I'm sorry if the title sounds confusing but that is the best way I can describe my problem. I have two GPU's in my system. I just replaced one of them. BOINC is still looking for the old one. The problem is, now only one of them is being used. How do I make BOINC start using the new GPU and stop looking for the old one that has been removed?

Reinstall the drivers.

And otherwise, please read When requesting help on these forums. It's so difficult to guess as to what you're on about.

---

Please forgive me, I feel I was as clear as I could be about my problem. I don't know what additional information I can provide so please ask me and I will provide.

Re-installing the drivers didn't help. It's only using one of the original GPU's. I figured out that I can look at the event log. In the log I can clearly see that it's trying to find the GPU that is no longer in the system. It's as if BOINC is trying to continue to use the hardware it knew about and not look for new hardware. How can I get it to look for the new GPU and stop trying to use the one that is no in the system anymore?

1) BOINC itself will silently forget about any GPU it can no longer find. The new GPU - provided you've installed a valid driver for it - should show up, but if it is significantly different from the remaining old one (faster, slower, bigger. smaller) BOINC won't use one of them by default (that will be marked in the Event Log at startup). You can enable <use_all_gpus> via a cc_config.xml file - see Client configuration

2) As above - BOINC won't be trying to find the missing GPU. But some project might be telling you that "app_version xxx refers to missing GPU yyy", or words to that effect. It would be better to copy the relevant lines from the event log window, and paste them here - that way, we get to see exactly what you are seeing.

If it is an application report from a project, the best way to attempt a first fix would be to set 'No New Tasks' for the project in question, and - when all active tasks for the project in question have been completed and reported - reset the project.

I added a cc_config.xml to include this:

<cc_config>
<options>
<use_all_gpus>1</use_all_gpus>
</options>
</cc_config>

I also see this in the event log:

4/11/2014 2:36:51 PM | SETI@home | Application uses missing ATI GPU
4/11/2014 2:36:51 PM | | app version refers to missing GPU type ATI
4/11/2014 2:36:51 PM | Einstein@Home | Application uses missing ATI GPU
4/11/2014 2:36:51 PM | | Host name: NickJohnson


All of these are messages that relate to a GPU that is no longer installed in the system. Here are messages about a GPU that is in the system:

4/11/2014 2:36:51 PM | | CUDA: NVIDIA GPU 0: Quadro FX 1700 (driver version 307.45, CUDA version 5.0, compute capability 1.1, 512MB, 303MB available, 88 GFLOPS peak)
4/11/2014 2:36:51 PM | | OpenCL: NVIDIA GPU 0: Quadro FX 1700 (driver version 307.45, device version OpenCL 1.0 CUDA, 512MB, 303MB available, 88 GFLOPS peak)
4/11/2014 2:36:51 PM | | OpenCL: NVIDIA GPU 0: Quadro FX 1700 (driver version 307.45, device version OpenCL 1.0 CUDA, 512MB, 303MB available, 88 GFLOPS peak)


I see nothing about the Quadro FX 3500 that is also installed. Why won't it attempt to use that GPU also?

4/11/2014 2:36:51 PM | | CUDA: NVIDIA GPU 0: Quadro FX 1700 (driver version 307.45, CUDA version 5.0, compute capability 1.1, 512MB, 303MB available, 88 GFLOPS peak)
4/11/2014 2:36:51 PM | | OpenCL: NVIDIA GPU 0: Quadro FX 1700 (driver version 307.45, device version OpenCL 1.0 CUDA, 512MB, 303MB available, 88 GFLOPS peak)
4/11/2014 2:36:51 PM | | OpenCL: NVIDIA GPU 0: Quadro FX 1700 (driver version 307.45, device version OpenCL 1.0 CUDA, 512MB, 303MB available, 88 GFLOPS peak)


I see nothing about the Quadro FX 3500 that is also installed. Why won't it attempt to use that GPU also?

Do you have a monitor connected to the Quadro FX 3500, or have the desktop extended onto it?

Claggy

Yes. There are three monitors total. On card has two, the other has one.

Yes. There are three monitors total. On card has two, the other has one.

Can you post the Boinc startup from the Event log please (the first 30 lines will do), it would be helpful knowing which version of Boinc you are running, and which OS (I'm assuming a Windows OS, (Rather than Linux, or OS X), But which one),
another read of When requesting help on these forums. would be helpful too.

