Because I forget things; how do I get my second GPU in use? New card just delivered :)

20/09/2014 12:03:23 | | CUDA: NVIDIA GPU 0: GeForce GTX 980 (driver version 344.11, CUDA version 6.5, compute capability 5.2, 4096MB, 3910MB available, 7468 GFLOPS peak)
20/09/2014 12:03:23 | | CUDA: NVIDIA GPU 1 (not used): GeForce GTX 670 (driver version 344.11, CUDA version 6.5, compute capability 3.0, 2048MB, 1957MB available, 2845 GFLOPS peak)
20/09/2014 12:03:23 | | OpenCL: NVIDIA GPU 0: GeForce GTX 980 (driver version 344.11, device version OpenCL 1.1 CUDA, 4096MB, 3910MB available, 7468 GFLOPS peak)
20/09/2014 12:03:23 | | OpenCL: NVIDIA GPU 1 (ignored by config): GeForce GTX 670 (driver version 344.11, device version OpenCL 1.1 CUDA, 2048MB, 1957MB available, 2845 GFLOPS peak)

---

http://boinc.berkeley.edu/wiki/Client_configuration

You're looking to use

<use_all_gpus>1</use_all_gpus>
If 1, use all GPUs (otherwise only the most capable ones are used).

in the options section.

You just need to create a cc_config.xml with notepad containing this,

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

and save it into your ProgramData BOINC folder.

[edit] looks like Richard got in while I was typing again. :-D

Cheers.

Thanks Richard, Wiggo. I tried the entry in various places but can't make it work
Here's what's in there at the moment, where does the entry go?

<app_config>
<app>
<name>setiathome_v7</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.2</cpu_usage>
</gpu_versions>
</app>
<app>
<name>astropulse_v6</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.2</cpu_usage>
</gpu_versions>
</app>
</app_config>

---

Different files, you posted the app_config.xml, that goes in the project folder,

The file you need to create is called cc_config.xml, that goes in the Boinc Data directory, you then need to restart Boinc, GPU detection is only done on startup, doing a read config files won't enable the 2nd GPU.

Claggy

Really appreciate everyone's help. Much better now, thanks.

20/09/2014 12:48:29 | | Starting BOINC client version 7.2.42 for windows_x86_64
20/09/2014 12:48:29 | | log flags: file_xfer, sched_ops, task, dcf_debug
20/09/2014 12:48:29 | | Libraries: libcurl/7.25.0 OpenSSL/1.0.1 zlib/1.2.6
20/09/2014 12:48:29 | | Data directory: C:\ProgramData\BOINC
20/09/2014 12:48:29 | | Running under account Martin
20/09/2014 12:48:29 | | CUDA: NVIDIA GPU 0: GeForce GTX 980 (driver version 344.11, CUDA version 6.5, compute capability 5.2, 4096MB, 3777MB available, 7468 GFLOPS peak)
20/09/2014 12:48:29 | | CUDA: NVIDIA GPU 1: GeForce GTX 670 (driver version 344.11, CUDA version 6.5, compute capability 3.0, 2048MB, 1957MB available, 2845 GFLOPS peak)
20/09/2014 12:48:29 | | OpenCL: NVIDIA GPU 0: GeForce GTX 980 (driver version 344.11, device version OpenCL 1.1 CUDA, 4096MB, 3777MB available, 7468 GFLOPS peak)
20/09/2014 12:48:29 | | OpenCL: NVIDIA GPU 1: GeForce GTX 670 (driver version 344.11, device version OpenCL 1.1 CUDA, 2048MB, 1957MB available, 2845 GFLOPS peak)
20/09/2014 12:48:29 | SETI@home | Found app_info.xml; using anonymous platform
20/09/2014 12:48:29 | | Host name: 600T
20/09/2014 12:48:29 | | Processor: 8 GenuineIntel Intel(R) Core(TM) i7-3770K CPU @ 3.50GHz [Family 6 Model 58 Stepping 9]
20/09/2014 12:48:29 | | 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 sse4_2 popcnt aes syscall nx lm vmx tm2 pbe
20/09/2014 12:48:29 | | OS: Microsoft Windows 7: Home Premium x64 Edition, Service Pack 1, (06.01.7601.00)
20/09/2014 12:48:29 | | Memory: 15.94 GB physical, 31.89 GB virtual
20/09/2014 12:48:29 | | Disk: 238.46 GB total, 142.66 GB free
20/09/2014 12:48:29 | | Local time is UTC +1 hours
20/09/2014 12:48:29 | SETI@home | Found app_config.xml
20/09/2014 12:48:29 | | Config: use all coprocessors
20/09/2014 12:48:29 | SETI@home | URL http://setiathome.berkeley.edu/; Computer ID 7280272; resource share 100
20/09/2014 12:48:29 | SETI@home | General prefs: from SETI@home (last modified 01-Feb-2014 18:42:03)
20/09/2014 12:48:29 | SETI@home | Host location: none
20/09/2014 12:48:29 | SETI@home | General prefs: using your defaults
20/09/2014 12:48:29 | | Reading preferences override file
20/09/2014 12:48:29 | | Preferences:
20/09/2014 12:48:29 | | max memory usage when active: 8163.31MB
20/09/2014 12:48:29 | | max memory usage when idle: 14693.96MB
20/09/2014 12:48:29 | | max disk usage: 100.00GB
20/09/2014 12:48:29 | | max CPUs used: 6
20/09/2014 12:48:29 | | (to change preferences, visit a project web site or select Preferences in the Manager)
20/09/2014 12:48:29 | | Not using a proxy
20/09/2014 12:48:30 | SETI@home | Starting task 03ap08af.16918.16023.438086664203.12.46_0
20/09/2014 12:48:30 | SETI@home | Starting task 02ap08ac.11703.263373.438086664207.12.106_0
20/09/2014 12:48:30 | SETI@home | Sending scheduler request: To fetch work.
20/09/2014 12:48:30 | SETI@home | Requesting new tasks for CPU and NVIDIA
20/09/2014 12:48:33 | SETI@home | Scheduler request completed: got 0 new tasks
20/09/2014 12:48:33 | SETI@home | Not sending work - last request too recent: 216 sec

