Message boards :
Number crunching :
48 WU limit
Message board moderation
Author | Message |
---|---|
Send message Joined: 29 Oct 10 Posts: 4 Credit: 78,568,389 RAC: 0 |
I understand that one is limited to 6 WU's in queue per CPU core. I am running an i7 2600k, and dual ATI 6950's. Therefore I am limited to 48 WU's in queue at any time. I process a typical WU every 42 seconds, so I only have about 30 minutes of work in queue. I also understand that having too many units in queue can lead to them not being completed in time, but being limited to a half hour of work in queue means being down whenever the server hiccups. Is there any way to increase the number of work units in queue? Other projects let you have 2-10 days of work in queue. |
Send message Joined: 12 Nov 07 Posts: 2425 Credit: 524,164 RAC: 0 |
Is there any way to increase the number of work units in queue? Other projects let you have 2-10 days of work in queue. No. Doesn't expecting the unexpected make the unexpected the expected? If it makes sense, DON'T do it. |
Send message Joined: 29 Oct 10 Posts: 4 Credit: 78,568,389 RAC: 0 |
Is there any chance that this limit will be relaxed in the future? From what I can tell, the 6 wu per core limit was instituted before there was GPU support, and work units took hours instead of minutes to complete? It makes no sense to have a GPU based project to have a queue based on CPU cores. They should base it on the speed of the GPU processor. I should be able to get more work than someone with a 3800, and I don't think it's wrong to expect 12 hours or more of work in the queue. That said, I guess I'll have to find a backup project to work on when MW@H is out of work. What other GPU based projects are optimized for ATI? |
Send message Joined: 8 Jan 09 Posts: 58 Credit: 53,161,741 RAC: 0 |
other AMD/ATI Projects: http://boinc.thesonntags.com/collatz/ - CAL-based Apps - best http://www.primegrid.com/ - OpenCL-based App - no so good (esp. with x2 Cards) |
Send message Joined: 4 Oct 08 Posts: 1734 Credit: 64,228,409 RAC: 0 |
This subject is brought up by new comers to the project about 0nce per months, and this has gone on for more than 2 years. Please read the previous threads as the the several reasons why the 6 WUs per core will not be raised. <sigh> Go away, I was asleep |
Send message Joined: 13 Mar 08 Posts: 804 Credit: 26,380,161 RAC: 0 |
|
Send message Joined: 9 May 10 Posts: 2 Credit: 12,553,698 RAC: 0 |
other AMD/ATI Projects: For me: milkyway 3.13 credits/sec primegrid 2.07 credits/sec collatz 2.18 credits/sec |
Send message Joined: 15 Jul 08 Posts: 383 Credit: 729,293,740 RAC: 0 |
Recently upgraded a couple machines from x4 to x6. One is getting the correct 36 WU cache but the other refuses to DL more than 24: > Milkyway@home 03-29-11 16:29 (reached limit of 24 tasks in progress) Tried updating BOINC, resetting project, detaching and re-attaching, automatic ATI client, optimized ATI client. Still won't get 36 WUs. Any ideas? Here's the client: http://milkyway.cs.rpi.edu/milkyway/show_host_detail.php?hostid=272957 |
Send message Joined: 15 Jul 08 Posts: 383 Credit: 729,293,740 RAC: 0 |
Doi, found out the problem. When I installed the x6, BOINC for some reason decided to change the "Use at most 100% of processors" to 75%. Don't know why. Was running 4 instances of the non-BOINC Wieferich client on the CPUs and that automatically discovered the 6 cores. Anyway, if upgrading to a processor with more cores, check the above setting as BOINC is behaving mysteriously again... |
©2024 Astroinformatics Group