Well I've woken my fastest PC up after a few months break, and yesterday downloaded BOINC 6.13.10 and the cpu usage was 0.05 CPU per GPU.
This morning it is 0.2 !
Strange ?
dunx
Out of interest why would you put 6.13.10 on it?
Its very much a work in progress version of BOINC. In fact its pretty much been rewritten. Oh and its up to 6.13.12 at the moment. There have been a few fixes since .10 came out.
Because I can't download software two days into the future !
dunx
Your first received Wu was at 12th Nov 2011 12:32:06 UTC, 6.13.11 was tagged for release at 10:07 10th Nov 11 (Pacific time i believe), and 6.13.12 was tagged for release at 09:39 11th Nov 11,
they don't always appear immediately in the download location, normally very soon after.
Because I can't download software two days into the future !
dunx
I think MarkJ's comment wasn't so much to do with running 6.13.10 instead of 6.13.11 or 6.13.12 but rather why run 6.13.X instead of a more stable (recommended) version? If you want to see an example of the grief that comes from the bleeding edge, take a look at this thread and particularly read Richard Haselgrove's explanation of the problem. If your intention is to help debug BOINC, then fine, but if you're trying to bring a host back to a more stable crunching environment, then you're using the wrong version. I'm pretty sure that was MarkJ's point.
If you want to use a 'less risky' test version, try to avoid those where new versions are appearing almost every day. If a test version has been around for a week or two, it's probably less likely to contain show stoppers. Also, with test versions, you are supposed to upgrade immediately there is a new release. The BOINC Devs aren't too interested in bug reports from earlier point releases.
I think Gary has summed it up better than I said it.
As 6.13 was a rewrite its probably one of the more error-prone versions. The 6.12 versions are more of an incremental change so tend to be more stable.
My other point was you may have grabbed the 6.13 simply because its the highest version there, without realising its problems.
Anyway its a major pain to go back to a 6.12 version now (the client_state file isn't backwards compatible without some editing), so you might as well stay with it unless its causing major issues for you.
I do post the BOINC change history on Seti and GPUgrid so people there can see whats going on with it, but refrain from posting on Einstein as I don't think people here would be interested.
It might be noted that the public test of OpenCL applications - primarily related to ATI GPUs - which Bernd announced yesterday is also testing the new OpenCL detection features in v6.13.12, and the new BOINC server code behind them.
My other point was you may have grabbed the 6.13 simply because its the highest version there, without realising its problems.
Anyway its a major pain to go back to a 6.12 version now (the client_state file isn't backwards compatible without some editing), so you might as well stay with it unless its causing major issues for you.
I do post the BOINC change history on Seti and GPUgrid so people there can see whats going on with it, but refrain from posting on Einstein as I don't think people here would be interested.
It might be noted that the public test of OpenCL applications - primarily related to ATI GPUs - which Bernd announced yesterday is also testing the new OpenCL detection features in v6.13.12, and the new BOINC server code behind them.
As far as I know, CPU usage always was around 10 to 25%, depending
on your CPU, brand,type, features, capabillity, also how many GPUs , i.e. a
desktop or work-station, a laptop, although differences quickly
are dissapearing, atleast getting lessser, always have to use a balance between
power and (accu) time. This
host, look at some CPU+GPU tasks.
And this
host, where 0.20CPU and 1 NVIDIA GPU is running those tasks for quite a while now.
I want to add, the GPU application has defenatly improved, I wonder what difference or advantage has a FERMI class GPU, over a 8000/9000/200 series of GPUs? Apart from running more then 1 instance per device.
By the way, very pleased the search has discovered quite a lot of new pulsars!
The version of BOINC has little or no effect on the Project Typical Applications, the science app' s, which are, ofcoarse, two different
things. BOINC is a Manger, handling communications, with all projects involved, detecting hardware, how to handle and compute a WU, anonymous app' s involved, among other things.
Well, I just had to nreinstall WIN 7, 64BIT, 2 Seagate harddrives gave out.
S.M.A.R.T. ALREADY was bad, but a copy and reverse to BOINC 6.10.60
from 6.12.34, ended in a mess.
And I aplologize to my winmen/women, all of them had to time out, since I
also was hospitalized for a week. :-/ Also had to recover and still are.
Now that the I7-2600 and 2 EAH5870 and running again and have some 1500GByte
of extra storage, BOINC didn't have enough diskspace, cause a piece was already bad and the only ' room' was for a Swap Page.
(And CPDN had a staggering 12 GBytes(?) space and 520 hour WUs).
And installed SAMSUNG, 1 x 500GB and 2 1TB SATA 2 drives, for a change ;-)
I would like to try an OpenCL app. on the ATI (AMD) GPUs, I'll read the forums
and there is already a (beta) test?!
Do later BOINC version make use of AVX, f.i.? Or the science app.?
As far as I know only the Seti optimisers have done a version to use it. I don't have a CPU that supports it so haven't checked other projects apart from the ones I normally run. From what I gather it wasn't a simple matter of recompile the app to use it, they had to make some code changes to get a speed advantage.
Given its only the newest CPU's that support AVX (latest i7's and AMD bulldozer) then it will take some time before apps start taking advantage.
