Hi, I've been using BOINC ever since it came out. I just upgraded to the latest version of the software and have been having nothing but problems since.

I'm using Win7 64 Pro which is up to date. Also have anti-virus up to date. I'm attached to Einstein@home, Milkyway@home and SETI@home. I have it set to that SETI gets 70% of the time and the other projects an equal share of the rest

My Pc is pretty old, it's only an Athlon dual core machine with an nVidea 8600 GPU +4Gb RAM

All was OK for a while after upgrading to the latest BOINC - no settings or preferences were changed - but I did notice that far fewer SETI workunits were being downloaded and processed, whilst a lot more Einstein ones were. I put this down to variances in workunit availability. Also, BOINC seems to get estimated processing times for Milkyway@homeunits wildy wrong, it keeps on giving 7000 hours to completion (even though 20% is already done in 2 hours) and this constantly forces BOINC to put them on high priority.

What has just started happening in that the PC is running inordinately slowly. Even simply tasks like web-browsing are almost unusable as scrolling through page sor moving the mouse pointer are very laggy.

Snoozing BOINC instantly solves the problem.

What might be causing this?

I've always allowed BOINC tasks to run at all times and to use the GPU but never had any issues in the past.

Two separate issues.

1) Milkyway are currently sending out some WUs with some very dubious estimates. I have one running at the moment which started with an estimate of over two years. It reached 10% after 100 minutes, so I'm confident that it will finish in less than a day. But BOINC - particularly the new v7.0.64 - seems inordinantly slow to accept the evidence of its own eyes when a unit is over-estimated: the task is still showing 17,913 hours remaining to run, and I suspect will remain in high priority for many hours to come. All you can do is pester the project to be more careful in their workunit preparation (you'll find me there already).

2) I'm guessing, but the screen slowness and laggy mouse pointer might be related to teething problems with the new applications launched at SETI last week. [It's most unlikely to be BOINC itself to be causing this]

Could you help to narrow it down, please, by suspending first one project at a time, then one task at a time, until you find a single task which clears the problem when you suspend it? Then, go to the message boards for the project responsible for the task, and post your experiences there - quote the full names of the Application and (task) Name that seemed to be the culprit. Project volunteers will be better placed to help you on that one.

Can you post a Host ID for it at Seti please.

Claggy

HostID 3080266 belonging to Richard Gregory matches the description, if you want to start the diagnostics before he gets back to us with confirmation.

Confirm that ID.

I fnd that suspending, for example, the Milkyway@home task immediately improves my PC's response times, but also caused BOINC to just download an additional Einstein@home task and begin to run it, bringing everything back down to a slow-motion crawl.

It gives every appearance that BOINC is simply allocating too much CPU resource to tasks running on the CPU. I'm assuming that it is running two concurrently, one on each core.

I tried changing BOINC preferences to "use at most 80% CPU", which made no difference I could perceive.

I am, by the way, aware that I need to update my GPU drivers to be able to use all available SETI tasks. Unfortunately, I'm bandwidth limited and am having some issues with my ISP, so don;t want to download a 200Mb file just now.

Perhaps you should try to set your preferences to "Suspend GPU work while computer is in use? yes" as a compromise between running always (on the CPU) and avoiding performance problems, as I think those stem from using your (rather weak) GPU for computations.

Gruß,
Gundolf

HostID 3080266 belonging to Richard Gregory matches the description, if you want to start the diagnostics before he gets back to us with confirmation.

First thing i'd do is disable Astropulse v6 in the project preferences, complete the OpenCL AstroPulse v6 Wu and see if there is an improvement just running the regular v7 Cuda Wu's,
my experience with running AP on a 9800GTX+ was it was quite laggy and performance poor compared with running v6 Cuda Wu's.

Claggy

Thanks. Having faffed about a lot with various WU's and projects, it looks as though Astropulse is the culprit. Though I never had any issues previously, am I correct in saying that there have been new versions of the SETI software released more or less at the same time as the new BOINC?

Oh, and maybe it's time I went ahead and built myself a new PC with a Haswell CPU and a graphics card that isn;t about a decade out of date LOL

am I correct in saying that there have been new versions of the SETI software released more or less at the same time as the new BOINC?

Yes, the Setiathome v7 applications have just been released, they add an extra search called autocorrelation,

Claggy

Since BOINC ver. 7.0.64 (x86), there has been a new "don't need" response
from the scheduler. I've managed to "suspend" at least one of the
other WU's (both CPDN), to force a download of a desired WU from another project. CPDN should not be adversely affected.

However, am not completely sure what situation the "don't need" response is supposed to address. I prefer to take a more hands-on approach, and allocate
different projects' priorities.

Is BOINC taking a more active mode in allocating different projects?

Since BOINC ver. 7.0.64 (x86), there has been a new "don't need" response
from the scheduler. I've managed to "suspend" at least one of the
other WU's (both CPDN), to force a download of a desired WU from another project. CPDN should not be adversely affected.

However, am not completely sure what situation the "don't need" response is supposed to address. I prefer to take a more hands-on approach, and allocate
different projects' priorities.

Is BOINC taking a more active mode in allocating different projects?

The 'Don't need' message isn't from the scheduler, it's from the client, if your host's cache is full, the client doesn't need any more work, the 'don't need' is just a message from the client saying why it didn't ask for work.

Claggy

OK, Claggy. Will forge ahead with testing, thus armed with your
info.

Thanks,

There are some 14 or so local reasons of why work is not fetched since v7. They were always in the client, just not printed before. Here's an example of me having suspended 1 task [ready to start] and having increased the buffer so that there is a shortfall. Upon hitting 'update' to speed things up, this rather pointy bolded message appears:

2306 World Community Grid 6/17/2013 7:58:49 AM update requested by user
2307 World Community Grid 6/17/2013 7:58:52 AM [sched_op] Starting scheduler request
2308 World Community Grid 6/17/2013 7:58:52 AM Sending scheduler request: Requested by user.
2309 World Community Grid 6/17/2013 7:58:52 AM Not requesting tasks: some task is suspended via Manager
2310 World Community Grid 6/17/2013 7:58:52 AM [sched_op] CPU work request: 0.00 seconds; 0.00 devices
2311 World Community Grid 6/17/2013 7:58:56 AM Scheduler request completed

In the case of 'don't need', translation is "buffer full" :D

---

Coelum Non Animum Mutant, Qui Trans Mare Currunt