Running GTX275 on ubuntu 10.04 x64, running 6.04 app

Everything was just fine, I was not even around of that computer. But I just noticed that there is no 2nd task. I updated manually and here's what i've got:

Tue 17 Aug 2010 06:08:12 PM EDT GPUGRID Sending scheduler request: To fetch work.
Tue 17 Aug 2010 06:08:12 PM EDT GPUGRID Requesting new tasks for CPU
Tue 17 Aug 2010 06:08:22 PM EDT GPUGRID Scheduler request completed: got 0 new tasks
Tue 17 Aug 2010 06:08:22 PM EDT GPUGRID Message from server: No work sent
Tue 17 Aug 2010 06:08:22 PM EDT GPUGRID Message from server: Fermi-class GPU needed

Smth happened to the rig or to the project? Or G200b is not supported any more?

---

What is your cache (additional work buffer) set at?
Try another manual update for GPUGrid.

The servers try to send Fermi Work Units first, and then non-Fermi work. You don’t always see these messages but I have seen this many times on several types of card and have been able to pick up new tasks after a short time. My GTX260-sp216 GT200b is running one task with one in the queue.

Our CC1.3 cards will be able to crunch here for a long time yet.

Hi CTAPbli

Looks perfectly normal to me.

The messages you refer to are typical BOINCish and simply translate into "no, there are no CPU tasks for GPUGRID".

Also, your RAC numbers suggest that you are running other GPU projects in parallel. Therefore, when enough other work is available BOINC will choose to not immediately download a second GPUGRID WU.

Nothing to worry about.

Happy crunching

Alain

CTAPbli, you have 2 tasks In Progress now; One running and one Queued.

smth weird happened this nite... in my understanding there was no non-Fermi WUs that is why I got that messages. now it looks OK, but i'll keep an eye on this.

thanks a lot guys for the help :-)

---

If it would be able to set "Preferences for this project" to
Use Central Processing Unit (CPU) "NO"
client would not ask for CPU work.
It is an website issue, when set to as mentioned, NO changes to YES after updating preferences.

I have a GTX295 v2(single pcb card), Everytime lately that I've tried to use It on gpugrid, I get the notice Fermi-class gpu needed and I was told to ignore It, Well I ignored It as the project seems to ignore Me, I run Boinc 6.10.56 x64(64 bit Boinc) and the video driver is 257.15 which works for Seti, Newer drivers cause trouble and the current Seti app for My 295 needs this driver, Yet the gpugrid WU's just error out, So My only conclusion is that XP x64 and Boinc 64bit is not supported here, So until My platform is supported properly and not given lip service, Count Me out. No I don't have money for even an MSI Cyclone GTX460 OC card and I won't for a good 3-6 months, As I live on a fixed income(SSI) as I'm a disabled person.

---

XP x64 and Boinc 64bit ARE supported here.

I dont know what you are basing your opinions on but you have not run any tasks recently. It is normal that the servers first try to send Fermi tasks, and normal that the server later sends non-Fermi tasks. You need to allow the system time to deliver.

XP x64 and Boinc 64bit ARE supported here.

I dont know what you are basing your opinions on but you have not run any tasks recently. It is normal that the servers first try to send Fermi tasks, and normal that the server later sends non-Fermi tasks. You need to allow the system time to deliver.

I did try and run the app a few months back on My GTX295 and I use the 257.15 Nvidia driver, Boinc 6.10.56 and most gpugrid WU's would just error out with maybe one or two WU's completing properly. I did allow time for the server to send the proper app, My problem did happen even when I had the right app, I wished It didn't happen of course, No I didn't notice any error numbers, I just wanted to let someone know this happened, If My current project is not able to send Me any work for the EVGA Co-OP GTX295 v2 card(single pcb), I'll reattach and try again so that You'll have some data.

---

XP x64 and Boinc 64bit ARE supported here.

I dont know what you are basing your opinions on but you have not run any tasks recently. It is normal that the servers first try to send Fermi tasks, and normal that the server later sends non-Fermi tasks. You need to allow the system time to deliver.

