1)
Message boards :
Problems and Help :
Colab & GNF-17 not working
(Message 161569)
Posted 56 days ago by waffleironhead
definitely a driver issue. its not seeing the T4 as an opencl device.
First and foremost, I am not a Linux user, so there could be a better way to do this.
I Added: !apt install nvidia-driver440
and things seem to be smooth sailing.
import os.path
!apt-get update
!apt install nvidia-driver-440
!apt-get install boinc boinc-client
!cp /usr/bin/boinc /content
!cp /usr/bin/boinccmd /content
if not os.path.exists('/content/slots'):
!mkdir slots
if not os.path.exists('/content/slots/0'):
!boinc --attach_project "http://www.primegrid.com/" "YOURUSERID"
else:
!boinc
|
2)
Message boards :
Problems and Help :
cpu and gpu tasks on short queue
(Message 161470)
Posted 63 days ago by waffleironhead
the easiest way to do this is via the boinc manager.
Head into options-->computing preferences
Under section other:
Store at least 0 days work
store up to an additional 0 days work
this will mostly limit to 1 cpu task per core if multithreading isnt set up, and one gpu task at a time
|
3)
Message boards :
Number crunching :
Geek Pride Day Challenge
(Message 155727)
Posted 371 days ago by waffleironhead
Is 12% high for aborted Tasks?
Doesn't really matter. Aborted tasks are probably from people who accidentally downloaded too many tasks and aborted them so they have a better chance of returning a result 1st and getting credit for finding a prime instead of being the DC'er.
Doesn't change anything in the challenge.
Aborted tasks are better than abandoned tasks. IE: Tasks that are removed by means other than the BOINC client whether it's deleting them from the host directly or issuing a project reset on the BOINC client.
Either way, it doesn't effect the scoring at all here.
Only negative effect is the person who downloads the resends will almost definitely be a DC'er.
Well my review is they are members that know what they are doing and doing before each Challenge.
You do not download 900 tasks and then aborted them and do the same the next day.
Most aborted tasks are Second run tasks and will not be 1st that is why I think know long time members do this.
I just do not care to complete tasks that have been aborted a day ago knowing that the tasks was completed already.
"Doesn't change anything in the challenge." But it is not all about the challenge but about finding the Prime.....
But I also see your Point.
Boinc is terrible at judging how many tasks to download with regards to multithreaded apps. Add into the fact that it also is terrible at judging task completion times for gpu and cpu tasks from the same project, and you end up with people like me who request 1 day worth of work and end up with 2 weeks worth.
It's nothing malicious.
|
4)
Message boards :
Number crunching :
LLR2 installed on all big LLR projects
(Message 146460)
Posted 908 days ago by waffleironhead
Just curious.
What kind of fault tolerance is built into the verification double check system, say the double check computer is overclocked and faulty. How does the server check for that?
|
5)
Message boards :
Number crunching :
Sophie Germain's Birthday Challenge
(Message 139284)
Posted 1160 days ago by waffleironhead
I'm crunching 4 * 4 on my 16 core machines.
Try 16 tasks one thread. MT is not a good idea for SGS.
I disagree, on my 2400g system 4 core/8 thread, my best results are from doing 2 units at a time using 4 threads each.
|
6)
Message boards :
Number crunching :
Sophie Germain's Birthday Challenge
(Message 139244)
Posted 1160 days ago by waffleironhead
I HAVE COMPLETED 4 TASKS BUT WHY I'M NOT ON TNE STATICS??????????
You will not show in the stats till your wingman has been completed and your task verified.
Edit: guess that is not the case. Misremembered. Carry on
|
7)
Message boards :
Number crunching :
Google Colab and its free GPUs
(Message 139064)
Posted 1172 days ago by waffleironhead
I have been using colab. figured I would post up some numbers based on the different gpus I get. I have been able to get a single instance 12 hours a day or so and then a break of 12 hours. non gpu instances are available during that time, but I get timeouts faster, so don't bother.
ppssieve.
p100 216 seconds
t4 206 seconds
p4 388 seconds
ap27
p100 1063 seconds
t4 1079 seconds
p4 1781 seconds
gfn15
p100 36 seconds
t4 27 seconds
p4 58 seconds
k80 56 seconds
|
8)
Message boards :
Number crunching :
syst.host
(Message 100850)
Posted 2396 days ago by waffleironhead
It looks like you are running windows 7. There is a known issue with windows update eating up a lot of processor time.
You might want to try the things suggested in this thread at anandtech forums and see if it helps.
https://forums.anandtech.com/threads/fix-for-windows-7-update.2471653/
|
9)
Message boards :
AP26 - AP27 Search :
App testing
(Message 96771)
Posted 2520 days ago by waffleironhead
Just tested the opencl64 app on my HD6950 1GB card.
All three tests completed successfully and took around 4 minutes each.
Windows 7 Running driver 15.7.1
So older VLIW4 card works even with only 1GB memory.
|
10)
Message boards :
Problems and Help :
Sierpinski Problem ESP/PSP/SoB (Sieve) Validate Error WU
(Message 94834)
Posted 2586 days ago by waffleironhead
You are welcome. Glad you got it sorted.
|
