| Author |
Message |
|---|
|
|
|
Hi all,
is for this project work for GPU? i see that cuda version is in aplications? is working?
JHAPA
____________
 |
|
|
|
|
|
It works! Not quite as fast as PPS, but still nice.
____________
My lucky numbers are 121*2^4553899-1 and 3756801695685*2^666669±1
My movie https://vimeo.com/manage/videos/502242 |
|
|
|
|
|
Hi,
It sorta seems to work. But I'm seeing computation errors. See this task and this task.
Both seem to have failed for "Maximum elapsed time exceeded" at 1:43 elapsed. Could someone please bump up the limit?
These were run on a GeForce 9400M on a MacBook Pro. Boinc manager 6.12.11.
Thanks,
Tony
|
|
|
|
|
Hi,
It sorta seems to work. But I'm seeing computation errors. See this task and this task.
Both seem to have failed for "Maximum elapsed time exceeded" at 1:43 elapsed. Could someone please bump up the limit?
These were run on a GeForce 9400M on a MacBook Pro. Boinc manager 6.12.11.
Thanks,
Tony
Tony,
The workunit values have been corrected. Unfortunately there were a number of workunits created before the problem was noticed. These will continue to be resent until they find a machine that can run them. It's a pain, but they will eventually all get done.
|
|
|
henningVolunteer tester
Send message
Joined: 29 Nov 08
Posts: 4
ID: 32332
Credit: 991,286,037
RAC: 2,125
                
|
|
All my CW sieve cuda workunits have failed so far on my Linux 64 bit system with a NVIDIA GTX260 using the 260.19.21 drivers. see for example this task.
The PPS cuda sieve btw runs great on my system.
|
|
|
 rroonnaalldd Volunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
            
|
Those nVIDIA 260.19.* drivers don't quite work right. Try 256.53. It's faster, too!
I found a posting that nvidia driver v256.X with latest SDK gives the same error as driver v260.19.X with an earlier 3.X SDK. Please take a look at Microcruncher* posting in thread CWPSieve CUDA testing.
Maybe the SDK 2.3 solve this problem.
____________
Best wishes. Knowledge is power. by jjwhalen
 |
|
|
|
|
The workunit values have been corrected. Unfortunately there were a number of workunits created before the problem was noticed. These will continue to be resent until they find a machine that can run them. It's a pain, but they will eventually all get done.
I'm still seeing these old WUs. ;-(
Tony
____________
  |
|
|
|
|
I'm still seeing these old WUs. ;-(
And I'm still seeing them. ;-(
Since I'm wasting GPU cycles on these, I'm going to pull back from CW for awhile. Folks, please lemme know if and when this gets fixed. Thanks.
____________
|
|
|
|
|
It works! Not quite as fast as PPS, but still nice.
I'm still seeing the Cullen/Woodall GPU app listed under applications;
but not available as a selection under PrimeGrid preferences. If it
were available, I'd run some on our tesla2's. -Bruce*
(If you're recommending app.info, do you know that the one
you have in mind works on C2050's? GeniferCuda definitely doesn't.) |
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
I'm still seeing the Cullen/Woodall GPU app listed under applications;
but not available as a selection under PrimeGrid preferences. If it
were available, I'd run some on our tesla2's. -Bruce*
(If you're recommending app.info, do you know that the one
you have in mind works on C2050's? GeniferCuda definitely doesn't.)
Tried and Cullen/Woodall (sieve) is available. Check Cullen/Woodall and activate "Use NVIDIA GPU".
If you want to crunch with an app_info you can use the thread app_info file as basis...
C2050/2070 or M2050/2070 are Fermi-based DP-cards. GeneferCuda needs DP and should be running. If not, post all relevant info's in the thread Project Staging Area.
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
|
I'm still seeing the Cullen/Woodall GPU app listed under applications;
but not [**listed as**] available as a selection under PrimeGrid preferences.
Tried and Cullen/Woodall (sieve) is available. Check Cullen/Woodall and activate "Use NVIDIA GPU".
If you want to crunch with an app_info you can use the thread app_info file as basis...
Yes, thanks. Looks like selecting both Cullen/Woodall and Proth for
GPU sieving got me 10 Proth tasks and 2 C/W tasks. These two GPU
sieving projects are listed symmetrically under apps; but even with
NVIDIA activated, there's an extended note in "preferences" for
Proth GPU sieving but no mention of GPU sieving for C/W. Perhaps
there'll be an upgraded profile once the Cullen/Woodall GPU sieving
is viewed as stable.
For those of us with ruby badges from PPS Sieving (44.8 million, and
counting), picking up the ruby badge from C/W sieving seems to be a
plausible diversion, pending a subsequent credit comparision. I like
having the GPU sieving available under "plain vanilla" boinc; not having
to worry about correct settings/tuning in a app_info. With c. 200 x86_64
cores running NFS sieving and with 1000 64-bit windows7 ECM jobs
running overnights, both under condor, I have enough to occupy the share
of my attention available for crunching.
Meanwhile, looks like our 32-bit xeons are doing better on LLR's since
I've gotten the condor jobs off --- PPS (239K), Cullen, Woodall and SG.
I'm seeing 1952 credit/task for Cullen LLR and 2559 credit/task for
Woodall LLR (with none of the 32-bit tasks in yet), so the extended runtime
seems plausible, credit-wise.
C2050/2070 or M2050/2070 are Fermi-based DP-cards. GeneferCuda needs DP and should be running. If not, post all relevant info's in the thread Project Staging Area.
Actually, as a math person, I'm not sure I'd be able to explain wanting
to find a generalized Fermat (at least Proths have a small/tiny chance
of giving a Fermat factor); likewise for 123primes. The install script wanted
to open xterms, while I'm using Xming or Xwin32 terms; and oughtn't to
distract the people in PSA from more likely productive inquiries.
Regards, Bruce* |
|
|
|
|
I'm still seeing these old WUs. ;-(
And I'm still seeing them. ;-(
And I'm still seeing them. ;-(
See this task, for example.
____________
|
|
|
|
|
|
Ha-ha!!!
Look at my GCW Stat:
37 / 7400 = precisely 0.005! )))
____________
|
|
|
