That would be great!
Can't crunch GPUGRID anymore on my GTX260 since you switched to CUDA 2.2...
____________

pixelicious.at - my little photoblog

Before the end of the month, we will try to see if cuda2.3 fixes some outstanding fft bug that people with gtx260 are experiencing.

Internal tests showed that it could be the case, but then everyone must run on 190.xx drivers.

No schedule for the change yet.

gdf

I'll wait for it! GTX260 from Gigabyte (192 cores) can't process gpugrid's Wus on x64 system.

But with this bug I have crunched few SETI cuda WUs and found a "Collatz Conjecture" project to make my gpu not be idle... :-)
____________
From Siberia with love!

Always me with one GTX 260+ + 8800 GT on vista 64 and boinc 6.6.36

I try 190.38 driver. And I have always errors. The sames since driver superior than 182.5.
With 182.5 my two graphics cards works well.

Why why why....

My WUs don't complete in 24h anymore,
are the new WUs longer or is the app slower or is it my computer?

Hi,
for a couple of months i had no problems with boinc, Seti, Gpugrid and vista64. But when i came back from vacation a week ago, something strange happened. Windows did a lot of updates and the graphic-driver (?) reports, that a reset was necessary. A running gpugrid-wu reports an error. Most seti-wu work, but not all. This driver-reset happens 3 or 4 times a day.
I updated to the latest nvidia-drivers and BOINC and tried to understand, what other boinc-users report about ther problems. I've seen, that the GT260 seems to have a problem, but i do not understand all the other post about SDK, upcoming versions aso.
I would be really happy, if anyone could explain, what happend and which solutions are available in words, that are understandable for non-DSK users and people, who do not think, that running BOINC is the primary goal for a computer.
Kind regards,
Alexander

Surely I have cuda 2.3

And I have always the same errors like other drivers sup than 182.5 ---->

14/08/2009 21:59:47 GPUGRID Output file 52-KASHIF_HIVPR_sub_so_ba1-15-100-RND7514_0_1 for task 52-KASHIF_HIVPR_sub_so_ba1-15-100-RND7514_0 absent
14/08/2009 21:59:47 GPUGRID Output file 52-KASHIF_HIVPR_sub_so_ba1-15-100-RND7514_0_2 for task 52-KASHIF_HIVPR_sub_so_ba1-15-100-RND7514_0 absent
14/08/2009 21:59:47 GPUGRID Output file 52-KASHIF_HIVPR_sub_so_ba1-15-100-RND7514_0_3 for task 52-KASHIF_HIVPR_sub_so_ba1-15-100-RND7514_0 absent

And Why I have always this message ?


With 182.5 I crunch witout pb. with other driver it's catastrophe. all wus crash

There is a lot of mess happening to my systems also...
1
winxp32
driver 190.
9600gt

2
winxp32
driver 190
9800gt

a lot of units fails now although (especially on the 9600gt) it worked flawless before. will got to nnw for gpugrid, try collatz

regards formulaman

My WUs don't complete in 24h anymore,
are the new WUs longer or is the app slower or is it my computer?

Well, is there any progress with cuda 2.3? When we will be able to start crunch again with our 260s? :-)
____________
From Siberia with love!

I'll grab one and check it out.

vista 32-bit, gtx 260. 190.38

edit: it's running.....

to me, it could just be the wu is a short one, or it's running faster with 2.3.....

3(rt) minutes = .5 %

link to test wu in progress:

http://www.gpugrid.net/result.php?resultid=1135107

Task 1134614 arrived safely with a cuda23 designation - including the 8MB cufft download!

I've bumped it to run immediately, but still estimating 18 hrs on the 9800GTX+. BOINC v6.6.37, nothing else needed (except the drivers) to get it to download.

I am the user of this day!
I will try on one gpu with errors:
20.8.2009 20:52:08 CUDA device: GeForce GTX 260 (driver version 19038, compute capability 1.3, 896MB, est. 96GFLOPS)
Task: http://www.gpugrid.net/result.php?resultid=1137304

est. 96GFLOPS ????


Bymark
____________
"Silakka"
Hello from Turku > Åbo.

I am the user of this day!
I will try on one gpu with errors:
20.8.2009 20:52:08 CUDA device: GeForce GTX 260 (driver version 19038, compute capability 1.3, 896MB, est. 96GFLOPS)
Task: http://www.gpugrid.net/result.php?resultid=1137304

est. 96GFLOPS ????


Bymark

Yep, that is a stock core value..
It is Now as it arrived from production line from china!
Testing,and edit: I have many friends from china it is not a country fault.

Regards

Bymark

Can't start ?

