Post at Guru3D.com here, with links to files for XP32, Vista/7 32 and 64. It shouldn't be too long until they appear on nVidias site. Anyone feel brave enough to try them?

An interesting side note: in post 24 a user complains that all 190 series drivers are switching clock speeds in an uncontrolled way on his system with Win 7 and a GTX 285. In games as well as if its just sitting on the desktop. This sounds similar to what PoorBoy and others have been reporting.

MrS

---

Scanning for our furry friends since Jan 2002

I have 2 GTX 260s. One with 216 shaders is happy with 190.38. The other with 192 shaders was never happy with anything except 182.xx so was on F@H. I have installed 190.62 on the 192 260 and ran OCCT for awhile and looked better than ever did, no artifacts or driver resets. So it has some GPUGrid WUs and we will see what happens. Fingers crossed.

Vista 64 and Boinc 6.6.36 on both.

1st one errored out. Back to F@H for this card. :-(

Seems likely for me too.... the combination of 190.62, 6.6.38, and 260 (192's) has completed 2 out of todays allotment of wu's the rest all errored out

First WU (73-KASHIF_HIVPR) on new GTX 260 and 190.62, BOINC v6.10.0 finished A-OK.

Post at Guru3D.com here, with links to files for XP32, Vista/7 32 and 64. It shouldn't be too long until they appear on nVidias site. Anyone feel brave enough to try them?

An interesting side note: in post 24 a user complains that all 190 series drivers are switching clock speeds in an uncontrolled way on his system with Win 7 and a GTX 285. In games as well as if its just sitting on the desktop. This sounds similar to what PoorBoy and others have been reporting.

MrS

I did get 3 260 Cards back Friday on a RMA and put 2 in 1 Box, within minutes 1 Wu erred & when I checked 1 of the Cards had reset it's self to a lower clock speed. I reset it but when I checked later 1 or 2 more Wu's had erred too so I just switched back to Collatz on that Box & it's been running fine since. I'm going to try the 190.62 Drivers & a few more Grid Wu's later on it but I'm not expecting any miracles at this point.

First WU (73-KASHIF_HIVPR) on new GTX 260 and 190.62, BOINC v6.10.0 finished A-OK.

Four completed out of 4 now, v190.62 is working fine here.

http://www.gpugrid.net/results.php?hostid=48008

The 190.62's didn't do anything for me, I tried them on 2 of the Cards I got back on as RMA replacements. 1 of the Cards wouldn't run the Wu's for more than a minute or two before erring the Wu. The other card ran for awhile & then erred & reset itself to a lower clock speed or just the opposite by resetting & then erring the Wu.

Poorboy, what brand are the bad cards?

I did get 3 260 Cards back Friday on a RMA and put 2 in 1 Box, within minutes 1 Wu erred...

I don't want to sound arrogant.. but this is just what I expected :/

MrS

---

Scanning for our furry friends since Jan 2002

Poorboy, what brand are the bad cards?

10 bucks says EVGA!

1 was a EVGA & 4 were BFG's, still waiting for 2 RMA'ed BFG's, all the cards run the Collatz Wu's okay so that's what I'm running on them for now.

Poorboy, could you try to generate some errors on one machine with a 190 drivers so that we can look at the problem.

gdf

my 250 and 9800GTX+
have bsd WU
but Collatz Wu's okay (((

---

I have a 9800GTX+ running ok under 190.62 - at least it is now. GPUGRID stresses the hardware much more than other CUDA projects, you can and do get the situation where other WUs that stress the hardware less, run, and GPUGRID's dont.

It often occurs when the driver is corrupted. Deinstall 190.62 via windows, reboot into safe mode, run Guru3D's Driver Sweeper and get rid of all the old driver bits you will find there, reboot (it will go into 640 mode - thats fine), then reinstall 190.62.

There is yet another example of this kind of corruption over on the number crunching forum - in that case it was SETI CUDA running, and GPUGRID didnt.

Regards
Zy

Today I lost 2 WUs on my GTX295 after a emergency shutdown. The card runs at 95°C (normal is 68-72°C) and 100% fan speed (normal is 55%). After rebooting the next WU runs fine, like all the other WUs since installing the 190.63.

The error was in line 66 (Mersenne Twister) in this WU

1179359

The other core had a fft_data_swizzle_in error in line 78

1178552


Under this circumstands (card is running hot) it looks difficult to run GPUGrid without babysitting. Such a card is not a cheap card :(

---

Can you be sure the card reached this temperature due to GPU-Grid? I don't think they changed their algorithm that drastically. Heating a chip up that much more would require an enormous increase in code efficiency (granted you don't start running useless loops).

Could it be that the fan control software failed to set the fan to 100% due to some hickup?

MrS

---

Scanning for our furry friends since Jan 2002

i'm going to say that the auto fan did auto adjust to the temp.

That's what the software reported. But is it true?

MrS

---

Scanning for our furry friends since Jan 2002