Each computer can only achieve 100 tasks at a time?

Message boards : Number crunching : Each computer can only achieve 100 tasks at a time?
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile freestman
Volunteer tester
Avatar

Send message
Joined: 14 Oct 01
Posts: 14
Credit: 18,618,633
RAC: 37,744
China
Message 1930765 - Posted: 18 Apr 2018, 23:22:03 UTC

Is there a way to get more tasks?
ID: 1930765 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 5292
Credit: 352,952,800
RAC: 1,095,638
United States
Message 1930770 - Posted: 19 Apr 2018, 0:04:40 UTC - in response to Message 1930765.  

No, not really. The 100 tasks per device is a server-side limit.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 1930770 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 10189
Credit: 136,217,017
RAC: 88,092
Australia
Message 1930802 - Posted: 19 Apr 2018, 4:50:17 UTC - in response to Message 1930765.  

Is there a way to get more tasks?

Install suitable video cards.

Systems are limited to 100 WU per device- at this time the devices are CPU & GPU. The way it works out is that limits is 100WU for CPUs (whether you have 1 CPU or 100 CPUs, the limit is 100WUs for that computer).
When it comes to GPUs, the limit is per device. So if you have a computer & it has 1 GPU, it can get up to 100WUs. If it has 5 GPUs, it will get up to 500WUs (plus up to 100 WUs for the CPU(s) if it is/they are allowed to do processing).
Grant
Darwin NT
ID: 1930802 · Report as offensive
Profile Wiggo "Socialist"
Avatar

Send message
Joined: 24 Jan 00
Posts: 14816
Credit: 191,887,800
RAC: 69,291
Australia
Message 1930832 - Posted: 19 Apr 2018, 7:54:33 UTC

Or either dual boot or run 2 instances of BOINC.

Cheers.
ID: 1930832 · Report as offensive
Profile freestman
Volunteer tester
Avatar

Send message
Joined: 14 Oct 01
Posts: 14
Credit: 18,618,633
RAC: 37,744
China
Message 1930867 - Posted: 19 Apr 2018, 10:11:09 UTC

Thanks!
ID: 1930867 · Report as offensive
Profile David@home
Volunteer tester
Avatar

Send message
Joined: 16 Jan 03
Posts: 740
Credit: 4,151,759
RAC: 48
United Kingdom
Message 1931218 - Posted: 21 Apr 2018, 9:27:04 UTC
Last modified: 21 Apr 2018, 9:27:24 UTC

Does BOINC segregate the GPUs? If you have a small and a large Nvidia card will the large Nvidia crunch most of the 200 allocated work units?
ID: 1931218 · Report as offensive
Profile Mike Special Project $75 donor
Volunteer tester
Avatar

Send message
Joined: 17 Feb 01
Posts: 31277
Credit: 65,529,304
RAC: 26,321
Germany
Message 1931220 - Posted: 21 Apr 2018, 9:42:16 UTC - in response to Message 1931218.  

Does BOINC segregate the GPUs? If you have a small and a large Nvidia card will the large Nvidia crunch most of the 200 allocated work units?


Each device is supposed to get 100 tasks .
With each crime and every kindness we birth our future.
ID: 1931220 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 10189
Credit: 136,217,017
RAC: 88,092
Australia
Message 1931221 - Posted: 21 Apr 2018, 10:02:03 UTC - in response to Message 1931218.  
Last modified: 21 Apr 2018, 10:40:18 UTC

If you have a small and a large Nvidia card will the large Nvidia crunch most of the 200 allocated work units?

Yep.
As Mike said, you get 100WUs for each GPU that is available to crunch, but any of the GPUs can crunch the WUs that are allocated for GPU processing- they are not allocated to a specific GPU.

