I have my computer set to use my GPU while I am on it, since I do not leave my computer on unless I am on it.

While im researching something online or coding something this is fine. However when I build my game since the GPU is at 100% usage it wont run <2fps. So I have to deal with the hassle of pausing the task and then resuming it after debugging a little (about every 5 mins)

Is there a way to set the GPU usage in Boinc to about 75%?

Also I am just curious I am using a 9800GT. Meaning 30Hrs per task. Is this too slow to even donate? Is there a project with smaller tasks?

Thanks in advance,
CoderWalker

I have my computer set to use my GPU while I am on it, since I do not leave my computer on unless I am on it.

While im researching something online or coding something this is fine. However when I build my game since the GPU is at 100% usage it wont run <2fps. So I have to deal with the hassle of pausing the task and then resuming it after debugging a little (about every 5 mins)

Right click on the task bar icon and select Snooze GPU or Snooze, if you are going to game for a short time. If gaming longer exit Boinc or if you are an advanced user add exceptions to Boinc config file.

Is there a way to set the GPU usage in Boinc to about 75%?

That would not help you game.

Also I am just curious I am using a 9800GT. Meaning 30Hrs per task. Is this too slow to even donate?

Probably, saying as you don't leave the computer on when you are not using it.

Is there a project with smaller tasks?

- This project has the longest tasks. You might want to try Folding@Home (non-Boinc) or Einstein.

Is there a project with smaller tasks?

You might want to try MilkyWay (Boinc) or Folding@Home (non-Boinc).

MilkyWay requires double precision cards, so the 9800 isn't suitable.
Dnetc and Collatz will work, if you like the science. I don't, so I don't recommend, but it's a possibility.

Folding would be my recommendation for windows computers.

---

Gruesse vom Saenger

For questions about Boinc look in the BOINC-Wiki

Einstein is now also worth considering when your machine is running Windows: they have recently made considerable improvements to their CUDA application.

Thanks I have switched to the Einstein@Home.

What I mean is can I limit the Gpu usage to say 80%?
When it uses 100% of my Gpu my whole Windows interface lags terribly.

Also PrimeGrid is OK for this card.

Gives (PrimeGrid 2300) more credits then Einstein (500).

Takes about 20 minutes per WU!

---

Ton (ftpd) Netherlands

Also PrimeGrid is OK for this card.

Gives (PrimeGrid 2300) more credits then Einstein (500).

Takes about 20 minutes per WU!

Depends which excites you more - prime numbers or binary pulsars!

I ran an Einstein task and it only used 40% of the GPU, so you would not have to worry about trying to throttle it.

Unfortunately the credit systems of Boinc using Projects are not universal, and are therefore meaningless when comparing one project to another:
If your 2300 credits for a 20min task on a GT240 is accurate, then 165,600 credits per day for a GT240 is just being silly. I think this undermines the credibility of Boinc.

So it is best to run a science project that you are happy to contribute to.
For me the maths projects are of no interest whatsoever, and I don’t care to find ET either. I certainly wouldn’t spend money on a GPU never mind the electric for those projects - even if they added ten more zero’s to their stupid credit systems. I barely contribute to the Physics, materials, chemistry and astronomy projects, but at least they are real science and some could have an impact on health, medicine, poverty...
If crunching brings a cure for cancer, HIV, malaria... then it’s worthwhile. If crunching finds another prime number, or even an M class planet 50M light years away, so what?

Of course it's up to the cruncher to decide, credits or credibility.

I ran an Einstein task and it only used 40% of the GPU, so you would not have to worry about trying to throttle it.

Unfortunately the credit systems of Boinc using Projects are not universal, and are therefore meaningless when comparing one project to another:
If your 2300 credits for a 20min task on a GT240 is accurate, then 165,600 credits per day for a GT240 is just being silly. I think this undermines the credibility of Boinc.

So it is best to run a science project that you are happy to contribute to.
For me the maths projects are of no interest whatsoever, and I don’t care to find ET either. I certainly wouldn’t spend money on a GPU never mind the electric for those projects - even if they added ten more zero’s to their stupid credit systems. I barely contribute to the Physics, materials, chemistry and astronomy projects, but at least they are real science and some could have an impact on health, medicine, poverty...
If crunching brings a cure for cancer, HIV, malaria... then it’s worthwhile. If crunching finds another prime number, or even an M class planet 50M light years away, so what?

Of course it's up to the cruncher to decide, credits or credibility.

You say "For me the maths projects are of no interest whatsoever,".

OK it's up to the cruncher to decide, but i think that this project is based on FFT (Fast Fourier Transform) and this is a mathematical object.

Best Regards and Happy crunching

You say "For me the maths projects are of no interest whatsoever,".

OK it's up to the cruncher to decide, but i think that this project is based on FFT (Fast Fourier Transform) and this is a mathematical object.

