Jim, my old X4 630 (no overclock) runs 2 stock spec'd 660's using, -sbs 1024 -period_iterations_num 10 -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 and is lag free in use.

I can't comment on the other 2 cards as my other 2 rigs use 3GB 1060's that replaced 5xx cards.

Cheers.

Jim, my old X4 630 (no overclock) runs 2 stock spec'd 660's using, -sbs 1024 -period_iterations_num 10 -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 and is lag free in use.

I can't comment on the other 2 cards as my other 2 rigs use 3GB 1060's that replaced 5xx cards.

Cheers.

I was just Now getting ready to put the Lunatics app in all 3 Puters,
Then I checked in here again...Now I'm totally confuzzed again.
OH Dear now what to do, Oh My, Oh Me, Oh My. (Mind Frying...Brain Fart, FUZzzzzzzzzzzzaaaaaaappp)! ~~*:o/

I was just Now getting ready to put the Lunatics app in all 3 Puters,
Then I checked in here again...Now I'm totally confuzzed again.
OH Dear now what to do, Oh My, Oh Me, Oh My. (Mind Frying...Brain Fart, FUZzzzzzzzzzzzaaaaaaappp)! ~~*:o/

Use the same Application for all three, and the same configuration file (just with different values for period_iterations if it proves necessary), and change your app_config.xmls so that only 1 WU runs at a time on each GPU.

---

Grant
Darwin NT

I was just Now getting ready to put the Lunatics app in all 3 Puters,
Then I checked in here again...Now I'm totally confuzzed again.
OH Dear now what to do, Oh My, Oh Me, Oh My. (Mind Frying...Brain Fart, FUZzzzzzzzzzzzaaaaaaappp)! ~~*:o/

Use the same Application for all three, and the same configuration file (just with different values for period_iterations if it proves necessary), and change your app_config.xmls so that only 1 WU runs at a time on each GPU.

And make sure that each SoG MB, or OpenCL AP, task gets the support of a full CPU core as I was just looking at your X4 630 CPU task times and they're taking a lot longer to complete than mine is which indicates to me that you're CPU is over committed. ;-)

As Grant said, use the same GPU SoG app on all 3 rigs and only run 1 task on each of them (that's what mine are doing).

Forget using any of those old out of date Cuda apps.

Cheers.

This is the app that I have been running for over a year...It lives in BOINC data folder in the S@H folder with the WU's in it.

<app_config>
<app_version>
<app_name>setiathome_v8</app_name>
<plan_class>cuda50</plan_class>
<avg_ncpus>0.20</avg_ncpus>
<ngpus>0.40</ngpus>
</app_version>
<app_version>
<app_name>setiathome_v8</app_name>
<plan_class>cuda42</plan_class>
<avg_ncpus>0.20</avg_ncpus>
<ngpus>0.40</ngpus>
</app_version>
<project_max_concurrent>5</project_max_concurrent>
</app_config>

What the heck do I put where when I install the Optimized App.
And shouldn't I eliminate all of the cuda stuff?
I have Never modified anything to to do with guppies and they only run 1 at a time on the GPU's.

This is the app that I have been running for over a year...It lives in BOINC data folder in the S@H folder with the WU's in it.

<app_config>
<app_version>
<app_name>setiathome_v8</app_name>
<plan_class>cuda50</plan_class>
<avg_ncpus>0.20</avg_ncpus>
<ngpus>0.40</ngpus>
</app_version>
<app_version>
<app_name>setiathome_v8</app_name>
<plan_class>cuda42</plan_class>
<avg_ncpus>0.20</avg_ncpus>
<ngpus>0.40</ngpus>
</app_version>
<project_max_concurrent>5</project_max_concurrent>
</app_config>

What the heck do I put where when I install the Optimized App.
And shouldn't I eliminate all of the cuda stuff?

This is the app_config.xml I use

<app_config>
 <app>
  <name>setiathome_v8</name>
  <gpu_versions>
  <gpu_usage>1.00</gpu_usage>
  <cpu_usage>1.00</cpu_usage>
  </gpu_versions>
 </app>
</app_config>

It will run 1 WU at a time, and reserve 1 CPU thread for that GPU WU.

The Lunatics installer will overwrite your existing app_info.xml with the appropriate information.

---

Grant
Darwin NT

This is the app that I have been running for over a year...It lives in BOINC data folder in the S@H folder with the WU's in it.