Edit- That is the case if all your GPUs are the same type (eg AMD), although if you have mixed Nvidia/iGPU/AMD GPUs then I think it's a case of WUs allocated to particular types of GPU have to be processed by that type, but if you have multiple GPUs of a particular type then any of them can process WUs allocated to that GPU type.
Anyone able to confirm/refute?
Grant
Darwin NT
ID: 1931221 · Report as offensive
Profile Zalster Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 27 May 99
Posts: 4650
Credit: 292,898,314
RAC: 482,575
United States
Message 1931276 - Posted: 21 Apr 2018, 21:42:33 UTC - in response to Message 1931221.  

If you have a small and a large Nvidia card will the large Nvidia crunch most of the 200 allocated work units?

Yep.
As Mike said, you get 100WUs for each GPU that is available to crunch, but any of the GPUs can crunch the WUs that are allocated for GPU processing- they are not allocated to a specific GPU.

Edit- That is the case if all your GPUs are the same type (eg AMD), although if you have mixed Nvidia/iGPU/AMD GPUs then I think it's a case of WUs allocated to particular types of GPU have to be processed by that type, but if you have multiple GPUs of a particular type then any of them can process WUs allocated to that GPU type.
Anyone able to confirm/refute?


100 per GPU and 100 for total CPU (despite how many cores) 2 GPUs and 1 CPU will get you 300 total work units. Now, he's question regarding big vs small. I think he's talking high end vs low end GPUs. The high end will do the work much quicker therefore will do more per hour than the low end. So it would look like the high end is hogging all the work but in reality it's just processing faster than the low end card.
ID: 1931276 · Report as offensive
Profile freestman
Volunteer tester
Avatar

Send message
Joined: 14 Oct 01
Posts: 14
Credit: 18,618,633
RAC: 37,744
China
Message 1931329 - Posted: 22 Apr 2018, 3:32:37 UTC

It would be more reasonable if we could distribute tasks according to the number of threads in CPU.
ID: 1931329 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 10189
Credit: 136,217,017
RAC: 88,092
Australia
Message 1931335 - Posted: 22 Apr 2018, 5:20:48 UTC - in response to Message 1931329.  

It would be more reasonable if we could distribute tasks according to the number of threads in CPU.

GPUs can process way more work than CPUs, even those with lots of threads. It would be nice if we could carry more work, but the database just isn't up to it.
Grant
Darwin NT
ID: 1931335 · Report as offensive
Profile Mike Special Project $75 donor
Volunteer tester
Avatar

Send message
Joined: 17 Feb 01
Posts: 31277
Credit: 65,529,304
RAC: 26,321
Germany
Message 1931345 - Posted: 22 Apr 2018, 8:24:46 UTC - in response to Message 1931335.  

It would be more reasonable if we could distribute tasks according to the number of threads in CPU.

GPUs can process way more work than CPUs, even those with lots of threads. It would be nice if we could carry more work, but the database just isn't up to it.


My Ryzen is always the first empty device during long outages.
With each crime and every kindness we birth our future.
ID: 1931345 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 10189
Credit: 136,217,017
RAC: 88,092
Australia
Message 1931349 - Posted: 22 Apr 2018, 9:12:36 UTC - in response to Message 1931345.  

It would be more reasonable if we could distribute tasks according to the number of threads in CPU.

GPUs can process way more work than CPUs, even those with lots of threads. It would be nice if we could carry more work, but the database just isn't up to it.


My Ryzen is always the first empty device during long outages.

But if you had a video card of similar performance level, it would run out of work first.

High end CPU, High end GPU, the GPU will run out of work first.
Low end CPU, low end GPU, the GPU will run out of work first.
High end CPU, mid range GPU, would be pretty even
High end CPU, lower-mid/low end GPU, CPU will run out of work first.
And of course the exception of all exceptions- a low end octa-socket multi-thread CPU system will run out of work before a high end GPU.

The vast majority of systems will run out of GPU work before they run out of CPU work. Of course for every rule, there are exceptions.
Grant
Darwin NT
ID: 1931349 · Report as offensive

Message boards : Number crunching : Each computer can only achieve 100 tasks at a time?


 
©2018 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.