Advanced search

Message boards : News : For Fermi-CUDA3.0 users.

Author Message
Profile GDF
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist
Send message
Joined: 14 Mar 07
Posts: 1957
Credit: 629,356
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 19231 - Posted: 4 Nov 2010 | 10:37:39 UTC

If you have a GTX400 series card and are using our cuda30 application, please not that we have now deprecated it and only cuda31 applications will be distributed.

To keep receiving workunits you need to update to CUDA3.1 drivers (256. or newer).

thanks,
gdf

Profile Lanfear
Send message
Joined: 2 Nov 10
Posts: 3
Credit: 466,932
RAC: 0
Level

Scientific publications
watwat
Message 19253 - Posted: 4 Nov 2010 | 19:31:27 UTC
Last modified: 4 Nov 2010 | 19:47:31 UTC

Hi!
Im new on Boinc and GPUgrid!
I have a very new GTS 450(Fermi clas),with ForceWare driver 260.99(latest)
I have start a GPUgrid calculation,but it say that "Fermi class GPU not supported by Cuda 2.2",I am on Windows XP
How can i do calculation with my GPU?
Im French,so exscuse my for my bad english^^
Thanks!

EDIT:GPU-Z say that only CUDA 2.1 is supported!

Romain Cavallini

Aman
Send message
Joined: 24 Sep 10
Posts: 1
Credit: 18,770,850
RAC: 0
Level
Pro
Scientific publications
watwatwatwatwatwatwatwatwatwatwat
Message 19282 - Posted: 5 Nov 2010 | 23:15:10 UTC - in response to Message 19231.

Hi,

I Have GeForce GTX 460 Card and the Latest Driver 260.99 still Message says "Fermi-Class GPU, Not supported by CUDA 2.2"

Please help and let me know what corrective measures I should be taking.

Thanks

Aman Dutt

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 19285 - Posted: 6 Nov 2010 | 0:21:01 UTC - in response to Message 19253.

Lanfear, you are presently running this task, p24-IBUCH_26_PQpYEEIPI_101019-10-40-RND3589.

Aman, I expect you will pick up a task in a while.

Geras
Send message
Joined: 17 Nov 10
Posts: 1
Credit: 0
RAC: 0
Level

Scientific publications
wat
Message 19630 - Posted: 22 Nov 2010 | 20:40:47 UTC - in response to Message 19285.

Hello,

I've got several tasks from GPUGRID (GTX460, nVidia 260.99). First tasks started to run and hold almost to a finish.
Example:
Name p14-IBUCH_3_opt1_pYEEI_101027-18-20-RND3520_0
Workunit 2084590
# There is 1 device supporting CUDA
# Device 0: "GeForce GTX 460"
# Clock rate: 0.81 GHz
# Total amount of global memory: 1041694720 bytes
# Number of multiprocessors: 7
# Number of cores: 56
# Using device 0

Than it started to make errors:
# There are 3 devices supporting CUDA
# Device 0: "�"
# Clock rate: 0.00 GHz
# Total amount of global memory: 4475442 bytes
# Number of multiprocessors: 1221788
# Number of cores: 9774304
# Device 1: "�"
# Clock rate: 0.00 GHz
# Total amount of global memory: 4475442 bytes
# Number of multiprocessors: 1221788
# Number of cores: 9774304
# Device 2: "�"
# Clock rate: 0.00 GHz
# Total amount of global memory: 4475442 bytes
# Number of multiprocessors: 1221788
# Number of cores: 9774304
SWAN: FATAL : No device found

New task did not start to compute even.
The details you can see in my Tasks-View.

What should be the reason and how to correct it?
Thank you.

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 19654 - Posted: 23 Nov 2010 | 21:16:55 UTC - in response to Message 19630.

Some of your specs:
Boinc 6.10.58, GTX460 1GB, nVidia 260.99, AMD Phenom(tm) II X6 1090T, Win7 x86 3GB RAM.

Same error repeatedly:
Client state Compute error
Exit status -40 (0xffffffffffffffd8)

From your reports it looks like the driver loaded but then lost its configuration at some stage during one of the runs. To speculate, this might happen if the system rolled back to the last known good configuration, restarts abruptly (power) or crashes, resulting in a hard drive corruption. A memory problem would clear up on a restart, or the system would just fail. The app/Boinc now seems to think there are 3 GPUs but does not recognise them. As it did install and work for a while, I would be inclined to do a disk check and reinstall the driver. My guess is that you have a corrupt driver but it’s possible Boinc has a corruption, and it could be another problem altogether. Did the system crash or freeze any time recently? You might also want to check that the card has not physically moved. If you suspect it has (when putting the video cable in for example), remove the card, restart, shutdown and reinsert the card.

