The following is not new information but can be found if one searches around long enough.

Procedure to run multiple Milkyway WU’s on a single GPU card:

Step 1 – Determine if you want to run multiple WU’s

A) Go to http://www.techpowerup.com/gpuz/ and download, unzip, and install the GPU-Z utility

B) With a single Milkyway GPU WU running, start the GPU-Z utility. Go to the Sensors tab and check your GPU usage (%) and memory usage.

C) If you have sufficient resources to run another unit, go to Step 2

Step 2 – Setup Milkyway to Run Multiple WU’s Simultaneously

A) Install BOINC 7.0.42 or later (Current version is 7.0.44 Beta)

B) Place the following in a text file using your favorite text editor:

<app_config>
<app>
<name>milkyway</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.05</cpu_usage>
</gpu_versions>
</app>
</app_config>

Note: The gpu usage can be set to .33 to run 3 WU’s simultaneously, etc.

C) Save the text file as app_config.xml in your Milkyway directory (….\BOINC\projects\milkyway.cs.rpi_milkyway)

D) Restart BOINC

In my case, a single Milkyway unit took about 1200 seconds, used 55% of the GPU, about 20% of the GPU memory, and the GPU temp was 62C. Two simultaneous WU’s take about 1300 seconds, use 99% of the GPU, 40% of the GPU memory, with a GPU temp of 70C. .

The following is not new information but can be found if one searches around long enough.

Procedure to run multiple Milkyway WU’s on a single GPU card:

<app_config>
<app>
<name>milkyway</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.05</cpu_usage>
</gpu_versions>
</app>
</app_config>

THANK YOU I have been looking for a working one of those ever since I heard about them!! I question though, if you changed the name from Milkyway to Poem, or whatever, do you think it would work? Yours does not seem project specific except in that spot. I understand that not ALL projects can handle crunching multiple wu' at once, nor is it always helpful.

---

sorry for the double post!!

<app_config>
<app>
<name>milkyway</name>
<max_concurrent>4</max_concurrent>
<gpu_versions>
<gpu_usage>.5</gpu_usage>
<cpu_usage>.05</cpu_usage>
</gpu_versions>
</app>
</app_config>

my configuration is a littlebit different. at using multiple gpus u must enter a maximum number of tasks who will be excuted.

by using 2 gpu i get approximately 100% more wus to be calculated. havend testet 3 wus per gpu, thought i did this before but-> no advantage.

greetings from germany :)

edit: must be worth a news bulleting, are´n it !?

I have a quad processor.

I DO NOT have the <max_concurrent>4</max_concurrent> in my app_config file.

My computer runs 4 non-GPU and 2 GPU WU's (total 6) concurrently.

The information provided should work for any application that allows multiple WU's in the GPU. However, I haven't tried any other application. The only other GPU WU's I have are SETI and one SETI WU consumes 95% of my GPU, so there was no sense trying to run two.

so there was no sense trying to run two.

thats true. but milkyway does benefit of it.

The information provided should work for any application that allows multiple WU's in the GPU. However, I haven't tried any other application. The only other GPU WU's I have are SETI and one SETI WU consumes 95% of my GPU, so there was no sense trying to run two.

THANKS!!

so there was no sense trying to run two.

thats true. but milkyway does benefit of it.

Yeah soon do and some don't, testing can tell the difference as can gpu-z.

Just curious what card you're doing this with. I'm running single MW WUs on a Radeon HD 6950. WUs take about 90 seconds at 98% GPU load and approx 10% GPU memory. I'd like to try multiple WUs at a time, but it doesn't look like I have much extra room.

Part of the gain are the few seconds where the WU runs on the CPU and your GPU is idle. With faster GPUs that occur quite often, in your case every 90 seconds. If you'd be running two WUs at the same time, your GPU would never be idle.

Also using app_info allowes you to tweek the app a bit more, for example disable checkpointing.

---

Just curious what card you're doing this with. I'm running single MW WUs on a Radeon HD 6950. WUs take about 90 seconds at 98% GPU load and approx 10% GPU memory. I'd like to try multiple WUs at a time, but it doesn't look like I have much extra room.

I am right now using 2 Nvidia 560Ti's, but also have some AMD 5870's that haven;t been here in a long time, but will work just fine. I have NOT tested them for single of double units though.

Part of the gain are the few seconds where the WU runs on the CPU and your GPU is idle. With faster GPUs that occur quite often, in your case every 90 seconds. If you'd be running two WUs at the same time, your GPU would never be idle.

Also using app_info allowes you to tweek the app a bit more, for example disable checkpointing.

I am moving more towards using the app_info file as opposed to the cc_config file, but it is a slow process for me. Too many machines means I have to log into each one and change it, test it and then perhaps change it back. It is working, for the most part, as it is and I am not into testing right now.

---

Matt -

I have a single EVGA GTX 660. Two Milkyway WU's run with less than 100% GPU usage.

Just curious what card you're doing this with. I'm running single MW WUs on a Radeon HD 6950. WUs take about 90 seconds at 98% GPU load and approx 10% GPU memory. I'd like to try multiple WUs at a time, but it doesn't look like I have much extra room.

On a 7950, running two at once only netted me a gain of roughly 7% in speed, but similar methods can be used to make a MW job combine well with other jobs/projects. I'm running two fast AMD cards with an AMD hexacore CPU, and find that if I want to run POEM on all of it, my system becomes totally CPU bound, and half of my cards' capacity go unutilized.

On the other hand, using default settings for something like MW or donate@home will get close to keeping my cards busy, while my CPU twiddles its thumbs.

Something like 3 POEM jobs and 1 MW (or other GPU-intensive, CPU-light job) on each card uses up everything very neatly & efficiently, and neither project's jobs are slowed much if you find the right balance between them.

Just curious what card you're doing this with. I'm running single MW WUs on a Radeon HD 6950. WUs take about 90 seconds at 98% GPU load and approx 10% GPU memory. I'd like to try multiple WUs at a time, but it doesn't look like I have much extra room.

On a 7950, running two at once only netted me a gain of roughly 7% in speed, but similar methods can be used to make a MW job combine well with other jobs/projects. I'm running two fast AMD cards with an AMD hexacore CPU, and find that if I want to run POEM on all of it, my system becomes totally CPU bound, and half of my cards' capacity go unutilized.

On the other hand, using default settings for something like MW or donate@home will get close to keeping my cards busy, while my CPU twiddles its thumbs.

Something like 3 POEM jobs and 1 MW (or other GPU-intensive, CPU-light job) on each card uses up everything very neatly & efficiently, and neither project's jobs are slowed much if you find the right balance between them.

Yes finding that sweet spot between projects and over or under utilizing the cpu and gpu's is quite a good feeling!! Sometimes though raw power can be fun too!

I am running a single MW client at a time, and starting a new one every 54 seconds ...

---

Go away, I was asleep

I have been trying to run the app_config file shown above and whenever I do the Boinc event log shows an error message saying "App file not found". Has anyone else had this problem?

I have been trying to run the app_config file shown above and whenever I do the Boinc event log shows an error message saying "App file not found". Has anyone else had this problem?

Is the file in the projects directory?

I have been trying to run the app_config file shown above and whenever I do the Boinc event log shows an error message saying "App file not found". Has anyone else had this problem?

Is the file in the projects directory?

Yes. I simply substituted it for the app_info file I normally use.