Hello S@h gurus,

I was about to start testing my GTX 1060 with 5 tasks at a time but encountered the following situation detailed below before the code section (which is only needed if this is a Boinc or S@h bug).
So for the gurus, please jump to the code section to see if there is a simple config issue.

Scenario:
I cached 1000 tasks with my semi-automated script for deviceQueueOptimisation (similar to Mr Kevvy's)
in both of my HP Z400 rigs thinking they would both last 4 days before WoW.
The rig with two GTX 750 Ti was 200 kms away and went down 5 days before WoW.
I only brought it back up 3 days before WoW.

As for the other rig with the new GTX 1060, I underestimated the amount of work it could do (because of shorties) and it had emptied its GPU queue (of nonVLARs) almost a full day prior to WoW.

So I decided to swap the GTX 1060 with one of the GTX 750 Ti.
And that's when I got the following "CUDA: NVIDIA GPU 1 (not used)":

CUDA: NVIDIA GPU 0: GeForce GTX 1060 6GB (driver version 368.81, CUDA version 8.0, compute capability 6.1, 4096MB, 3566MB available, 4374 GFLOPS peak)	
CUDA: NVIDIA GPU 1 (not used): GeForce GTX 750 Ti (driver version 368.81, CUDA version 8.0, compute capability 5.0, 2048MB, 1689MB available, 1421 GFLOPS peak)	
OpenCL: NVIDIA GPU 0: GeForce GTX 1060 6GB (driver version 368.81, device version OpenCL 1.2 CUDA, 6144MB, 3566MB available, 4374 GFLOPS peak)	
OpenCL: NVIDIA GPU 1 (ignored by config): GeForce GTX 750 Ti (driver version 368.81, device version OpenCL 1.2 CUDA, 2048MB, 1689MB available, 1421 GFLOPS peak)	
OpenCL CPU: Intel(R) Xeon(R) CPU           W3550  @ 3.07GHz (OpenCL driver vendor: Advanced Micro Devices, Inc., driver version 1800.11 (sse2), device version OpenCL 1.2 AMD-APP (1800.11))

I wasn't able to attempt to fix the issue until 6hrs later, and the weirdest part is that the GTX 1060 (processing 4 tasks with Cuda50 as the 750s were) is taking just as long as a 750: ~50mins/4nonVLARs
When the 1060 was alone in the other rig, it was processing 4nonVLARs in ~28mins.

FYI, the only thing I've tried since is to reinstall Lunatics v0.45 beta3 to see if it would fix the "ignored by config" message...but sadly no!

Let me know if you need additional info.
Cheers,
RobG :-)

---

Are you running a cc_config.xml in the BOINC folder telling it to run use all GPUs?

if not, then that is probably the problem.

Will need to look to see where I left my copy of that xml

Look for an edit.

Edit...

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

---

It works! Thanks Zalster :-D

Any idea why it was working with 2 GTX 750 Ti
and not with the 1060 and 1 GTX 750 Ti?

BOINC only recognizes the most powerful GPU in the system with stock settings. Your two GTX 750 Ti's just appeared as one card to BOINC. With different classes of cards in a system, to make BOINC see them all you have to use the cc_config.xml option Zalster showed.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

BOINC only recognizes the most powerful GPU in the system with stock settings. Your two GTX 750 Ti's just appeared as one card to BOINC. With different classes of cards in a system, to make BOINC see them all you have to use the cc_config.xml option Zalster showed.

Thanks Keith!

Is this a config functionality that could be added to Lunatics v0.4x in the future?
If so, what about also specifying the number of tasks/gpu?

Both of these would be great ~selling points~ for using Lunatics.
Or "reusing" it, as in my scenario above!

Cheers,
RobG :-)

BOINC only recognizes the most powerful GPU in the system with stock settings. Your two GTX 750 Ti's just appeared as one card to BOINC. With different classes of cards in a system, to make BOINC see them all you have to use the cc_config.xml option Zalster showed.

Thanks Keith!

Is this a config functionality that could be added to Lunatics v0.4x in the future?
If so, what about also specifying the number of tasks/gpu?

Both of these would be great ~selling points~ for using Lunatics.
Or "reusing" it, as in my scenario above!

Cheers,
RobG :-)

That would have to be a Richard Haselgrove solution as he is the author and maintainer of the Lunatics installers. If he got ambititious, I could see adding another configuration panel which would offer to write a app_config.xml file. Have choices for how many tasks you want to run concurrently for each application for example. Like a menu tick list where you could select all the various app_config keyword options. As it now stands, I view the Lunatics installers as more of a power cruncher tool. Only those users that have much experience and follow all the thread discussion to see what is currently in vogue are the main users. They know enough to read the readmes and experience in trashing tasks through poor editing of app_info and app_config to know the pitfalls.

For example, I made sure to read Richard's post on the new Lunatics Beta 4 and the gotcha that could befall you if you didn't pay attention to the change in the MB parameter file name that Raistmer put in play for the 3500 CL app. I didn't want to ghost 400 tasks like I did last time I messed with changing the MB app. I'm still seeing 25 ghosts hanging around after 2 months. Live and learn and expect failures. I don't know if the Lunatics installer will ever be the main tool deployed to casual crunchers. Not sure I even would want that. I'm pretty sure the installer wouldn't be updated as often if it was deployed as stock to Main.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

I had pointed out that change in nomenclature to Raistmer a while back but nothing ever came from it.

