080'/>

<!-- User Information -->
<passkey v=''/>
<team v='11108'/>
<user v='Scottie_314159'/>

<!-- Folding Slots -->
<slot id='2' type='SMP'/>
<slot id='0' type='GPU'>
<gpu-index v='1'/>
</slot>
<slot id='1' type='GPU'>
<cuda-index v='1'/>
<gpu-index v='0'/>
<opencl-index v='1'/>
</slot>
</config>

This is the config file that is now working for me, i copied the two lines missing in slot 0 from slot 1 and adjusted the indexes.

<config>
<!-- Folding Slot Configuration -->
<gpu v='true'/>

<!-- Network -->
<proxy v=':8080'/>

<!-- User Information -->
<passkey v=''/>
<team v='11108'/>
<user v='Scottie_314159'/>

<!-- Folding Slots -->
<slot id='2' type='SMP'/>
<slot id='0' type='GPU'>
<cuda-index v='0'/>
<gpu-index v='1'/>
<opencl-index v='0'/> </slot>
<slot id='1' type='GPU'>
<cuda-index v='1'/>
<gpu-index v='0'/>
<opencl-index v='1'/>
</slot>
</config>

As a reminder Slot 0: is my Geforce 480 and slot 1 is my geforce 260

thanks

[Athlonite]

Have you tried swapping the cards around so that the 480 is GPU 0 and the 9600 is GPU 1
so it runs 480+9600 instead of other way round

[melephant]

Where can I find the config file?

[Napoleon]

The config.xml file is in the client's Data directory. Unlike v6 client's config, it's OK to edit it manually in a text editor. To be on the safe side, shut down the client before editing and restart after you've saved your changes.

[melephant]

The issue is solved and, I feel, this time for goood. Thanks to all who spent their time helping me.

[bruce]

I finally got two GPUs running in the same Linux system. There have been numerous reports of inconsistencies in how GPUs are enumerated so I'm not surprised but I'm still going to report what I see.

My first clue was that I seemed to be folding on the "wrong" GPU. It turns out that ocore, using OpenCL enumerates devices in the opposite order that both lspci and nvidia-smi (and FAHClient) do -- at least on this system.

If the two GPU were supported by different versions of OpenCL (say one is NVidia and one in AMD/ATI) or even if they were different gnerations of GPUs from the same vendor) this could be a serious problem. In my case, they're both modern NV GPUs so it's not so serious, just confusing, because the fast GPU is apparently slower than the slow GPU.

Code: Select all

22:49:53:******************************* System ********************************
22:49:53:        CPU: Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz
22:49:53:     CPU ID: GenuineIntel Family 6 Model 26 Stepping 4
22:49:53:       CPUs: 8
22:49:53:     Memory: 5.82GiB
22:49:53:Free Memory: 3.93GiB
22:49:53:    Threads: POSIX_THREADS
22:49:53: OS Version: 3.13
22:49:53:Has Battery: false
22:49:53: On Battery: false
22:49:53: UTC Offset: -7
22:49:53:        PID: 10313
22:49:53:        CWD: /var/lib/fahclient
22:49:53:         OS: Linux 3.13.0-32-generic x86_64
22:49:53:    OS Arch: AMD64
22:49:53:       GPUs: 2
22:49:53:      GPU 0: NVIDIA:4 GM107 [GeForce GTX 750 Ti]
22:49:53:      GPU 1: NVIDIA:3 GK106 [GeForce GTX 650 Ti]
22:49:53:       CUDA: 3.0
22:49:53:CUDA Driver: 6000

I'm running two p2901 WUs. According to FAHControl,
WU03 has a tpf of 21 mins 08 secs and WU01 has a tpf of 4 mins 35 secs
Also according to the information shown above, GPU 0 is the GM107 is in slot FS01 and
the GK106 is in slot FS02. That's apparently false.

Code: Select all

02:50:48:WU01:FS02:0x17:Completed 3850000 out of 5000000 steps (77%)
02:55:22:WU01:FS02:0x17:Completed 3900000 out of 5000000 steps (78%)
02:59:57:WU01:FS02:0x17:Completed 3950000 out of 5000000 steps (79%)
03:04:31:WU01:FS02:0x17:Completed 4000000 out of 5000000 steps (80%)
03:07:50:WU03:FS01:0x17:Completed 2250000 out of 5000000 steps (45%)
03:09:06:WU01:FS02:0x17:Completed 4050000 out of 5000000 steps (81%)
03:13:41:WU01:FS02:0x17:Completed 4100000 out of 5000000 steps (82%)
03:18:15:WU01:FS02:0x17:Completed 4150000 out of 5000000 steps (83%)
03:22:50:WU01:FS02:0x17:Completed 4200000 out of 5000000 steps (84%)
03:27:24:WU01:FS02:0x17:Completed 4250000 out of 5000000 steps (85%)
03:28:58:WU03:FS01:0x17:Completed 2300000 out of 5000000 steps (46%)
03:31:59:WU01:FS02:0x17:Completed 4300000 out of 5000000 steps (86%)
03:36:33:WU01:FS02:0x17:Completed 4350000 out of 5000000 steps (87%)

Code: Select all

bruce@u192:~$ nvidia-smi
Mon Jul 28 20:21:57 2014       
+------------------------------------------------------+                       
| NVIDIA-SMI 331.38     Driver Version: 331.38         |                       
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 750 Ti  Off  | 0000:03:00.0     N/A |                  N/A |
| 57%   70C  N/A     N/A /  N/A |    248MiB /  2047MiB |     N/A      Default |
+-------------------------------+----------------------+----------------------+
|   1  GeForce GTX 650 Ti  Off  | 0000:04:00.0     N/A |                  N/A |
| 24%   67C  N/A     N/A /  N/A |     43MiB /  1023MiB |     N/A      Default |
+-------------------------------+----------------------+----------------------+

Code: Select all

bruce@u192:~$ lspci|grep VGA
03:00.0 VGA compatible controller: NVIDIA Corporation GM107 [GeForce GTX 750 Ti] (rev a2)
04:00.0 VGA compatible controller: NVIDIA Corporation GK106 [GeForce GTX 650 Ti] (rev a1)
bruce@u192:~$

Code: Select all

bruce@u192:~$ Downloads/ocore_601_OpenCL_v19 --devices --nospoiler
OpenCL compatible devices:
name: GeForce GTX 650 Ti | platformId: 0 deviceId: 0
name: GeForce GTX 750 Ti | platformId: 0 deviceId: 1
bruce@u192:~$

In this case, it really looks like OpenCL should detect the 650Ti as deviceid 1 and the 750Ti as deviceid 0 but I'm not sure if somebody else's configuration would suggest something else. Is that an inconsistency in NV's OpenCL, or is ocore changing things?