BOINC 6.13.10 X64 CPU + GPU usage
)
A CPU usage of 20% for the host and GPU like yours is quite normal for the current batch of GPU tasks at Einstein@Home. Nothing to be concerned about.
HBE
RE: Well I've woken my
)
Out of interest why would you put 6.13.10 on it?
Its very much a work in progress version of BOINC. In fact its pretty much been rewritten. Oh and its up to 6.13.12 at the moment. There have been a few fixes since .10 came out.
BOINC blog
Erm.... Because I can't
)
Erm....
Because I can't download software two days into the future !
dunx
RE: Erm.... Because I
)
Your first received Wu was at 12th Nov 2011 12:32:06 UTC, 6.13.11 was tagged for release at 10:07 10th Nov 11 (Pacific time i believe), and 6.13.12 was tagged for release at 09:39 11th Nov 11,
they don't always appear immediately in the download location, normally very soon after.
http://boinc.berkeley.edu/trac/timeline
Claggy
RE: Erm.... Because I
)
I think MarkJ's comment wasn't so much to do with running 6.13.10 instead of 6.13.11 or 6.13.12 but rather why run 6.13.X instead of a more stable (recommended) version? If you want to see an example of the grief that comes from the bleeding edge, take a look at this thread and particularly read Richard Haselgrove's explanation of the problem. If your intention is to help debug BOINC, then fine, but if you're trying to bring a host back to a more stable crunching environment, then you're using the wrong version. I'm pretty sure that was MarkJ's point.
If you want to use a 'less risky' test version, try to avoid those where new versions are appearing almost every day. If a test version has been around for a week or two, it's probably less likely to contain show stoppers. Also, with test versions, you are supposed to upgrade immediately there is a new release. The BOINC Devs aren't too interested in bug reports from earlier point releases.
Cheers,
Gary.
I think Gary has summed it up
)
I think Gary has summed it up better than I said it.
As 6.13 was a rewrite its probably one of the more error-prone versions. The 6.12 versions are more of an incremental change so tend to be more stable.
My other point was you may have grabbed the 6.13 simply because its the highest version there, without realising its problems.
Anyway its a major pain to go back to a 6.12 version now (the client_state file isn't backwards compatible without some editing), so you might as well stay with it unless its causing major issues for you.
I do post the BOINC change history on Seti and GPUgrid so people there can see whats going on with it, but refrain from posting on Einstein as I don't think people here would be interested.
BOINC blog
It might be noted that the
)
It might be noted that the public test of OpenCL applications - primarily related to ATI GPUs - which Bernd announced yesterday is also testing the new OpenCL detection features in v6.13.12, and the new BOINC server code behind them.
(message 114911)
Yup ! signed up for Albert as
)
Yup ! signed up for Albert as well !
dunx
P.S. Been away for a few months and just jumped back in with the most obvious download... ( Don't mind a few problems along the way ! )
RE: RE: RE: My other
)
As far as I know, CPU usage always was around 10 to 25%, depending
on your CPU, brand,type, features, capabillity, also how many GPUs , i.e. a
desktop or work-station, a laptop, although differences quickly
are dissapearing, atleast getting lessser, always have to use a balance between
power and (accu) time.
This
host, look at some CPU+GPU tasks.
And this
host, where 0.20CPU and 1 NVIDIA GPU is running those tasks for quite a while now.
I want to add, the GPU application has defenatly improved, I wonder what difference or advantage has a FERMI class GPU, over a 8000/9000/200 series of GPUs? Apart from running more then 1 instance per device.
By the way, very pleased the search has discovered quite a lot of new pulsars!
The version of BOINC has little or no effect on the Project Typical Applications, the science app' s, which are, ofcoarse, two different
things. BOINC is a Manger, handling communications, with all projects involved, detecting hardware, how to handle and compute a WU, anonymous app' s involved, among other things.
Well, I just had to nreinstall WIN 7, 64BIT, 2 Seagate harddrives gave out.
S.M.A.R.T. ALREADY was bad, but a copy and reverse to BOINC 6.10.60
from 6.12.34, ended in a mess.
And I aplologize to my winmen/women, all of them had to time out, since I
also was hospitalized for a week. :-/ Also had to recover and still are.
Now that the I7-2600 and 2 EAH5870 and running again and have some 1500GByte
of extra storage, BOINC didn't have enough diskspace, cause a piece was already bad and the only ' room' was for a Swap Page.
(And CPDN had a staggering 12 GBytes(?) space and 520 hour WUs).
And installed SAMSUNG, 1 x 500GB and 2 1TB SATA 2 drives, for a change ;-)
I would like to try an OpenCL app. on the ATI (AMD) GPUs, I'll read the forums
and there is already a (beta) test?!
Do later BOINC version make use of AVX, f.i.? Or the science app.?
Re: AVX instructions As
)
Re: AVX instructions
As far as I know only the Seti optimisers have done a version to use it. I don't have a CPU that supports it so haven't checked other projects apart from the ones I normally run. From what I gather it wasn't a simple matter of recompile the app to use it, they had to make some code changes to get a speed advantage.
Given its only the newest CPU's that support AVX (latest i7's and AMD bulldozer) then it will take some time before apps start taking advantage.
BOINC blog