---

20/09/2014 12:03:23 | | CUDA: NVIDIA GPU 0: GeForce GTX 980 (driver version 344.11, CUDA version 6.5, compute capability 5.2, 4096MB, 3910MB available, 7468 GFLOPS peak)

Will be interesting to see the run times on that card.
Are you running 1, 2 or 3 WUs at a time?

---

Grant
Darwin NT

I've been running 3 at a time since last night because that gets me around 99 % GPU load compared to the 93 % if I just run 2 so I'm guessing that's the best configuration.
As for run times, it seems to complete in around 2/3 of the time of the GTX 670 I'm running but it's so difficult to tell because all WUs have different run times. I remember using a neat app a while back that someone had made that would benchmark using a standard WU and you could select running 1, 2, 3 or 4 at a time and it would calculate what was best. Is that still around and has anyone got a link to that?
One thing I will say about the card though is that it runs hot out of the box. Temps quickly got up to 80 C and the chip frequency throttled, and yet the fan was only running at 50 %. I had to use MSI Afterburner to use a custom fan profile to get the temp down so now the fan's at 88 % and the temp is down to 71 C with a 100 MHz overclock that ran stable overnight. This is a Zotac reference design.
There is a removable panel on the card that apparently helps with cooling if you take it off so I'll try that later if I'm ever 'allowed' to shut down the minecraft server that's currently eating into my RAC.

---

I've been running 3 at a time since last night because that gets me around 99 % GPU load compared to the 93 % if I just run 2 so I'm guessing that's the best configuration.

I found with my GTX750Tis that 2 at a time was the sweet spot. They have no problems handling the longer running WUs, but processing the shorter ones takes a lot longer, so it offset the good run times with the longer running WUs.
Hence sticking with only 2 WUs at a time. Given the extra CUDA cores on the 980 over the 750, 3 at a time may give the best throughput.

And once they come up with an application that takes advantage of Maxwell's architecture, the throughput will be even greater.

---

Grant
Darwin NT

I remember using a neat app a while back that someone had made that would benchmark using a standard WU and you could select running 1, 2, 3 or 4 at a time and it would calculate what was best. Is that still around and has anyone got a link to that?

SetiPerformance 1.8 + Added apps from Lunatics v0.41
http://setiathome.berkeley.edu/forum_thread.php?id=73524

(It's up to you if you want to add and text newer apps)

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

Doesn 't make the GPU's running at 99% load, a lot off noise?
Reminds me doing MilkyWay, GPU's @ 99% making sound like an jet, taking off.

Something completely different, still running BOINC 7.2.42 (stock),
using an I7-2600 and 2 AMD/ATI HD5870 GPU's.
[In Einstein@home I run 2 WU's on each GPU].
And I already experience heat problems on the CPU. which has a stock-cooler.
CPU on 90% --> temp >/= 97C, GPU 's 58 and 64C.[2 E@Home WU's on each GPU.]
[This rig draws 465 Watt].
A better cooler for the CPU is necessary when running optimized app. 's in fact
100% use sometimes shut's the PC off, thermal shutdown @ 110C !
PSU has enough capacity 1KWatt.

---

I remember using a neat app a while back that someone had made that would benchmark using a standard WU and you could select running 1, 2, 3 or 4 at a time and it would calculate what was best. Is that still around and has anyone got a link to that?

SetiPerformance 1.8 + Added apps from Lunatics v0.41
http://setiathome.berkeley.edu/forum_thread.php?id=73524

(It's up to you if you want to add and text newer apps)

Thanks for the link. This really is a neat app. Maybe it would be good to get it linked in a sticky so that others can find it easily?
It looks like running 3 at a time is slightly faster on the 980 but quite a lot slower on the 670. So I'll set it to run 2.

---

Doesn 't make the GPU's running at 99% load, a lot off noise?

The fans are quite quiet really and actually I like a little white noise but I've set the card to run a bit hotter at 75 C so the fan runs a bit slower to keep the noise down and maybe keep some life in the bearings.

---

Doesn 't make the GPU's running at 99% load, a lot off noise?
Reminds me doing MilkyWay, GPU's @ 99% making sound like an jet, taking off.

Depends on the video card.
Maxwell based cards use considerably less power than previous cards, so even when running at 100% load, there is a lot less heat to get rid of.

---

Grant
Darwin NT

oops, replied to a question already answered...

---

.

Hello, from Albany, CA!...

I"d rather use 2 high end video cards, the bigger the better.
I now use AMD/ATI 2 HD 5870 GPU's. They are great using OpenCL speed up 500 times faster are realistick, Collatz C ; MilkyWay; GPUGrid had very efficient use and thus 200 - 800 times faster.
Both video cards and CPU maxed out, the rig uses 575 Watt. It has a 1000Watt PSU.
In that case the whole rig gets too hot and a big fan 40cm diameter 9600 rpm max. is used to get rid off the heat and for now I sometimes use it.

---