<app_config>
<app_version>
<app_name>setiathome_v8</app_name>
<plan_class>cuda50</plan_class>
<avg_ncpus>0.20</avg_ncpus>
<ngpus>0.40</ngpus>
</app_version>
<app_version>
<app_name>setiathome_v8</app_name>
<plan_class>cuda42</plan_class>
<avg_ncpus>0.20</avg_ncpus>
<ngpus>0.40</ngpus>
</app_version>
<project_max_concurrent>5</project_max_concurrent>
</app_config>

What the heck do I put where when I install the Optimized App.
And shouldn't I eliminate all of the cuda stuff?

This is the app_config.xml I use

<app_config>
 <app>
  <name>setiathome_v8</name>
  <gpu_versions>
  <gpu_usage>1.00</gpu_usage>
  <cpu_usage>1.00</cpu_usage>
  </gpu_versions>
 </app>
</app_config>

It will run 1 WU at a time, and reserve 1 CPU thread for that GPU WU.

The Lunatics installer will overwrite your existing app_info.xml with the appropriate information.

I Know that I'm Getting a bit slow anymore. However With my Wife's problems and All of our Vehicles Down for 2 weeks I just Ain't ready to quit the Project Yet. :o)
The MAIN thing is Thank You Sir And Others For YOUR Expertise, Time and Patients.

---

I Desire Peace and Justice, Jim Scott (Mod-Ret.)

with lunatics i use this on a 6core AMD @3.5Ghz and GeForce960 with 2GB Vram ( on other pc with 700er Geforces only 2WU/Card) :

cc_config:
<cc_config>
<log_flags>
<unparsed_xml>1</unparsed_xml>
</log_flags>
<options>
<process_priority_special>4</process_priority_special>
<use_all_gpus>1</use_all_gpus>
<ncpus>12</ncpus>
</options>
</cc_config>

app_config.xml:
<app_config>
<app>
<name>setiathome_v8</name>
<gpu_versions>
<gpu_usage>0.25</gpu_usage>
<cpu_usage>0.25</cpu_usage>
</gpu_versions>
</app>
</app_config>

app_info.xml done by lunatics no own editings there.
boincmanager options settings: use 95%cpu and use 90%processortime so i can surf and response of system is very quick still.

---

with lunatics i use this on a 6core AMD @3.5Ghz and GeForce960 with 2GB Vram ( on other pc with 700er Geforces only 2WU/Card) :

cc_config:
<cc_config>
<log_flags>
<unparsed_xml>1</unparsed_xml>
</log_flags>
<options>
<process_priority_special>4</process_priority_special>
<use_all_gpus>1</use_all_gpus>
<ncpus>12</ncpus>
</options>
</cc_config>

app_config.xml:
<app_config>
<app>
<name>setiathome_v8</name>
<gpu_versions>
<gpu_usage>0.25</gpu_usage>
<cpu_usage>0.25</cpu_usage>
</gpu_versions>
</app>
</app_config>

app_info.xml done by lunatics no own editings there.
boincmanager options settings: use 95%cpu and use 90%processortime so i can surf and response of system is very quick still.

Thanks wykinger

---

I Desire Peace and Justice, Jim Scott (Mod-Ret.)

with lunatics i use this on a 6core AMD @3.5Ghz and GeForce960 with 2GB Vram ( on other pc with 700er Geforces only 2WU/Card) :

cc_config:
<cc_config>
<log_flags>
<unparsed_xml>1</unparsed_xml>
</log_flags>
<options>
<process_priority_special>4</process_priority_special>
<use_all_gpus>1</use_all_gpus>
<ncpus>12</ncpus>
</options>
</cc_config>

app_config.xml:
<app_config>
<app>
<name>setiathome_v8</name>
<gpu_versions>
<gpu_usage>0.25</gpu_usage>
<cpu_usage>0.25</cpu_usage>
</gpu_versions>
</app>
</app_config>

app_info.xml done by lunatics no own editings there.
boincmanager options settings: use 95%cpu and use 90%processortime so i can surf and response of system is very quick still.

But you are using the old and out dated Cuda50 app and Jim should not be using that app_info contents.

In fact if you use the SoG app you will do a lot better than what you're doing now.

Cheers.

