Hi,
Is there a way to use 2 out of 3 GPUs only? Running seti on the GPU connected to the monitor is creating lots of lag (linux).

So I am wondering if there is a way to set a preference for the main GPU to stop during computer use (in preferences) while the other two are crunching.

Or completely disable the main GPU and leave the other 2 running?

---

See for example https://setiathome.berkeley.edu/forum_thread.php?id=82007&postid=1892940 this thread, which uses device exclusion as solution.

---

_\|/_
U r s

awesome, thanks that worked. posting the below for future reference....

        <exclude_gpu>
             <url>http://setiathome.berkeley.edu/</url>
             <device_num>0</device_num>
             <type>ATI</type>
             <app>astropulse_v7</app>
        </exclude_gpu>
        <exclude_gpu>
             <url>http://setiathome.berkeley.edu/</url>
             <device_num>0</device_num>
             <type>ATI</type>
             <app>setiathome_v8</app>
        </exclude_gpu>
    </options>

---

Michel -
Insert the appropriate one in your cc_comfig.xml

<ignore_intel_dev>N</ignore_intel_dev> or <ignore_nvidia_dev>N</ignore_nvidia_dev>. Subsitute "N" with the device #, which you can get from the beginning of your event log.

On the other hand, I see that you have an i7/4770K, go into your bios and select the iGPU for your primary monitor and recycle the machine and insert this instead <ignore_intel_dev>N</ignore_intel_dev>. That way you can use your monitor for just about everything unless you're also a gamer. Hey MOM, I now have 3 x 1070s crunching.

I've used the internal iGPU on both my 4770k & 7770K and have no problems with movies, simple games, etc. with no lagging. You might have to adjust your CPU % a bit and you might lose those tasks that were assigned to the iGPU , but everything should work out.

---

I don't buy computers, I build them!!

It's a good idea to switch back to the iGPU and use the additional card. It is getting cold here... Now I'll have to get over my laziness and get to the back of the tower

---

Not sure how linux/AMD builds works, but with the windows/nvidia builds you could configurate the builds to reduce lag until it actualy allmost not noticeable.

Mainly ussing the <cmdline> tune option. I sugest you try.

---

i have been away for a while from this project that I cannot seem to remember where to put the cmdline... and i cannot find any hints so far

edit: figured that out, in the cmdline text file under /var/lib/boinc-client (for debian anyway)

---

i have been away for a while from this project that I cannot seem to remember where to put the cmdline... and i cannot find any hints so far

AFAIK there are few ways...
For me the easy way is to add it to the app_config file.
Just for example, this is my app_config file. Be aware mine is a 1070 GPU / 6 cores CPU mainly cruncher only host, so you need to adjust to your GPU and your host usage. To learn about the parameters used see the doc files included on the builds and they are a lot of talk about in the forums

<app_config>
<project_max_concurrent>12</project_max_concurrent>
<app_version>
<app_name>setiathome_v8</app_name>
<plan_class>opencl_nvidia_SoG</plan_class>
<avg_ncpus>1.0</avg_ncpus>
<ngpus>0.5</ngpus>
<cmdline>-use_sleep -tt 500 -hp -period_iterations_num 1 -high_perf -sbs 2048 -spike_fft_thresh 4096 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 64 -oclfft_tune_cw 64</cmdline>
</app_version>
<app_version>
<app_name>astropulse_v7</app_name>
<plan_class>opencl_nvidia_100</plan_class>
<avg_ncpus>1.0</avg_ncpus>
<ngpus>0.5</ngpus>
<cmdline>-use_sleep -unroll 28 -oclFFT_plan 256 16 256 -ffa_block 12288 -ffa_block_fetch 6144 -tune 1 64 4 1 -tune 2 64 4 1 -hp </cmdline>
</app_version>
</app_config>

---

For me the easy way is to add it to the app_config file.

Easier still is just to put it in the command line file itself, which in Windows is in the project data directory, and for the stock Windows application is called
mb_cmdline-8.22_windows_intel__opencl_nvidia_SoG.txt
Put it in there and save it, and the next WU that starts picks up those values.

With it in the app_config.xml file you have to exit & restart BOINC, or use the Manager and Options, read config files to get the new values to take effect.
The advantage of app_config.xml is that you have those settings in 1 file, instead of possibly several different files.

---

Grant
Darwin NT

i have added the below:

 -tt 1500 -sbs 1024 -period_iterations_num 4 -spike_fft_thresh 4096 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 64 -oclfft_tune_cw 64

to the file mb_cmdline-8.01-cuda60.txt...

still suffering from severe lag... any suggestions?

---

This is OpenCL command line whilst you are running Cuda app.
Won`t work.

---

With each crime and every kindness we birth our future.

this is what the server is sending me, i haven't modified anything... am i missing something? shouldn't I be getting SoG?

edit: needed to install nvidia-opencl-icd package, now i am getting opencl.

---

this is what the server is sending me, i haven't modified anything... am i missing something? shouldn't I be getting SoG?

You don't get SoG or CUDA Work Units. A WU is a WU is a WU. SoG & CUDA are the applications that process the WUs.

You are running the stock installation for Linux. So the Manager will try different applications out to determine which one performs the best. Most of the time it gets it right (although sometimes it does get it wrong).
From the looks of things it's settled on CUDA60, although there are a couple of WUs there that will be processed with SoG. If you wish to use the values you posted when they are processed you need to put them in the mb_cmdline-8.01-SoG.txt (or whatever it is actually called) file.
The file you have put it in is for the CUDA application & the arguments & values are completely different for that application.

---

Grant
Darwin NT

yes i am not new here but i have been away for a year... i was actually missing some packages, that's why it chose CUDA60. Now i am running SoG, with the above cmdline, and suffering from severe lag as well...

---

yes i am not new here but i have been away for a year... i was actually missing some packages, that's why it chose CUDA60. Now i am running SoG, with the above cmdline, and suffering from severe lag as well...

 -tt 1500 -sbs 1024 -period_iterations_num 4 -spike_fft_thresh 4096 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 64 -oclfft_tune_cw 64

No big surprise -period_iterations_num 4 is hardcore value and very likely to cause screen lags.

Just increase this value in steps of 10 until screen lag is tolerable.
Default is 50

---

With each crime and every kindness we birth our future.