Hello everyone,

Another error issue here. I was watching the results of a friend of me and I saw it throws the error Incorrect Function (Onjuiste Functie in Dutch) way too much. I couldn't find a solution for it on this forum either. You can check his results here: http://www.gpugrid.net/results.php?hostid=16213

It looks like it only errors out on a specific kind of WU, and quite fast after it begun. AFAIK (I'll ask him soon) the card is stock clocked and it is about half a year old. It's a single card, the 9800 GTX+ which runs on Windows Vista x64, BOINC 6.4.5.

Thanks in advance!

That's the same error I get in the second GPU of my GTX295 dual card. All WUs which grant 3718 points end in "Compute error":

<message>
Funci�n incorrecta. (0x1) - exit code 1 (0x1)
</message>

Sometimes at the very beginning but normally after a long time of processing.

But, as I have said, only when processed by the second GPU, not the first one.

Windows Vista Home Premium 32 bits
Nvidia Drivers 180.87
Diferent versions of Boinc, 6.5.0, 6.6.3, etc.

You can see it here: http://www.gpugrid.net/results.php?hostid=17157

There are a lot of them...

Looking at the overall results for this type of WU, the failure rate doesn't seem to be specially higher...
And looking at these failed WUs, they get successfully completed at some point. If they were wrong they would all (and always) fail.

And remember that apparently some cards do actually come already (slightly?) overclocked from the manufacturer...

ignasi

My gtx295 has no overclock (nor mine nor from factory as I can see with GPU-Z) and, what it's the strangest issue, it only gets "compute error" with one kind of units and in the same gpu. With any other unit this gpu works fine.

I have just installed the new beta Nvidia drivers 182.05 and it seems it is better than 180.87. With this gtx295 (EVGA) I could not install drivers 181.20 nor 181.22. They simply didn't work. But with this one (182.05)I can now configure more easily than before the gtx295 to work as two independent gpus. I shall try gpugrid again tomorrow.

Could you be a bit more specific about the "easier to configure"? Do you mean it just works as expected? If so it would be worth to start a new thread for this (with a title which helps 295 users to find it).

MrS

---

Scanning for our furry friends since Jan 2002

Signed up awhile ago, and just now have the proper h/w, s/w, and i-net connection to actually start contributing.

9600gso, stock clock.

what gives ?!

i suspended gpugrid, to avoid failing on every single wu that gets thrown to me.

whats happening ?!

Which driver do you use? Judging by the error message it may be too old.

MrS

---

Scanning for our furry friends since Jan 2002

ok, d/l'ing 181.22_xp32 now...

EDIT --> Updated, but still don't know if that was the problem. Apparently there is a 7 wu / day limit, and I am unable to d/l any new wu's to see if the updated driver will allow me to crunch a wu w/o erroring out...

The actual limit is higher, but it gets reduced when you repeatedly return erroneous results.

MrS

---

Scanning for our furry friends since Jan 2002

He has now updated the drivers from 180.84 to 181.22, but it seems even worse right now...

There are 2 strange issues in his task output: his 9600GSO reports 128 shaders instead of 96 and the "feature is not yet implemented [by the driver]" error message. Something must be very strange about his software setup.. or the hardware? Is his 9600GSO based on a G80 chip?

Bad Peng, would you mind running GPU-Z and tell us which chip it detects? It's a small utility which should be easy to find.

MrS

---

Scanning for our furry friends since Jan 2002

Thrashing through the paperwork, It looks like he has the G92 9600 GSO w/384mb...unless the detection script in the server is screwed-up. But it still shows as having 128 sp's in the canonical result...Hmmmm

I don't know of a G80 9600 GSO...But there is a new G94 9600 GSO, with 48 sp's on the market....

That would be the question, which one of those does he have for real..? Or is his driver wonky..?

---

Consciousness: That annoying time between naps......

Experience is a wonderful thing: it enables you to recognize a mistake every time you repeat it.

He shows 384mb memory on the card. All the G94 based 9600GSO cards are either 512mb or 1G, so it is definitely a G92 (though the 8800gs can falsely report as a 9600GSO I think--same card basically and still with 96 shaders).

I think the following line might point to something amiss with the card itself:

"Failed to set low-cpu sync mode"

Is this maybe installed on a motherboard with NVIDIA onboard video with its own unified shaders? I wonder if it is possible that an onboard video might not be turning itself off properly?

Well, this message goes hand in hand with the "feature not yet implemented" as its direct cause. That's why I assumed an old non-CUDA driver.. which 181.22 certainly isn't.

MrS

---

Scanning for our furry friends since Jan 2002