Claggy

4/14/2014 8:48:43 AM | | Starting BOINC client version 7.2.42 for windows_x86_64
4/14/2014 8:48:43 AM | | log flags: file_xfer, sched_ops, task
4/14/2014 8:48:43 AM | | Libraries: libcurl/7.25.0 OpenSSL/1.0.1 zlib/1.2.6
4/14/2014 8:48:43 AM | | Data directory: C:\ProgramData\BOINC
4/14/2014 8:48:43 AM | | Running under account njohnson
4/14/2014 8:48:43 AM | | CUDA: NVIDIA GPU 0: Quadro FX 1700 (driver version 307.45, CUDA version 5.0, compute capability 1.1, 512MB, 405MB available, 88 GFLOPS peak)
4/14/2014 8:48:43 AM | | OpenCL: NVIDIA GPU 0: Quadro FX 1700 (driver version 307.45, device version OpenCL 1.0 CUDA, 512MB, 405MB available, 88 GFLOPS peak)
4/14/2014 8:48:43 AM | | OpenCL: NVIDIA GPU 0: Quadro FX 1700 (driver version 307.45, device version OpenCL 1.0 CUDA, 512MB, 405MB available, 88 GFLOPS peak)
4/14/2014 8:48:43 AM | | app version refers to missing GPU type ATI
4/14/2014 8:48:43 AM | Einstein@Home | Application uses missing ATI GPU
4/14/2014 8:48:43 AM | | app version refers to missing GPU type ATI
4/14/2014 8:48:43 AM | SETI@home | Application uses missing ATI GPU
4/14/2014 8:48:43 AM | | Host name: NickJohnson
4/14/2014 8:48:43 AM | | Processor: 8 GenuineIntel Intel(R) Xeon(R) CPU E5405 @ 2.00GHz [Family 6 Model 23 Stepping 6]
4/14/2014 8:48:43 AM | | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss htt tm pni ssse3 cx16 sse4_1 nx lm vmx tm2 dca pbe
4/14/2014 8:48:43 AM | | OS: Microsoft Windows 7: Professional x64 Edition, Service Pack 1, (06.01.7601.00)
4/14/2014 8:48:43 AM | | Memory: 16.00 GB physical, 31.99 GB virtual
4/14/2014 8:48:43 AM | | Disk: 465.66 GB total, 339.12 GB free
4/14/2014 8:48:43 AM | | Local time is UTC -5 hours
4/14/2014 8:48:43 AM | | Config: use all coprocessors
4/14/2014 8:48:43 AM | climateprediction.net | URL http://climateprediction.net/; Computer ID 1316594; resource share 100
4/14/2014 8:48:43 AM | Einstein@Home | URL http://einstein.phys.uwm.edu/; Computer ID 10504669; resource share 100
4/14/2014 8:48:43 AM | Milkyway@Home | URL http://milkyway.cs.rpi.edu/milkyway/; Computer ID 564004; resource share 100
4/14/2014 8:48:43 AM | SETI@home | URL http://setiathome.berkeley.edu/; Computer ID 7220983; resource share 100
4/14/2014 8:48:43 AM | | No general preferences found - using defaults
4/14/2014 8:48:43 AM | | Reading preferences override file
4/14/2014 8:48:43 AM | | Preferences:
4/14/2014 8:48:43 AM | | max memory usage when active: 8190.83MB
4/14/2014 8:48:43 AM | | max memory usage when idle: 14743.49MB
4/14/2014 8:48:43 AM | | max disk usage: 342.44GB
4/14/2014 8:48:43 AM | | don't use GPU while active
4/14/2014 8:48:43 AM | | suspend work if non-BOINC CPU load exceeds 25%
4/14/2014 8:48:43 AM | | (to change preferences, visit a project web site or select Preferences in the Manager)
4/14/2014 8:48:43 AM | | Not using a proxy
4/14/2014 8:49:09 AM | | Suspending GPU computation - computer is in use
4/14/2014 8:49:43 AM | Milkyway@Home | Starting task de_separation_79_DR8_Rev_7_2_001_1396965303_2111727_0
4/14/2014 8:50:43 AM | climateprediction.net | work fetch suspended by user
4/14/2014 8:51:06 AM | climateprediction.net | Task hadam3p_anz_n9pr_2012_1_008600207_0 exited with zero status but no 'finished' file
4/14/2014 8:51:06 AM | climateprediction.net | If this happens repeatedly you may need to reset the project.
4/14/2014 8:51:06 AM | climateprediction.net | Task hadam3p_anz_n2hn_2012_1_008592584_1 exited with zero status but no 'finished' file
4/14/2014 8:51:06 AM | climateprediction.net | If this happens repeatedly you may need to reset the project.

Quadro FX 1700 supports CUDA, FX 3500 doesn't. NVIDIA, Wikipedia.


Isn't it nice when bigger number means better, more capable product?

Wow, that is crazy. So basically the 3500 at this time cannot do anything for BOINC?

Yep. (At least I can't think of anything.)

If it's any consolation, I've got a Quadro FX 1500, and it doesn't support CUDA or OpenCL either.

Mind you, I did buy the computer in November 2006, a couple of years before BOINC supported CUDA - so I didn't know what I was going to be missing. The computer is still going strong, and has had a replacement CUDA card running for five of those years: it's about to get its second transplant.

4/14/2014 8:48:43 AM | | CUDA: NVIDIA GPU 0: Quadro FX 1700 (driver version 307.45, CUDA version 5.0, compute capability 1.1, 512MB, 405MB available, 88 GFLOPS peak)
4/14/2014 8:48:43 AM | | OpenCL: NVIDIA GPU 0: Quadro FX 1700 (driver version 307.45, device version OpenCL 1.0 CUDA, 512MB, 405MB available, 88 GFLOPS peak)
4/14/2014 8:48:43 AM | | OpenCL: NVIDIA GPU 0: Quadro FX 1700 (driver version 307.45, device version OpenCL 1.0 CUDA, 512MB, 405MB available, 88 GFLOPS peak)

Knowing the the FX 3500 isn't Cuda compatible now, the next question is why Boinc displays the Quadro FX 1700 OpenCL device twice, because two monitors are connected to it?, I do remember someone with a HD5700 showing this once here too.

Claggy