Oh well, why we have to be diligent when install these, lol

I had pointed out that change in nomenclature to Raistmer a while back but nothing ever came from it.

If there is 2 different app flavors they should be named differently.
Windows OpenCL app naming scheme is:
AP7_win_x86_SSE2_OpenCL_ATI_r2742.exe
<subproject>_<OS>_<platform>_<device_flavour>_<revision>.exe

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

I had pointed out that change in nomenclature to Raistmer a while back but nothing ever came from it.

If there is 2 different app flavors they should be named differently.
Windows OpenCL app naming scheme is:
AP7_win_x86_SSE2_OpenCL_ATI_r2742.exe
<subproject>_<OS>_<platform>_<device_flavour>_<revision>.exe

I think the question was more 'where to put the SoG' - we might call that _<variant>_

With r3500, your package (which I've followed in the Beta4 installer) has

       MB8_win_x86_SSE3_OpenCL_NV_SoG_r3500.exe
mb_cmdline_win_x86_SSE3_OpenCL_NV_SoG.txt

OK, formatted like that, it makes more sense: you're treating the variant as an integral part of the _<device_flavour>_, and (as usual, and wisely) not using revision numbers for cmdline placeholders so they are inheritable from revision to revision.

OK, formatted like that, it makes more sense: you're treating the variant as an integral part of the _<device_flavour>_, and (as usual, and wisely) not using revision numbers for cmdline placeholders so they are inheritable from revision to revision.

Exactly, some prev builds namings were made by hands and had SoG after changing rev number. That was temporary and did not follow general naming scheme. Current build follows.

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

Hi, maybe you will know why this started happening last week.

8/20/2016 9:43:15 PM | | CUDA: NVIDIA GPU 0: GeForce GTX 750 Ti (driver version 368.81, CUDA version 8.0, compute capability 5.0, 2048MB, 1689MB available, 1606 GFLOPS peak)
8/20/2016 9:43:15 PM | | CUDA: NVIDIA GPU 1: GeForce GTX 550 Ti (driver version 368.81, CUDA version 8.0, compute capability 2.1, 1024MB, 816MB available, 691 GFLOPS peak)
8/20/2016 9:43:16 PM | | App version needs OpenCL but GPU doesn't support it
8/20/2016 9:43:16 PM | Milkyway@Home | Application uses missing NVIDIA GPU
8/20/2016 9:43:16 PM | | App version needs OpenCL but GPU doesn't support it
8/20/2016 9:43:16 PM | Milkyway@Home | Application uses missing NVIDIA GPU
8/20/2016 9:43:16 PM | | App version needs OpenCL but GPU doesn't support it
8/20/2016 9:43:16 PM | Milkyway@Home | Application uses missing NVIDIA GPU
8/20/2016 9:43:16 PM | | App version needs OpenCL but GPU doesn't support it
8/20/2016 9:43:16 PM | SETI@home | Application uses missing NVIDIA GPU
8/20/2016 9:43:16 PM | | App version needs OpenCL but GPU doesn't support it

I had been running WU's successfully for a week after the Windows 10 Anniversary update. And then after another reboot I got these messages. It finished a bunch of GPU WU's and now the ones that are partial and un-started are not doing anything.

---

Crunch! Crunch! Crunch!

The same happening here after the last windows update.

Reinstalling the nvidia driver (clean) and was fixed

---

First, why are you updating your Windows? Nah, just kidding...

When windows updated, it looks like it replaced the drivers for your GPUs with it's own. When they did that, they installed the ones without any OpenCL.

You should never allow windows to update your drivers. Go to

http://www.nvidia.com/Download/index.aspx?lang=en-us

input your information, download new drivers, do a clean install and then restart your computer and check the BOINC log to make sure the new drivers have installed correctly.

Good luck

Zalster

---

I did not "LET" windows update the drivers, but it seems they had, despite the Device manager stating that it was using the Nvidia Drivers. But now that they have just been updated, I now see this.

8/21/2016 11:40:41 PM | | CUDA: NVIDIA GPU 0: GeForce GTX 750 Ti (driver version 372.54, CUDA version 8.0, compute capability 5.0, 2048MB, 1689MB available, 1606 GFLOPS peak)
8/21/2016 11:40:41 PM | | CUDA: NVIDIA GPU 1: GeForce GTX 550 Ti (driver version 372.54, CUDA version 8.0, compute capability 2.1, 1024MB, 816MB available, 691 GFLOPS peak)
8/21/2016 11:40:41 PM | | OpenCL: NVIDIA GPU 0: GeForce GTX 750 Ti (driver version 372.54, device version OpenCL 1.2 CUDA, 2048MB, 1689MB available, 1606 GFLOPS peak)
8/21/2016 11:40:41 PM | | OpenCL: NVIDIA GPU 1: GeForce GTX 550 Ti (driver version 372.54, device version OpenCL 1.1 CUDA, 1024MB, 816MB available, 691 GFLOPS peak)

OpenCL is back.

---

Crunch! Crunch! Crunch!

Glad to hear that you got it back up and running.

Happy crunching..

Zalster

---