Regardless of which version of applicaiton you are using running 4 concurrent tasks on a GTX750 or GTX750ti will be far too many for the optimum output. With the current tasks the optimum is either one or two tasks per GPU, and each with more support from the CPU. So replace:

<gpu_usage>0.25</gpu_usage>
<cpu_usage>0.25</cpu_usage>

with (to run run one task per GPU)

<gpu_usage>1</gpu_usage>
<cpu_usage>0.5</cpu_usage>

or (to run two tasks per GPU)

<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.5</cpu_usage>

You are not looking for maximum GPU memory usage, but maximum GPU compute core usage, with minimum swapping between tasks and minimum CPU/GPU transfer (neither of which are monitored by the majority of GPU monitoring tools).

Even on the GTX1080 4 CUDA tasks per GPU is a long way down the performance (tasks per hour) curve), with the optimal being two or three, depending on the exact tasks type.

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

Is THIS Close to right,

<app_config>
<app_version>
<app_name>setiathome_v8</app_name>
<gpu_versions>
<gpu_usage>1.00</gpu_usage>
<cpu_usage>1.00</cpu_usage>
</gpu_versions>
</app_version>
</app_config>

Still running Stock for now.

---

I Desire Peace and Justice, Jim Scott (Mod-Ret.)

Is THIS Close to right,

Yep.
That will run 1 WU per GPU, and reserve 1 CPU thread per GPU WU to feed them.

Then putting the settings that we've posted previously in the thread in the configuration file for the SoG application will allow the cards to pump out considerably more work.

If not being used for daily work, just as a cruncher,
-tt 600 -hp -period_iterations_num 1 -high_perf -sbs 1024 -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 -cpu_lock
should result in plenty of work being crunched.

Running stock, you'd need to put it in the
mb_cmdline-8.22_windows_intel__opencl_nvidia_SoG.txt
file in the Seti project directory.

If running Lunatics then it goes in the
mb_cmdline_win_x86_SSE3_OpenCL_NV_SoG.txt
file, also in the Seti project directory.

Once the file has been saved, no need to do anything else- the values will be picked up when the next WU starts being processed.

---

Grant
Darwin NT

But you are using the old and out dated Cuda50 app and Jim should not be using that app_info contents.

In fact if you use the SoG app you will do a lot better than what you're doing now.

Cheers.

ok how do i get the new app and how do i configure it optimal ?


"on 750 4 WU´s are too much" - yes i use only on my pc with the gtx960 to run 4 WU´s . On the
one pc with an gts 450, gtx750 and gtx 680ti installed i run 2 WU´s per card cause there was 4WU´s per card slower.

---

ok how do i get the new app and how do i configure it optimal ?

Lunatics 0.45 Beta 6

1 task per GPU with the support of 1 CPU core.

Cheers.

ok how do i get the new app and how do i configure it optimal ?

Really messed up with that, rereading the whole tread here. Maybe my english isnt good enough.
SoG-App - i didnt found it. i have lunatics 0.45 beta6 from mikes world.
Or is SoG a special way to configure this lunatics ?
When installing lunatics and do propper answer the questions i allways end up in Cuda 50 :(

---

On page 3 of the installer Multibeam V8 NV cuda there is an option to choose the SoG app.
Its on the bottom.

---

With each crime and every kindness we birth our future.

On page 3 of the installer Multibeam V8 NV cuda there is an option to choose the SoG app.
Its on the bottom.

THANK YOU :)
i thought this is for old cards not fitting in the discription of the 3 other options ....
ok now i can fix it :)

UPDATE It works fine :)
<ncpus> removed from cc_config.
Question: 1 CPU+0.5 Nvidia GPU or 0.5 CPU+0.5 Nvidia GPU on GTX960 ?
Anything that can be optimized further for crunching ?
2.nd Question: on one PC i have an ATI GPU - no SoG avaible for ATI´ s ?

---

Check for a file called mb_cmdline_win_x86_SSE3_OpenCL_NV**.txt and paste this into it.
Its in the projects folder.

-sbs 384 -spike_fft_thresh 2048 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 32 -oclfft_tune_cw 32

Save as text.
On next task it will use this params.

On ATI SoG. is slower than non_SoG so i suggested not to put it into the installer.

---

With each crime and every kindness we birth our future.

Check for a file called mb_cmdline_win_x86_SSE3_OpenCL_NV**.txt and paste this into it.
Its in the projects folder.

-sbs 384 -spike_fft_thresh 2048 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 32 -oclfft_tune_cw 32

Save as text.
On next task it will use this params.

On ATI SoG. is slower than non_SoG so i suggested not to put it into the installer.

Now made a "how to" for my teammates so some others will optimize , too, and this will benefit seti@home. Thank you so much for help :)

DONE :)
Ah ok if ATI is slower with SoG its logical there is no option for that :)

---