Message boards :
Number crunching :
SETI@Home not getting CPU work; only running on GPU
Message board moderation
Author | Message |
---|---|
CalicoSkies Send message Joined: 20 May 99 Posts: 31 Credit: 1,352,098 RAC: 4 |
Yesterday I installed BOINC again on my 2 PCs so that I could get back into helping with SETI@Home (and other projects). I'm using the latest BOINC, 7.2.39, which I installed on both of my PCs. I've noticed that SETI@Home seems to only run on a GPU; it's not assigning tasks to run on the PC's CPU. My main PC has an Intel i7-3770K and an Nvidia GeForce GTX 560, and so far I've only seen SETI@Home tasks running on the GPU (none running on the CPU). My 2nd PC has an Intel i5-2500 using its built-in graphics (no discrete GPU), and that machine is not running any SETI@Home tasks at all. I have my option set to use Intel GPUs, so I'd think that machine should at least be running SETI@Home on the Intel GPU. I don't know why SETI@Home isn't running on the CPU. My main PC's CPU has 8 threads (4 cores with hyperthreading) and only one GPU - So I'd think there would be some benefit to running SETI@Home on the CPU as well as the GPU. Is there something I'm missing? |
ausymark Send message Joined: 9 Aug 99 Posts: 95 Credit: 10,175,128 RAC: 0 |
You may have to wait for the respective CPU work units to download too. Do your Seti settings on the website say to use CPU work units? |
Jord Send message Joined: 9 Jun 99 Posts: 15184 Credit: 4,362,181 RAC: 3 |
Your i5 is a Sandy Bridge type CPU, whose GPU does not support OpenCL. You need an Ivy Bridge type CPU, like your i7 at minimum to have that enabled. To be able to use the GPU built into the i7 you require Intel graphics drivers, and BOINC needs to have detected the GPU before it can be used. This may need a monitor attached, adummy VGA plug, or an extended desktop over the Intel GPU. |
CalicoSkies Send message Joined: 20 May 99 Posts: 31 Credit: 1,352,098 RAC: 4 |
Your i5 is a Sandy Bridge type CPU, whose GPU does not support OpenCL. You need an Ivy Bridge type CPU, like your i7 at minimum to have that enabled. That's fine, but it should at least get CPU work units. The i5 is not getting any CPU work units at all. And how did you know my i5 is a Sandy Bridge? I did not mention that in my first post. |
CalicoSkies Send message Joined: 20 May 99 Posts: 31 Credit: 1,352,098 RAC: 4 |
You may have to wait for the respective CPU work units to download too. Yes, I have the "Use CPU" option enabled in my SETI@Home account settings. Could it be that the SETI@Home project simply does not have any work units for CPUs right now? I'm not sure how that would make sense though.. I did see a post from late January (the 30th or 31st) saying the Arecibo telescope is closed due to an earthquake - Might that have something to do with no CPU work units being available? |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
Your i5 is a Sandy Bridge type CPU, whose GPU does not support OpenCL. You need an Ivy Bridge type CPU, like your i7 at minimum to have that enabled. By default BOINC allows your computers to be displayed via your profile Computers belonging to Eric Oulashin. This is often useful when troubleshooting. On your i5 What kind of response are you getting from the server when there is a work request. If you open Advanced > Event Log you should see something like this: 16-Feb-2014 19:37:12 [SETI@home] Requesting new tasks for CPU 16-Feb-2014 19:37:16 [SETI@home] Scheduler request completed: got 1 new tasks SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
You may have to wait for the respective CPU work units to download too. There is no such thing as CPU work or GPU work. Just work. It get assigned to a GPU or CPU when the client requests it. You can check the status of the project http://setiathome.berkeley.edu/sah_status.html to see if there is work on the server, "Results ready to send". Right now there are a few hundred thousand ready to go. When the client requests work it pulls from the feeder. Which only seems to hold so many at a time. So when things are very busy you may see a lot of "project is work of work" messages even tho there are lots ready to send out. The feeder has to refill to complete the client requests. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Batter Up Send message Joined: 5 May 99 Posts: 1946 Credit: 24,860,347 RAC: 0 |
I did see a post from late January (the 30th or 31st) saying the Arecibo telescope is closed due to an earthquake - Might that have something to do with no CPU work units being available? That post had an eye catching shock title, made me look, but totally false about "no new data for a while". |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