20.8.2009 22:02:24 Starting BOINC client version 6.6.36 for windows_intelx86
20.8.2009 22:02:24 log flags: task, file_xfer, sched_ops
20.8.2009 22:02:24 Libraries: libcurl/7.19.4 OpenSSL/0.9.8j zlib/1.2.3
20.8.2009 22:02:24 Data directory: C:\Documents and Settings\All Users\Application Data\BOINC
20.8.2009 22:02:24 Running under account tbymark
20.8.2009 22:02:24 Processor: 2 AuthenticAMD AMD Athlon(tm) 64 X2 Dual Core Processor 5600+ [x86 Family 15 Model 107 Stepping 2]
20.8.2009 22:02:24 Processor features: fpu tsc pae nx sse sse2 3dnow mmx
20.8.2009 22:02:24 OS: Microsoft Windows XP: Professional x86 Edition, Service Pack 3, (05.01.2600.00)
20.8.2009 22:02:24 Memory: 2.00 GB physical, 3.85 GB virtual
20.8.2009 22:02:24 Disk: 244.14 GB total, 234.25 GB free
20.8.2009 22:02:24 Local time is UTC +3 hours
20.8.2009 22:02:24 CUDA device: GeForce GTX 260 (driver version 19038, compute capability 1.3, 896MB, est. 96GFLOPS)
20.8.2009 22:02:24 Not using a proxy
20.8.2009 22:02:24 Collatz Conjecture URL: http://boinc.thesonntags.com/collatz/; Computer ID: 1201; location: (none); project prefs: default
20.8.2009 22:02:24 GPUGRID URL: http://www.gpugrid.net/; Computer ID: 35303; location: (none); project prefs: default
20.8.2009 22:02:24 No general preferences found - using BOINC defaults
20.8.2009 22:02:24 Reading preferences override file
20.8.2009 22:02:24 Preferences limit memory usage when active to 1228.30MB
20.8.2009 22:02:24 Preferences limit memory usage when idle to 1944.81MB
20.8.2009 22:02:24 Preferences limit disk usage to 100.00GB
20.8.2009 22:02:24 Suspending network activity - user is active

Edit: Running, slow but running....
____________
"Silakka"
Hello from Turku > Åbo.

2.3 Running good on XP64 with 2 EVGA 260s, previous version ran good as well.

I've noticed that running this new app, my gpu is running about 1-2 C cooler.

61 C.

Spoke too soon, having some compute errors..

2.3 Running good on XP64 with 2 EVGA 260s, previous version ran good as well.

Got one for test. 11 minutes - still normal flight.

Confirming that GPU core is working on appr. 4°С higher temperature. After manual setting of GPU cooler rotation (70% of max.) it fell down from 80°С to 72°С (overclocked GTX260 from Gigabyte).
____________
From Siberia with love!

Nope, Error while computing 189.69

I am the user of this day!
I will try on one gpu with errors:
20.8.2009 20:52:08 CUDA device: GeForce GTX 260 (driver version 19038, compute capability 1.3, 896MB, est. 96GFLOPS)
Task: http://www.gpugrid.net/result.php?resultid=1137304

est. 96GFLOPS ????


Bymark

that card must have stock core value..

my card is est. 106 gflops with 216 core.

Error...
____________
From Siberia with love!

Knowing from the news line, that from today only 6.6.36 BOIC mngr will able to run GPUGRID with 190.38 cuda 2,3 dirvers, yesterday was make a several attepts to update the system to the new, recommended level. Also decide to install additional GTX 260 card (2 are almost running weil).
So, system: Q9550 \ ASUS WS Evolution MoBo, Windows server 2008 SP2, 2 x ASUS GTX 260. BOIC version 6.5.0 (previosly recommended on GPUGRID site), drivers version 185.25.
Install BOIC 6.6.36. Reboot. BOINC shows just one GPU. OK, put typical cc_config enabling use of all gpu's. Reboot. BOINC shows two GPU's. No one is running. Previous jobs (from the cache) stays untouched with the "Waiting to run" message. BOIC is regularily asking MW server for WU's for GPU (???). Yes, I'm running non ATI CAL (gpu) version on the rest Q9550 cores, all works fine. But was no any wish to replace MW job to my GPUGRID gpu's.
Hmm.
Decided to install 3-rd GTX 260, drop it into the system, boot, install 185.25 drivers again, checking:
Windows ( in the device manager) showing 3 x GTX 260.OK. Launching nTune, as well, as GPU-Z. Both are reporting about 3 x GTX 260. OK. GPU frequencies, as well, as memory timings are adjustable on all 3 GTX 260. Seems, normal working behaviour.
Launching BOIC. BOIC is reporting about just two of GTX 260 cards & both are not reguesting any WU, just sitting in the sockets. Only rooling back to old BOIC 6.5.0 in combination with 185.25 drivers return system to normal behaviour ( two GPU's are working). No chance to push third GTX 260 to work.
Haven't any idea, how to screw 6.6.36 with 190.38 cuda 2,3 drivers into the system.
Having serious need for you support \ advices, any direct recommendations \ ideas, what could caused this behaviour ? What should I change in system ?