Ok You wanted something, So here It is:

pc1 GPUGRID 6.05 ACEMD2: GPU molecular dynamics (cuda) S493-TONI_HERGMETAXDOFE-17-50-RND2348_1 00:00:05 (00:00:04) 10/5/2010 7:16:18 PM 10/5/2010 7:17:23 PM 0.32C + 1NV (d0) Reported: Computation error (-40,)
pc1 GPUGRID 6.05 ACEMD2: GPU molecular dynamics (cuda) aet33-IBUCH_GCY_100914-18-160-RND1531_1 00:00:17 (00:00:12) 10/5/2010 7:15:48 PM 10/5/2010 7:17:23 PM 0.32C + 1NV (d0) Reported: Computation error (98,)
pc1 GPUGRID 6.05 ACEMD2: GPU molecular dynamics (cuda) h232f99r465g99-TONI_CAPBINDsp3-42-100-RND9535_1 00:00:07 (00:00:05) 10/5/2010 7:11:48 PM 10/5/2010 7:12:48 PM 0.32C + 1NV (d0) Reported: Computation error (-40,)

The files above were run today, Now You can see the errors, As can I and everyone else.

---

OK, so you are getting many failures, all within a few seconds, but you run one task to completion:

3091903 1959426 6 Oct 2010 2:16:57 UTC 6 Oct 2010 8:42:04 UTC Completed and validated 22,340.48 970.52 6,322.41 9,483.62 ACEMD2: GPU molecular dynamics v6.05 (cuda)

453-KASHIF_HIVPR_n1_unbound_so_ba1-33-100-RND0677_0

Tasks that fail almost immediately are annoying but they do not have a big impact on your contribution, so long as you keep getting some tasks to run and complete. Hopefully more are running and will complete for you.

If possible I would suggest you use the latest Boinc version and the latest NVidia driver. They may not be any better, but these updates do remove many bugs.

VW ... In the past when I have experienced the behavior you are seeing I had not turned off SLI (nvidia control panel disable multi gpu). Also, if you are overclocking, I have found that I can run higher shader frequencies on SETI than I can on GPUGrid.

---

Thanks - Steve

VW ... In the past when I have experienced the behavior you are seeing I had not turned off SLI (nvidia control panel disable multi gpu). Also, if you are overclocking, I have found that I can run higher shader frequencies on SETI than I can on GPUGrid.

Sli is off and I'm not overclocking.

---

I need to keep my GPU's under 75° to avoid failures. Just closing the windows brought it above 75° today & I got a tonne of failures. Try to check your temps & bring them down as much as possible.

---

I need to keep my GPU's under 75° to avoid failures. Just closing the windows brought it above 75° today & I got a tonne of failures. Try to check your temps & bring them down as much as possible.

My temps are usually much better than 75C, It's the one on the right, Next to the Green Square, The 4 temps to the left are for the cpu core temps.

---

Have you tried to suspend S@H4cuda while you run GPUGRID & see if that helps?

---

Have you tried to suspend S@H4cuda while you run GPUGRID & see if that helps?

Huh? Seti@Home is only running on the cpu while gpugrid is running as is normal, Look for the word Running in the supplied image...

---

No need to get mad, I'm just trying to suggest what it could be. BTW, it only says waiting to run & I don't know if you've set it to switch between tasks or not. Sometimes I get a failed WU simply by stopping & restarting it, don't know if you've noticed that too. Especially when using multiple GPU's even if they are the same type, when a WU running on GPU(1) restarts on GPU(0).

---

No need to get mad, I'm just trying to suggest what it could be. BTW, it only says waiting to run & I don't know if you've set it to switch between tasks or not. Sometimes I get a failed WU simply by stopping & restarting it, don't know if you've noticed that too. Especially when using multiple GPU's even if they are the same type, when a WU running on GPU(1) restarts on GPU(0).

I never get Mad, As Mad can mean Crazy, I think You and almost everyone I've ever heard of uses the wrong term, It should be Angry, And No I'm not angry.

As to Waiting to Run, Those tasks are waiting on the hdd and not in memory as I do not have anything set to wait in Memory. They will run next unless some other WU with a closer deadline shows up.

No I haven't set Boinc to switch between tasks. As I think very highly of 6.10.56 which runs just fine as does 257.15, As newer doesn't always equal better, I tried 6.10.58 and drivers newer than 257.15 and found I didn't like how they worked, So I'm staying where I am, Most improvements on the video driver are for Games that I have no interest in, But then I'm not a gamer really.

---

VW: up for more *suggestions*?

Still guessing but ...
For the moment, while S@H is down, would you be willing to try disconnecting from S@H and then do a reset on GPUGrid?
I had problems recently after running S@H and switching back to GPUGrid until I reset. My guess is there may be some incompatibility in the one of the core libraries each project downloads. Then again I had problems when switching hardware and a modular pcie cable came loose from the PSU (but not completely disconected) so what do I know :-)

p.s. I should rack up well over 100k when S@H gets back up and makes points available to stats :thumbsup:

---

Thanks - Steve