I want to ask a question to anyone who knows...

Which gpu is faster for crunching seti?
A GTX 285 (1242 ram) with 240 cuda cores, or a GTX 460 (1024 ram) with 336 cuda cores?

---

I want to ask a question to anyone who knows...

Which gpu is faster for crunching seti?
A GTX 285 (1242 ram) with 240 cuda cores, or a GTX 460 (1024 ram) with 336 cuda cores?

the 460 crunches real good.

---

I want to ask a question to anyone who knows...

Which gpu is faster for crunching seti?
A GTX 285 (1242 ram) with 240 cuda cores, or a GTX 460 (1024 ram) with 336 cuda cores?

the 460 crunches real good.

But compare to 285?

---

I want to ask a question to anyone who knows...

Which gpu is faster for crunching seti?
A GTX 285 (1242 ram) with 240 cuda cores, or a GTX 460 (1024 ram) with 336 cuda cores?

the 460 crunches real good.

But compare to 285?

If I had a choice I would go with the 460 card. I have 2 0f them. Get the 1gb of memory or more the more memory more wu at a time can be done.

---

You have 3 460 cards run 3 wu per card on them if you can.

---

No.. i do not have 3 cards.
2 of them are 285's.

But i would like to change them if 460's are better.

---

imo there is no huge gap earning credits between them, but the 460 consumes much fewer energy and runs cooler, so if u can change, do it.

---

- Performance is not a simple linear function of the number of CPUs you throw at the problem. -

No.. i do not have 3 cards.
2 of them are 285's.

But i would like to change them if 460's are better.

By pure specs alone, the 460 is better than the 285.

240 cores vs 336
648mhz vs 675mhz
1242mhz GDDR3 vs 1800mhz GDDR5

On your machine with the two 285's and the single 460, your times are pretty similar between the cards. However, you could get about a 40% increase in throughput on the 460's by switching to 2 WU's per card. The 285's will actually decrease throughput when running more than 1 WU at a time, so you'd have to pull them out before switching.

I run 3 460's on my i7 machine, and when I can keep a constant workload, the machine has been as high as about 62-63k RAC, and was still climbing when I had to shut down due to outside issues. I'd say it would've peaking in the upper 60's.

---

It's been a long time since I ran a 285, but back then it ran at the same speed or a little faster than a 460 does now. There have been changes to the scale though so it is really hard to make a comparison. You can't really go wrong with a 460 though...less power draw and pretty damn fast with the newer fermi apps.

Hi all.

Just wondering if changing video cards out in one machine would cause a mass of invalid results, as I changed out a old GTX 260 OC for a GTX 560 TI and the work units finished in a flash but I wound up with all of them being tagged as invalid?

---

Yes, you have to make sure that your optimised application is FERMI compatible.

Gruß,
Gundolf

Hi all.

Just wondering if changing video cards out in one machine would cause a mass of invalid results, as I changed out a old GTX 260 OC for a GTX 560 TI and the work units finished in a flash but I wound up with all of them being tagged as invalid?

Your Invalids are because you are using the old V12 opt app (it doesn't support Fermi hardware) so update it to the latest version and that will cure your problem. ;)

Cheers.

---

Downloaded and installed the latest I could find on the Lunatics site 0.38 installer, see if that helps when some cuda work becomes available, thanks for the reply guys :)

---

Downloaded and installed the latest I could find on the Lunatics site 0.38 installer, see if that helps when some cuda work becomes available, thanks for the reply guys :)

Certainly should - that's definitely the right app for your card.

Just keep an eye on the first few results back when work starts flowing - don't fiddle with anything else until you're sure that the basics are working properly.

You may find that there are still hurdles to be overcome - some models of the GTX 560 Ti have been sent out from the factories with settings that aren't quite right for the sort of intensive punishment we put them through here. One of the hardware guys can sort you out there, if needed - you would just need one of the software tuning packages which can adjust the voltages on the graphics chip, if needed. But see how yours goes with plain vanilla settings first, and ask if you see anything odd.

I've noticed that some (I haven't checked closely enough to see if it's all) of my GPU units are coming back "error while computing." The card in my new computer wasn't the top of the line when I bought it last spring, but it was fairly good and met my other requirements. I currently show 7 errors and all of them are CUDA_Fermi. What causes this and is there anything I can do about it?

David

---

David
Sitting on my butt while others boldly go,
Waiting for a message from a small furry creature from Alpha Centauri.

I've noticed that some (I haven't checked closely enough to see if it's all) of my GPU units are coming back "error while computing." The card in my new computer wasn't the top of the line when I bought it last spring, but it was fairly good and met my other requirements. I currently show 7 errors and all of them are CUDA_Fermi. What causes this and is there anything I can do about it?

David

Sounds like your card may have bad - can you test it on some non-SETI work?

---

I've noticed that some (I haven't checked closely enough to see if it's all) of my GPU units are coming back "error while computing." The card in my new computer wasn't the top of the line when I bought it last spring, but it was fairly good and met my other requirements. I currently show 7 errors and all of them are CUDA_Fermi. What causes this and is there anything I can do about it?

David

All seven are "Exit status -12"

That's nothing to worry about. There's a minor incompatibility between the data sent out by the project, and the applications - also sent out by the project. You'll get bitten by it once in a while, but it doesn't imply there's anything wrong with your card or your computer.

If you want to use a sledgehammer to crack a very small nut, you can install the optimised applications that Spectrum described - that doesn't quite eliminate the -12 problem, but reduces the frequency substantially. Or you can just carry on as before - it's not important either way.

Sounds like your card may have bad ...

No, I don't think so. See my note on -12 exit status.

Edit - he's got plenty of valid CUDA results. Card is A-OK.

All seven of your erroneous work units still shown in your task list have returned with a "-12 error":

SETI@home error -12 Unknown error
cudaAcc_find_triplets erroneously found a triplet twice in find_triplets_kernel

So, since that's only a small percentage of all your FERMI tasks, it's probably not the fault of your GPU!

Gruß,
Gundolf
[edit]Four minutes late ;-)[/edit]

Thanks, Richard and Gundolf. I won't worry about it then.

David

---

David
Sitting on my butt while others boldly go,
Waiting for a message from a small furry creature from Alpha Centauri.