You're totally right, seeing the 2nd GPU is not your problem.

And CUDA memtest.. well, would have been nice if it detected something. But memory doesn't fail that often over time - either it's broken from the beginning or it's the chip. Since your card worked fine and then started throwing errors, I' rather suppose the 2nd chip got damaged / decayed excessively.

72°C sounds fine.. but still, I'd test with the fan speed set to 100% (and leave the room). If that helps you have a clear case for RMA. If not I'd run the regular 3D stability tests I suggested above. If they fail you also have a clear case for RMA.

MrS

---

Scanning for our furry friends since Jan 2002

As a by note to fixing this maybe it means something who knows but under windows 7 my gtx 295 shows only 1 gpu to boinc no matter how i configure it. 2 monitors plugged in desktops extended and not. Any driver or version all only let boinc see the single gpu.

Just had a fun afternoon anyway playing quickly with W7 :)

As a by note to fixing this maybe it means something who knows but under windows 7 my gtx 295 shows only 1 gpu to boinc no matter how i configure it. 2 monitors plugged in desktops extended and not. Any driver or version all only let boinc see the single gpu.

Just had a fun afternoon anyway playing quickly with W7 :)

SLI is off?

I don't know how W7 handles SLI but there is an internal SLI mode on the 295 cards IIRC, XP Does not do that ... or at least none of my installs have done that (yet) ... then again, I have not had the courage to try Vista or W7 ...

THough I seem to increasingly hit the upper bound in memory which in XP is 2.something G ...

My system doesn't need to be tricked into "seeing 2 gpu's" there are 2 or it couldn't be failing or starting on the second one.

I meant working on 2 GPU's.
What I've seen elsewhere is that the second GPU being the only one to throw errors is the Drivers/CUDA are not happy with the config. This is why I've been suggesting what I have. I'm sorry if I sounded like a broken record.

Well, I have 3 systems all with 2 or more GPUs running and my error rate is vanishingly small. I mean in almost all cases of error it has turned out to be a bad task or some event where I have caused an issue.

To put it another way, I don't see a problem here ...

how can you not see a problem with the 2nd gpu always erroring w/u Paul. Just because your system is not erroring constantly then all systems are ok? i dont understand that statement.I had the same errors as you to start, the occasional w/u here and there. Then bang, all w/u that 2nd gpu does fail everytime.

Look through my tasks and try to find the last completed w/u on gpu 1.


Popandbob. I salute you for your suggestions i dont mean to be grumpy honest. I can only see the one way of changing sli config. There is no bios option for it and nvidia control panel just has the 2 important ones, SLI and Physx.

I'm nearly back to where i started again with no light at end of tunnel yet but am about to try fresh instal of 64 vista and 182.50 drivers first. Just doing the last updates before i instal boinc again.

OK i'm 2hours 46 minutes into running 2 tasks on it.
723126
723205
Fingers crossed and they both finish so i can sleep easy.

May sound silly but its always failed by now...........

The difference.........

Turned Aero interface off........

I'm just running on the vista basic theme so all the fancy graphic card utilising bumf is off

Could it be that simple for my problem!

how can you not see a problem with the 2nd gpu always erroring w/u Paul. Just because your system is not erroring constantly then all systems are ok? i dont understand that statement.I had the same errors as you to start, the occasional w/u here and there. Then bang, all w/u that 2nd gpu does fail everytime.

Here, as in my house ... I guess I was not clear ... obviously...

I did not say that there was not a problem THERE ...

The way I interpreted the earlier posts was that all systems with 2 or more GPUs are seeing errors on the second GPU. I was saying that I don't see that problem, nothing more or less ...

Sudden changes like what you report sound to me more like a failure of the card than anything else. Do you have other cards? I mean, can you swap things around?

As a by note to fixing this maybe it means something who knows but under windows 7 my gtx 295 shows only 1 gpu to boinc no matter how i configure it. 2 monitors plugged in desktops extended and not. Any driver or version all only let boinc see the single gpu.

Just had a fun afternoon anyway playing quickly with W7 :)

SLI is off?

I don't know how W7 handles SLI but there is an internal SLI mode on the 295 cards IIRC, XP Does not do that ... or at least none of my installs have done that (yet) ... then again, I have not had the courage to try Vista or W7 ...

THough I seem to increasingly hit the upper bound in memory which in XP is 2.something G ...

I tried W7 with 2 GTX285's and could only get boinc to see one card regardless of what settings I made.
Has anyone gotten W7 to use more then one GPU?