Your hosts do have CPU work from other projects, it is possible you don't need CPU work from Seti because your cache is already full, did Boinc even ask for CPU work from Seti?, what does the Event log say? Claggy |
CalicoSkies Send message Joined: 20 May 99 Posts: 31 Credit: 1,352,098 RAC: 4 |
On your i5 What kind of response are you getting from the server when there is a work request. If you open Advanced > Event Log you should see something like this: On my i5 machine, it's reporting no tasks are available: 2/16/2014 9:51:48 PM | SETI@home | Sending scheduler request: To fetch work. 2/16/2014 9:51:48 PM | SETI@home | Requesting new tasks for CPU 2/16/2014 9:51:50 PM | SETI@home | Scheduler request completed: got 0 new tasks 2/16/2014 9:51:50 PM | SETI@home | No tasks sent 2/16/2014 9:51:50 PM | SETI@home | No tasks are available for SETI@home Enhanced 2/16/2014 9:51:50 PM | SETI@home | No tasks are available for SETI@home v7 2/16/2014 9:51:50 PM | SETI@home | No tasks are available for AstroPulse v6 For comparison, these are some log entries from my i7 machine: 2/16/2014 3:06:29 PM | SETI@home | Sending scheduler request: To fetch work. 2/16/2014 3:06:29 PM | SETI@home | Reporting 1 completed tasks 2/16/2014 3:06:29 PM | SETI@home | Requesting new tasks for NVIDIA 2/16/2014 3:06:31 PM | SETI@home | Scheduler request completed: got 19 new tasks So this brings to mind a few of questions: - Why is my i7 machine requesting tasks only for NVIDIA? It can only run one task on NVIDIA at a time; it should be able to use my CPU cores too. - Why are no new tasks being given to my i5? - Why is it asking for tasks specifically for NVIDIA or specifically for CPU? As you've said, there's no such thing as CPU work or GPU work; just work. |
Jord Send message Joined: 9 Jun 99 Posts: 15184 Credit: 4,362,181 RAC: 3 |
And how did you know my i5 is a Sandy Bridge? I did not mention that in my first post. Because I have an i5-2500K myself and keep up with its and its brethren their capabilities. See Specifications of Intel HD Graphics series for all the HD Graphics/other Intel built-in graphics models that can and cannot do OpenCL. Mind, this is only the graphics, as long as you have an APP driver installed, all of the Intel Core CPUs can do OpenCL perfectly well. But an OpenCL application for on the CPU isn't available here (or at any other project, AFAIK). |
Jord Send message Joined: 9 Jun 99 Posts: 15184 Credit: 4,362,181 RAC: 3 |
On my i5 machine, it's reporting no tasks are available So, when you go to https://setiathome.berkeley.edu/prefs.php?subset=project, which venue (---, home, school, work) are you using, and what are your settings for: Use CPU Use ATI GPU Use NVIDIA GPU Use Intel GPU Default computer location Run only the selected applications If no work for selected applications is available, accept work from other applications? |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
On your i5 What kind of response are you getting from the server when there is a work request. If you open Advanced > Event Log you should see something like this: As Claggy said your machines have CPU work from other projects. It looks like the CPU on your i7 and i5 are busy running Rosetta@home at the moment. The BOINC client tries to fill up with as much as as work as it has been told to retain, based on each projects resource share setting, to keep all of your resources busy. The BOINC client work on a debt system. So it may run one project for a given amount of time before even asking for work from other project. Because it has enough work to not need work from any other project to keep it busy. So your i7 probably has enough work to not request any CPU work from SETI@home right now. Your i5 might just be unlucky in its work requests to SETI@home. Looking over my history I do see several times where my machines asked for work but didn't get any. Which would be an empty feeder when the client request was made. You could set the other projects on that machine to No New Tasks to see if it will pickup SETI@home tasks. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
CalicoSkies Send message Joined: 20 May 99 Posts: 31 Credit: 1,352,098 RAC: 4 |
The BOINC client work on a debt system. So it may run one project for a given amount of time before even asking for work from other project. Because it has enough work to not need work from any other project to keep it busy. I see. But I thought the resource share settings in BOINC were supposed to allow you to adjust the relative priority of the various projects you're running. Right now I have SETI@Home configured to get the highest resource share of all the projects I'm running, yet it seems that most of the work my machines are getting is from rosetta@home right now. |
rob smith Send message Joined: 7 Mar 03 Posts: 22200 Credit: 416,307,556 RAC: 380 |