Much appreciated,

Jet

First one finished and reported, and the new app seems to be slightly slower than previously.

If you look at the full task list for the host, you can see it's got seven GIANNI_BIND tasks currently showing. The latest cuda23 task is the only one showing more than 50,000 seconds elapsed: five others were in the 49,000 range, and one in the 48,000 range. So not a big difference, but (with a sample point of one!) a move in the wrong direction.

I believe you are correct, I set to stock speeds and it is running good now.

Mark Henderson, your case seem to be due to overclocking.
It's not the error FFT or PME related.

gdf

Dear GDF,

My current client is 6.5.0, with previous driver version 185.25.
I know., how to install drivers. Does your question means, that you will recommend to make complete drivers uninstall & make a "clera" install again ?
Basically I've two key targets:
1. To install 3-rd GTX 260 & make it running with GPUGRID.
2. To make it works ( both 3 GTX 260) under newest versions of BOINC \ CUDA drivers.

Any suggestions are highly appreciated !
Thanks in advance,

Jet

http://www.gpugrid.net/result.php?resultid=1137304

Bymark,
which result did error out for you?

gdf

my wu completed last night...

FYI Driver 190.62 released on Aug 21.

ETA, thanks for advices. You are totally right, running multiple GPU's is a real challenge, in my case it just make it even more interesting.
Anyhow, I would continue attempts to push all GPU's running together for GPUGRID.
Basic action plan (:-) is exactly, as you described, to start it again, practically from scratch, consequentially, clean install etc.
Thanks for the ideas,

Regards,

Jet

testing it as we speak, and which crackhead had them change the number of task in queue????? i liked it at 2.

We have now selectively assigning a cuda2.3 application to people with a 190 driver only. This is a test. We keep it on if it does not give problems.
It should be transparent to older driver users.

You need a recent BOINC version to receive the new application.
Only for Windows so far.

gdf

We have now selectively assigning a cuda2.3 application to people with a 190 driver only. This is a test. We keep it on if it does not give problems.
It should be transparent to older driver users.

You need a recent BOINC version to receive the new application.
Only for Windows so far.

gdf

Where do you get the 190 drivers? I just upgraded to BOINC 6.6.36 as requested, so where do I find the new driver?

Windows Vista 64 SP2 with a 9800 GT

nvidia.com

http://www.gpugrid.net/result.php?resultid=1145066

wu with 190.62 complete.

So, there is short story:
1. Using Zydor's advice, all nVidia drivers was uninstalled ( with Driver Sweeper).
2. Old version BOINC (6.5.0) was replaced by new one 6.10.0.
3. New driver (198.38 CUDA 2,3) was installed, according to the Zydor's methologie (under safe mode). Due to the fact, that nVidia's web site is down over 12 hours, was unable to get more recent driver version (198.62, for example.
4. A cc_config.xtml was added to use all gpu's.
5. First attempt with total new software brings practically zero result, I was unable to use 3 GPU's, BOIC was showing just two of them. Device manager, nVtune & GPU-Z still shows 3 GPU's. Kind'a tricky, yeh.
6. Decided to make some relocation of GPU's on the MoBo (ASUS WS Evolution, having 4 PCI-E slots). After movement of GTX 260 to another slot (closer to other GTX), strated BOINC again & suddenly found, that all GTX 260 is in the system & recognized by BOINC. Very happy. That means, that PCI-E IRQ (or else) distribution managed by MoBo is quite important for systems with multiple GPU's.
7.So, get a couple of WU's, crunching with new system less than a day. Almostcompleted 6 WU's up to now, looks, that most running fine.
Having a very strong feeling, that new system ( driver, BOIC & 3 x GTX 260) is a little slower at the moment. New drivers ? More "heavy" contemporary WU's ? I don't know exctly.

So, the first battle result is acceptable. Let's look into the future.

Thanks for the advices & support. Special thanks to Zydor!

Jet

Hi,

where did you find boinc 6.10.0? The highest version, that is offered, is 6.6.36 at http://boinc.berkeley.edu

I've also tried to install drivers 190.62, but WU's fail after 10 to 15%.

Vista64, E8400, 6GB, GTX260 and a lot of troubles.

Regards,

Alexander

Have a look in http://boinc.berkeley.edu/dl/ for Boinc manager 6.10.

Another two broken Wus...

http://www.gpugrid.net/result.php?resultid=1153163
http://www.gpugrid.net/result.php?resultid=1153166
____________
From Siberia with love!

Sorry I'm reverting to the 18x drivers as I'm having too many problems with the 190.x dropping into power save mode and staying there, getting such poor performance is killling any gameplay.