Best Regards and Happy crunching

Skgiven is placing more emphasis on the end result rather than the process. Yes, a lot of distributed computing projects use the FFT, but that is mainly used as a means to an end goal.

Personally, I contribute to DNETC@home whenever GPUGrid goes down (distributed.net was my first distributed computing project seven years ago). But I agree with skgiven that more emphasis should be placed on projects that would result in an improvement of the human condition.

You say "For me the maths projects are of no interest whatsoever,".

OK it's up to the cruncher to decide, but i think that this project is based on FFT (Fast Fourier Transform) and this is a mathematical object.

Best Regards and Happy crunching

Skgiven is placing more emphasis on the end result rather than the process. Yes, a lot of distributed computing projects use the FFT, but that is mainly used as a means to an end goal.

Personally, I contribute to DNETC@home whenever GPUGrid goes down (distributed.net was my first distributed computing project seven years ago). But I agree with skgiven that more emphasis should be placed on projects that would result in an improvement of the human condition.

OK, i respect your position. I think different.

Is there any plans to release smaller WUs? I tried to run the standard WU and after 11 mins the WU was only barely got any work done. Based on percent the done, the WU would take 4 days to finish. I don't mind leaving the machine one 24/7, but WUs shouldn't take more then 1 day.

If not, can someone recommend another medical project that uses GPU? I didn't see any on the boinc page.

edit grammer

Is there any plans to release smaller WUs? I tried to run the standard WU and after 11 mins the WU was only barely got any work done. Based on percent the done, the WU would take 4 days to finish. I don't mind leaving the machine one 24/7, but WUs shouldn't take more then 1 day.

If not, can someone recommend another medical project that uses GPU? I didn't see any on the boinc page.

edit grammer

Folding@Home is the only other med project I know of. What GPU are you running?
Cheers.

---

Dataman, I am using an old Nvidia Quadro Fx 1800.

I see. You are not going to find many projects for that card. You might try Einstein (binary pulsars) as they are short and have small files. No med projects though. Here they are moving toward even larger files for high-end cards. Cheers.

---

I've seen a program that claims to be able to limit GPU use:

http://www.efmer.eu/boinc/

I've installed it but haven't found time to read the manual yet.

So far, it seems to limit BOINC CPU use too much, but does limit the temperatures as claimed.

Please note that Folding@Home is a non-BOINC project.

Also note that POEM@Home, which IS a BOINC project related to medical research, is planning to start offering GPU workunits "soon". No details yet. Currently CPU-only, so they shouldn't insist on fast GPUs especially soon.

You say "For me the maths projects are of no interest whatsoever,".

OK it's up to the cruncher to decide, but i think that this project is based on FFT (Fast Fourier Transform) and this is a mathematical object.

It's a little off topic, but I'd like to throw my 2 cents in here anyway:

Math is extremely important. Without analytics we'd be wasting our time solving "trivial" problems numerically. And without numerics our other "real world research" would be horribly inefficient.

However, that's not what the current GPU based BOINC math projects are about. And that's where I agree that I see little value in them:

- DNETC: we know a solution exists. We've developed, tested and benchmarked the algorithm to find it via brute force. The time t needed to find the solution can be calculated easily for any amount of crunching power available. On average we're going to find the solution after t/2. However, if it's going to be t/2, tomorrow or t is entirely random / good luck / bad luck. And finding the solution brings absolutely no gain in knowledge. Personally I think this project is about the largest waste of electricity which ever happed to distributed computing.

- Collatz: we can not prove the Collatz Conjecture in a mathematical sense by number crunching. All we can do is to verify it for a certain finite range of numbers. Or find a counter example, which would be great and might explain why none has been able to prove it up to now, but seems unlikely given the range we've already covered.

- PrimeGrid: finding new primes is.. nice. But at least it's an intrinsic property of nature :)

MrS

---

Scanning for our furry friends since Jan 2002

I agree with you, and would be running some BOINC project related to medical research on the GPU of my laptop if I could find one that could use a G105M.

It seems that BOINC insists on looking for GPU workunits first if there is a GPU that is possible for workunits to use, and only proceeds to looking for CPU workunits after it has found enough GPU workunits. Collatz Conjecture seems to be the best provider of those GPU workunits I've found, even though I don't especially care if their project is ever finished or not. Therefore, I have my laptop running Collatz Conjecture GPU workunits for now, even though I plan to switch to running BOINC GPU workunits related to medical research as soon as I can find a suitable BOINC project. I'm also looking into learning a few more computer languages so that I can offer some programming help to starting such a project, but I'll need to choose computer languages for which enough online instruction is available.

If a future BOINC version offers a way to tell it to go on to requesting CPU workunits faster, I'll consider whether to save electricity by using that instead of Collatz Conjecture.

You can deselect GPU usage via your cc_config.xml file:
Add <no_gpus>1</no_gpus> to the <options> section of the cc_config.xml file.

If POEM@Home wasn't just starting toward GPU workunits, I'd do that.