I'm sure its not the card.
I see the new host (37017)
currently 4 errors all on the second GPU of the 295.
Nothing yet from the first GPU...
Sounding more like a config issue as there would be errors from the first GPU as well if something was wrong...

Bob

I have a PC with a gtx295 and a 8800GT. The GPUs assignation is:

GPU#0: gtx295
GPU#1: 8800GT
GPU#2: gtx295

I get many erros with the WUs that are processed in GPU#1. I thought it was dued to mixed GPU chips (g92 and gt200). I changed the 8800GT with a gtx280 with this new GPU assignation:

GPU#0: gtx280
GPU#1: gtx295
GPU#2: gtx295

The same, many WUs being processed on GPU#1 give error. Note that GPU#1 is a different graphic card in each case.

I thought it could be the mobo. I installed the gtx295 in another PC (with only one PCIe for graphics). The GPUs assignation was:

GPU#0: gtx295
GPU#1: gtx295

The same: many WUs being processed in GPU#1 give error...

Windows Vista Home Premium 32 bits and Nvidia Driver 182.50 in boht PCs.

Now I have the PC with gtx295 + 8800GT configured so It only process WUs in the gtx295 (GPU#0 and GPU#2) (the GPU#1 -> 8800GT not processing) and I have NO problem (no errors) but I have one Graphic Card wasted...

Its in a parcel at the post office atm :) will update as it pans out

Edboard,

oh.. what a ****** mess! What does that mean for other people? Take a close look at which GPU errors out. Otherwise.. it can't be a general problem, as some people are running fine. That makes me point to the combination of OS and driver. Would you mind checking one of the 185 series drivers?

MrS

---

Scanning for our furry friends since Jan 2002

My problem started very suddenly and nothing i did changed anything in the slightest, drivers, O/S, whole different pc which has none of the problems this card gave me and i still get the same result.

Before this started the GTX295 was running both w/u with just the occasional fail on either core so either something is wrong enough with the gpu to fail w/u which it does in 2 different systems or who knows. the other system i tried it in is running 2 w/u on its own without a problem until i put this card in then it also fails on GPU 1.

I want to state with facts that many WUs processed in any GPU assigned #1 fails, mainly with the newest units.

I begun to have this kind of problems some time ago with Folding. But NO with GpuGrid/SETI. Now I have this problem with GpuGrid too. SETI works fine.

It's Gainward card??? I have ONE GTX 295 from Gainward - one GPU works fine GPUGrid and the other crashed after a few seconds. But its ok for short WUs like SaH. And this not a problem from XP, Vista, Linux or the driver/Boinc version or an unused monitorport, it's the card... Perhaps you can chance it to another...

ok, you are getting the same errors as I was.

I tried mine in 2 computers (with different OS's) and errored on both.

I suspected I had a Bad card but I can run benchmarks with no errors.

I'm just waiting for the argument with customer support when they say they can find nothing wrong. I have a feeling they wont like my answer that its not fit for the purpose i bought it for, namely gpugrid.

Cant be long now till they get back in touch with me, they have had it since Thursday afternoon :(

Check what kind of slots you have your cards plugged into on the motherboard. Not all PCI-E slots are the same. Some run at x16, some at X8, some at X4, etc. I searched hard when shopping for my motherboard to find one that had 3 full speed x16 slots and that was also capable for triple SLI (though you need SLI disabled when running CUDA like GPUgrid). Then, I made sure to use the x16 slots for my cards. Oh, and the x16 slots have to be spaced correctly too since the newer GTX200 series cards are double thick, or they won't fit. Then, of course, the case has to have room for that setup. Check for motherboard BIOS and chipset driver updates as well. Anyway, just a thought.

I'm getting a ton of errors with my gtx 295 as well. =[
I just started gpugrid, and this doesn't make me very happy.
I gave up on using the 2nd gpu and just tried using the first.... and the 1st gpu errors out everything... some within second and some after 20 minutes.
My card is not overclocked, and the temps are in a very good range.

Win7, 185.55, 6.6.31

Annoying to say the least.

The product GX-191-AS on RMA RMA****** has been replaced, this will be sent on a new order with order number *******.

This will be despatched as soon as possible, subject to stock availability.

Well thats that sorted....they tested the card and the 2nd gpu was indeed faulty. It was ok for a bit then the fault showed under load.

Pharg my slots are both PCIe 2.0 16x so that wasn't the problem.

My system ran ok for 2 weeks till the fault showed itself. There where no system/driver changes to blame between working and not working.

Thanks to you all for keeping me sane while i waited for their decision.

:) cant wait to take my emergency EAX800 XT out and slide a 295 back in. It will be a lot quieter believe it or not :)