Note, this guesswork is just an attempt to point you in the direction to help yourself, without all the facts that’s all I can offer. Run away failures can mean you stop picking up tasks for 24h.

Some usual culprits to also check are the PSU, system resource usages (CPU, RAM, HDD), temperature, fan speeds. GPUZ might help with the GPU.

Profile Fred J. Verster
Send message
Joined: 1 Apr 09
Posts: 58
Credit: 35,833,978
RAC: 0
Level
Val
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 19726 - Posted: 30 Nov 2010 | 23:59:01 UTC - in response to Message 19654.


Hi,

I Have GeForce GTX 460 Card and the Latest Driver 260.99 still Message says "Fermi-Class GPU, Not supported by CUDA 2.2"

Please help and let me know what corrective measures I should be taking.

Thanks

Aman Dutt.



Upgrading to CUDA 3.1 libraries, these cards have Compute Capabillity 2.0/2.1
Why not, since it clearly doesn't work OK...
The GTX580 is next best?


____________

Knight Who Says Ni N!

Profile robertmiles
Send message
Joined: 16 Apr 09
Posts: 503
Credit: 727,920,933
RAC: 155,858
Level
Lys
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 19737 - Posted: 3 Dec 2010 | 2:20:55 UTC

I've seen a message somewhere that the current workunits and the 260.99 driver don't work together well.

Therfore, I'm using the 258.96 driver with my 9800 GT to give it CUDA 3.1. However, I'm currently not getting any GPUGRID workunits. Do the current workunits have problems with that combination also?

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 19747 - Posted: 3 Dec 2010 | 18:15:55 UTC - in response to Message 19737.

You are getting lots of errors running the 6.13app. If you uninstall your present driver and install (reboot to safe mode) the 197.45 driver you will pick use the 6.12app to run tasks. The 6.12app may be more stable than the 6.13app for CC1.1 cards, and is faster.

Werkstatt
Send message
Joined: 23 May 09
Posts: 121
Credit: 321,525,386
RAC: 177,358
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 19782 - Posted: 6 Dec 2010 | 9:52:29 UTC

Hi,
I see here some posts referring to problems with drivers or setup.
I needed to update to 260.99; MW has released the first openCL app and I wanted to test that.
I have currently running:
win7-64, 6GB, BM 6.10.58, 260.99, I use an app_info forcing 6.13 (cuda31) apps.
I'm currently crunching:
249-GIANNI_DHFR-1000...
151-KASHIF_HIVPR_n1...
I encountered configuration problems at the beginning; the system has an ATI 4850 which was not detected by BOINC. The solution atm is: Monitor attached to the ATI-Card, nVidia runs without monitor. ATI crunches MW (cal).
It needed several reboots, reinstalling the drivers and two GPUGRID wu's failed immediate. But keep on experimenting, there is a way to get it running!

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 19785 - Posted: 6 Dec 2010 | 10:56:16 UTC - in response to Message 19782.

What are you using in your app_info file?

I see you are getting Anonymous platform reports for Application.

The error said energies have become nan. I think this just means that the molecules energies have went flat, zeroed out, so this might just have been a task issue and nothing to do with your setup.

Some of the more recent NVidia drivers allow for multiple GPUs use without a cable or KVM attached, including the 260.99 driver (recommended for the GTX460). I guess your ATI driver can't do this and has to be plugged in.

Werkstatt
Send message
Joined: 23 May 09
Posts: 121
Credit: 321,525,386
RAC: 177,358
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 19786 - Posted: 6 Dec 2010 | 11:02:44 UTC - in response to Message 19785.

What are you using in your app_info file?

I see you are getting Anonymous platform reports for Application.

The error said energies have become nan. I think this just means that the molecules energies have went flat, zeroed out, so this might just have been a task issue and nothing to do with your setup.

Some of the more recent NVidia drivers allow for multiple GPUs use without a cable or KVM attached, including the 260.99 driver (recommended for the GTX460). I guess your ATI driver can't do this and has to be plugged in.


pls check your PM

Post to thread

Message boards : News : For Fermi-CUDA3.0 users.

//