BOINC will try and balacne out the work done over a LONG period, not a matter of hours or even days. I've found (and others may well have found differently) you have to have a big difference between you first project and other to get a sensible balance. So, if you want SETI to be be your number one project set it to 1000, set your second project to 100, your third to 10. Setting 100, 90, 80 for example will not have the desired effect unless you are prepared to wait a very long time... Another thing to remember is that you can always set a "reserve" project, with a share of zero - this will only run when other projects have no work available. Bob Smith Member of Seti PIPPS (Pluto is a Planet Protest Society) Somewhere in the (un)known Universe? |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
The BOINC client work on a debt system. So it may run one project for a given amount of time before even asking for work from other project. Because it has enough work to not need work from any other project to keep it busy. It never works exactly as we users think that it should. Say Rosetta were configured for 50% & SETI 50% on a 4 core system. Instead of always running 2 Rosetta & 2 SETI tasks all of the time it will run one project or the other for an overall equal amount of time. Using an app_config.xml BOINC can somewhat be made to work in that fashion by limiting the numbers of instances of a project application. So only 2 of each would be allowed to run at once. However if one project were out of work the other would still be limited to 2 instances. On my machines that run SETI@Hom Beta I limit them to only 1 instance. So that BOINC does not fill the whole queue with Beta work. There is a special condition setting where you can assign a resource share of 0 to other projects you prefer not to run unless your main one is out of work. Then the BOINC client will only request work from the other project(s) when your main project(s) are not providing work. It will only request enough work to keep your computer resources busy instead of filling your work queue. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
CalicoSkies Send message Joined: 20 May 99 Posts: 31 Credit: 1,352,098 RAC: 4 |
This past Saturday I noticed several SETI@home work units that reported compute errors. Looking at these tasks on the SETI@home web site, it looks like these were all running AstroPulse v6 v6.01 (rather than AstroPulse v6 v6.04 (opencl_nvidia_100), which the other SETI@home tasks are using). I don't know if it's possible to confirm whether the compute errors happened on my CPU or GPU, but if they were on my CPU, I wonder if that's why I'm not getting any CPU-bound SETI@home tasks. I'd hate to think my CPU has a problem, but I haven't had a problem with it doing anything else, and my system has always been stable since I built it. These are the links to my recent tasks with compute errors: http://setiathome.berkeley.edu/result.php?resultid=3390293164 http://setiathome.berkeley.edu/result.php?resultid=3390293136 http://setiathome.berkeley.edu/result.php?resultid=3390293115 http://setiathome.berkeley.edu/result.php?resultid=3390293039 http://setiathome.berkeley.edu/result.php?resultid=3390292707 |
betreger Send message Joined: 29 Jun 99 Posts: 11361 Credit: 29,581,041 RAC: 66 |
Since the error tasks do not say AstroPulse v6 v6.04 (opencl_nvidia_100)as a sucessful GPU task from your computer says it seems the failed tasks were CPU. |
CalicoSkies Send message Joined: 20 May 99 Posts: 31 Credit: 1,352,098 RAC: 4 |
Since the error tasks do not sayAstroPulse v6 v6.04 (opencl_nvidia_100)as a sucessful GPU task from your computer says it seems the failed tasks were CPU. But the debug output has the following line: In ap_gfx_main.cpp: in ap_graphics_init(): Starting client. That makes it sound like it's trying to initialize the graphics component (GPU). I just thought of something.. The motherboard in my i7 machine does not have a video output port, so I am not able to use the Intel graphics - Could there be a bug in the SETI@Home software that would cause it to fail on a CPU where the system is not using internal graphics? |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
Since the error tasks do not sayAstroPulse v6 v6.04 (opencl_nvidia_100)as a sucessful GPU task from your computer says it seems the failed tasks were CPU. No, the "ap_graphics_init()" refers to the screensaver graphics included with the CPU worker application. If the Intel graphics component was available as an OpenCL 'Intel GPU' (not every i7 is capable of this, it has to be 3rd generation 'Ivy Bridge' or 4th generation 'Haswell'), a completely separate Intel GPU application would be downloaded to make use of it. It wouldn't interfere with the standard CPU application at all. |
©2024 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.