All cuda tasks end up with CreateProcess () failed.

Example Task:

http://setiathome.berkeley.edu/result.php?resultid=1187239039

CPU: Intel Q9300
GPU: 9400 GT

Installed the newest nVidia driver.
The last complete package from Raistmer
and the last x64 Update V10b.

Create an cc_config.xml with ncpus = 5.

Anyone any idea what's wrong?

---

All cuda tasks end up with CreateProcess () failed.

Example Task:

http://setiathome.berkeley.edu/result.php?resultid=1187239039

CPU: Intel Q9300
GPU: 9400 GT

Installed the newest nVidia driver.
The last complete package from Raistmer
and the last x64 Update V10b.

Create an cc_config.xml with ncpus = 5.

Anyone any idea what's wrong?

The OS on this host is:

Microsoft Windows XP
Professional x86 Editon, Service Pack 2, (05.01.2600.00)

and the update is:

x64 Update V10b

Try the x86 Update, not the x64.

Claggy

All cuda tasks end up with CreateProcess () failed.

Example Task:

http://setiathome.berkeley.edu/result.php?resultid=1187239039

CPU: Intel Q9300
GPU: 9400 GT

Installed the newest nVidia driver.
The last complete package from Raistmer
and the last x64 Update V10b.

Create an cc_config.xml with ncpus = 5.

Anyone any idea what's wrong?

The OS on this host is:

Microsoft Windows XP
Professional x86 Editon, Service Pack 2, (05.01.2600.00)

and the update is:

x64 Update V10b

Try the x86 Update, not the x64.

Claggy

You're completly right.
I was too stupid.
It's running :-)

Thanks

---

This update ONLY for those who use BOINC only for SETI MB crunching. No other projects, not even AstroPulse, ONLY SETI Enhanced (MultiBeam).

So since I'm running other projects besides Seti, I cannot use this optimized app package at all?

---

You will be assimilated...bunghole!

This update ONLY for those who use BOINC only for SETI MB crunching. No other projects, not even AstroPulse, ONLY SETI Enhanced (MultiBeam).

So since I'm running other projects besides Seti, I cannot use this optimized app package at all?

That means you don't need this particular update.

I got some validation error here:
1187078058
1186604655
1184920466
1184799046
1182550790
1181304343

It started in CPU then moved to GPU. What would be the cause? Why did it move?

---

This update ONLY for those who use BOINC only for SETI MB crunching. No other projects, not even AstroPulse, ONLY SETI Enhanced (MultiBeam).

So since I'm running other projects besides Seti, I cannot use this optimized app package at all?

That means you don't need this particular update.

Which package should I download then if I want to run CPU MB + CUDA at the same time? Thanks!

Edit - Oh and is it possible to download the packages without having to register for the forums? Thanks again!

---

You will be assimilated...bunghole!

I got some validation error here:
1187078058
1186604655
1184920466
1184799046
1182550790
1181304343

It started in CPU then moved to GPU. What would be the cause? Why did it move?

It moved because idle GPU was detected (as stderr stated in result).
About the cause of validation error - more likely it's some hardware problems.
It can be some software (more specifically, precision-related rounding problem. CPU and GPU give slightly but different results in computations with floating point single precision) problem too, but many results (CPU/GPU combos too) already pass validation.
So, your first step could be to remove (or at least to lower a little) GPU/CPU OCing and see if these invalid results will vanish.
If it happens even w/o OC at all, try latest V10b update that will forbid CPU->GPU re-scheduling (if you run anything but SETI MB it could reduce host performance though).
If invalid results will occur even after these steps - post again.

This update ONLY for those who use BOINC only for SETI MB crunching. No other projects, not even AstroPulse, ONLY SETI Enhanced (MultiBeam).

So since I'm running other projects besides Seti, I cannot use this optimized app package at all?

That means you don't need this particular update.

Which package should I download then if I want to run CPU MB + CUDA at the same time? Thanks!

Edit - Oh and is it possible to download the packages without having to register for the forums? Thanks again!

V10a.
No, for now I don't provide any alternative hosting for V10a packs.
Blame ugly Kaspersky AV false detections.

If it happens even w/o OC at all, try latest V10b update that will forbid CPU->GPU re-scheduling (if you run anything but SETI MB it could reduce host performance though).
If invalid results will occur even after these steps - post again.

What's the indication that i'm using 10b? IIRC I have updated my rigs to 10b.

---

If it happens even w/o OC at all, try latest V10b update that will forbid CPU->GPU re-scheduling (if you run anything but SETI MB it could reduce host performance though).
If invalid results will occur even after these steps - post again.