The problem is under 18x drivers I get no WUs with the error about needing 190.38 or better for CUDA 2.3

Surely I have cuda 2.3

And I have always the same errors like other drivers sup than 182.5 ---->

14/08/2009 21:59:47 GPUGRID Output file 52-KASHIF_HIVPR_sub_so_ba1-15-100-RND7514_0_1 for task 52-KASHIF_HIVPR_sub_so_ba1-15-100-RND7514_0 absent
14/08/2009 21:59:47 GPUGRID Output file 52-KASHIF_HIVPR_sub_so_ba1-15-100-RND7514_0_2 for task 52-KASHIF_HIVPR_sub_so_ba1-15-100-RND7514_0 absent
14/08/2009 21:59:47 GPUGRID Output file 52-KASHIF_HIVPR_sub_so_ba1-15-100-RND7514_0_3 for task 52-KASHIF_HIVPR_sub_so_ba1-15-100-RND7514_0 absent

You need at least 185, the error message is misleading.
g

The 186.08 (sorry not .06 I think I must need new glasses) are working fine now. No sign of the downclocking or crashed WUs the 190s were giving.

One guy (svol) from team "Russia" said that he's OK with 190.xx driver and GTX260. I'll try to get a couple Wus for testing...
____________
From Siberia with love!

GDF,

I have a machine, 20013, that refused to run on anything in the 185.xx and higher series. It has an factory OC eVGA 260-216. I backed it off to 182.50 and so far, it appears to be working. I did this yesterday and will report how it runs for longer term. I originally had this on driver 182.06 and it ran fine there. I only moved it when it didn't work with 182.06 anymore. This machine went through the 2.2 upgrade with the old driver, during the period where you guys were still sending out work it could crunch.

I'm happy the Grid went back to allowing 182.xx machines back into the game. I was thinking I would have to drop off the Grid with this one...

I'm also having a side issue of the newer (185.xx and above) drivers not working through the KVM on my farm. They come up with a blank screen. Driver 182.50 works fine.

The farm is running a mix of 185.85, 190.38, and 190.62. I have a workaround which involves plugging the monitor directly into the machine during boot, then putting it back through the KVM when it's up. This is a PITA, but it works.

Let's hope that nVidia gets it together soon...

Regards,
Bob

One guy (svol) from team "Russia" said that he's OK with 190.xx driver and GTX260. I'll try to get a couple Wus for testing...

So, another one broken WU...

28.08.2009 4:42:29 GPUGRID Computation for task p335000-IBUCH_r_pYEEI_us5xy1_2408-3-5-RND3015_0 finished
28.08.2009 4:42:31 GPUGRID Output file p335000-IBUCH_r_pYEEI_us5xy1_2408-3-5-RND3015_0_1 for task p335000-IBUCH_r_pYEEI_us5xy1_2408-3-5-RND3015_0 absent
28.08.2009 4:42:31 GPUGRID Output file p335000-IBUCH_r_pYEEI_us5xy1_2408-3-5-RND3015_0_2 for task p335000-IBUCH_r_pYEEI_us5xy1_2408-3-5-RND3015_0 absent
28.08.2009 4:42:31 GPUGRID Output file p335000-IBUCH_r_pYEEI_us5xy1_2408-3-5-RND3015_0_3 for task p335000-IBUCH_r_pYEEI_us5xy1_2408-3-5-RND3015_0 absent
28.08.2009 4:42:34 GPUGRID Started upload of p335000-IBUCH_r_pYEEI_us5xy1_2408-3-5-RND3015_0_0
28.08.2009 4:42:44 GPUGRID Finished upload of p335000-IBUCH_r_pYEEI_us5xy1_2408-3-5-RND3015_0_0


But why it exited after 5+hours of computing?

I'll try now newest BM and drivers, if it applicable to Win 7 x64...
____________
From Siberia with love!

Well, I have another error. now without gaming (and USING desktop at all!) while computing, without overclocking, with 85% of MAX GPU cooler speed, with 190.62 video driver (BM 6.6.37). As usually, Win 7 Ultimate x64.

Now I've got another WU, 'cause BM 6.10.1 is installed. Let's see, what will happen...

I'll make it work even if it will require a very old driver...
____________
From Siberia with love!

Now you should be able to run again drivers 182.xx.
However, please confirm that 182.xx does work.
Also the error message should be now more appropriate.

gdf

No, using BM 6.10.1 x64 did not help. Win 7 Ult., 190.62.

Old drivers are being loaded...
____________
From Siberia with love!

An old driver is installed:

01.09.2009 8:51:12 NVIDIA GPU 0: GeForce GTX 260 (driver version 18581, CUDA version 2020, compute capability 1.3, 896MB, est. 96GFLOPS)
01.09.2009 8:51:12 No CAL Runtime Libraries installed.

____________
From Siberia with love!