What's the indication that i'm using 10b? IIRC I have updated my rigs to 10b.

No special indication provided. Just no CPU->GPU re-scheduling.

Thanks Raistmer for your help. I have optimized MB, Cuda, and AP apps on all three of my systems. As mentioned in the documentation, I do occasionally have issues where my GPU is idle, but I am playing around with resource share and hopefully I'll get it to where I don't need to babysit it anymore. :) As I understand it, the upcoming 6.6.x version of BOINC will take care of this on it's own, so either way it'll be fixed. Thanks again!

---

You will be assimilated...bunghole!

Here V10/11 CUDA MB update (old filename used to simplify updating).
Just replace your current CUDA MB executable with this one.
It has some performance gain due to decreased CPU consumption during startup.

ADDON: Also, it should do more smooth %done increase indication.

I use BOINC V6.4.7 and Raistmer's V7 mod.

I gone to my rig and saw only one CUDA running.
The other task was frozen at ~ 00:02:05 CPU time. Remaining time ~ 00:20:00.
Normally the tasks are finshed after ~ 00:02:00 CPU time and the remaining time peak max. ~ 00:12:00.
I can't remember at which %.
I made a reboot and two CUDA starts.

I guess the task was finished but couldn't finish well.


Hmm..

So where was the prob?


AFAIK, BOINC V6.4.5 don't rise the priority of the CUDA WUs.
V6.4.7 do this.

So maybe the BOINC V6.4.7 and Raistmer V7 combi are not well?



Somebody made also experiences with similar CUDA probs?

---

V7 was deprecated long ago.
Current pack is V10. Alternatively you can try BOINC 6.6.17 w/o teamed pack.

AFAIK, only with your V7 mod the CUDA apps are free over all CPU-Core.
Or not?

I used your V9 mod also and in this mod the CUDA apps were fixed at CPU-Core #0.

In V10 the CUDA apps are also fixed at CPU-Core #0?


I crunch only on GPUs. :-)
So I have more performance if one of the GPUs not idle.. ;-)


I used BOINC V6.6.5 and .11 and this two versions have a BUG with work fetch.
Because this versions thought I have all the time two idle CPU-Core and asked all the time for new work.

This BUG is solved in V6.6.17?

---

AFAIK, only with your V7 mod the CUDA apps are free over all CPU-Core.
Or not?

I used your V9 mod also and in this mod the CUDA apps were fixed at CPU-Core #0.

In V10 the CUDA apps are also fixed at CPU-Core #0?


I crunch only on GPUs. :-)
So I have more performance if one of the GPUs not idle.. ;-)


I used BOINC V6.6.5 and .11 and this two versions have a BUG with work fetch.
Because this versions thought I have all the time two idle CPU-Core and asked all the time for new work.

This BUG is solved in V6.6.17?

The work-fetch in 6.6.17 is much improved. I have no idle cores - 4 CPU + 2 GPU.

F.

---

The work-fetch in 6.6.17 is much improved. I have no idle cores - 4 CPU + 2 GPU.

F.

I allow only GPU crunching with app_info.xml .
Only CUDA entry.

In the 'Computing preferences' I insert 4, 2 and finally 0 CPU.
But all the time BOINC thought I have two idle CPU-Core.
[V6.6.11]

So with V6.4.7 it's well with only GPU crunching.


You have all the time 4 x MB and 2 x CUDA?
[This will happen if all stock?
Also with opt. apps?]

I used up to V6.6.11 and sometimes I had no CUDA running but 5 or 6 MBs on CPU.
[And with V6.6.14 very bad CUDA performance with Raistmer's V7 mod]

---

You have all the time 4 x MB and 2 x CUDA?
[This will happen if all stock?
Also with opt. apps?]

No, my CPU cores run mainly optimised AP. Have you noticed that there is now an option on your website preferences that allows you to select NOT to use the CPU? Surely that would achieve what you are aiming for?

F.

---

You have all the time 4 x MB and 2 x CUDA?
[This will happen if all stock?
Also with opt. apps?]

No, my CPU cores run mainly optimised AP. Have you noticed that there is now an option on your website preferences that allows you to select NOT to use the CPU? Surely that would achieve what you are aiming for?

F.

OPS.. Thanks! ..something new in 'SETI@home preferences'.

Why they didn't published this? ;-)


O.K., this will solve 'my' problem.. ;-)


But, if I insert 'Use Central Processing Unit (CPU)' -> NO, the GPU-CPU-support will not stop?
[I don't think so.. but.. sometimes there are BUGs.. ;-)]

---