(SORRY, double post)

I am having a lot of CPU Quantum Chemistry WUs error out as soon as they start on my Ubuntu machine. Some work but most error out. Is this inherent with the CPU app or is it my machine?

There's a bug in the app, which prevents more than 1 task starting simultaneously. When the task started successfully, you can start another task manually. Since there's no automated way for this, you should pause all but one task before you shut down your computer, then start them one by one. The other option is to limit the concurrently running QC apps to 1. Since this app uses only 4 threads (cores) you should utilize your other CPU cores with a different project.
To do this you should create / modify your app_config.xml file in the projects\www.gpugrid.net folder.

<app_config>
  <app>
    <name>QC</name>
    <max_concurrent>1</max_concurrent>
  </app>
  <app_version>
    <app_name>QC</app_name>
    <plan_class>mt</plan_class>
    <avg_ncpus>4</avg_ncpus>
  </app_version>
</app_config>

So I understand why I receive "error while computing" with almost 0 seconds of compute time (multiple WUs starting at once) but I don't understand the fairly high number of failed WUs with multiple thousand second or even over ten thousand second run times. Linked below are my error rates:

http://www.gpugrid.net/results.php?hostid=424454&offset=0&show_names=0&state=5&appid=

Does anyone think they can explain why the non WUs starting at once errors are occurring?

Did you stop the computer at any point, and let them all RE-start in unison - from whatever point they'd reached?

Did you stop the computer at any point, and let them all RE-start in unison - from whatever point they'd reached?

From my knowledge, the computer only turned off once last week. It is on 24/7 otherwise. It looks like these errors are all over the place in terms of date.

We are losing CPU Volunteers, can anyone help?

We are losing CPU Volunteers, can anyone help?

It's up to the scientists to sort. They will need a Windowa version to have any hope of getting a reasonable throughput.

---

Radio Caroline, the world's most famous offshore pirate radio station.
Great music since April 1964. Support Radio Caroline Team -
Radio Caroline

We are losing CPU Volunteers, can anyone help?

It's up to the scientists to sort. They will need a Windowa version to have any hope of getting a reasonable throughput.

Or just linux app without 'starting many at once' issue. I could add more than 100 cores but I cant because of that.

We will talk this week with the devs of the QM CPU software to see how we can make a Windows build. Once we know the status and how much work it is we will update. But believe me we are extremely keen on a Windows version...

Since you are using the wrapper from LHC@home why don't you use VirtualBox? This would eliminate the need for a Windows version. All Windows user of LHC@home can run LHC@home tasks, written in Scientific Linux, using Virtual Box.
Tullio

changed my desktop settings to include cpu tasks again. (im in the pool so it wont show on my account). hope my contribution, small as it is, helps!

Since you are using the wrapper from LHC@home why don't you use VirtualBox?

No, please.
Now i'm using a Linux vm with Virtual Box for this project and it's a nightmare.

Since you are using the wrapper from LHC@home why don't you use VirtualBox?

No, please.
Now i'm using a Linux vm with Virtual Box for this project and it's a nightmare.

The LHC@home performance using Virtualbox vs direct from linux is a ridiculous loss in efficiency and performance, not to mention substantial ram usage. If we can avoid virtualbox outright I would consider that a win.

The LHC@home performance using Virtualbox vs direct from linux is a ridiculous loss in efficiency and performance, not to mention substantial ram usage. If we can avoid virtualbox outright I would consider that a win.

You keep saying that, which I think just shows that you have had problems with VirtualBox. Even if so, I don't know why that is a reason for GPUGrid not to use it. A lot of people can get it to work.

It is not a "ridiculous loss in efficiency", but about the same as using Windows on some projects optimized for Linux (I do both VirtualBox ATLAS and native ATLAS on LHC, and have a basis for comparison).

And the ram usage depends mainly on the project. It is reasonable enough (about 2 GB) when running 4 cores on Cosmology. If that is too much for you, then upgrade your hardware or stop hassling other people that do have it.

We talked with the devs. The idea is to collaborate over summer for a Windows version. So I hope by the end of summer we should have a new release.

We talked with the devs. The idea is to collaborate over summer for a Windows version. So I hope by the end of summer we should have a new release.

Great news!!
Meantime, please fix problems with linux

All CPU tasks fail on my SuSE Linux Leap 42.3. All GPU tasks complete and validate on my GTX 750 Ti giving me huge credits, more than Einstein@home and SETI@home GPU tasks.
Tullio

All CPU tasks fail on my SuSE Linux Leap 42.3. All GPU tasks complete and validate on my GTX 750 Ti giving me huge credits, more than Einstein@home and SETI@home GPU tasks.
Tullio

Hi, Credits cannot be compared from project to project. And even though GPU tasks give so much credit here at GPUGrid there is still substantial scientific weight and benefit to GPUGrid's CPU Work Units even though they don't give as much credit. Please keep this in mind.

All CPU tasks fail on my SuSE Linux Leap 42.3. All GPU tasks complete and validate on my GTX 750 Ti giving me huge credits, more than Einstein@home and SETI@home GPU tasks.
Tullio

Hi, Credits cannot be compared from project to project. And even though GPU tasks give so much credit here at GPUGrid there is still substantial scientific weight and benefit to GPUGrid's CPU Work Units even though they don't give as much credit. Please keep this in mind.

Since BOINC credits are defined as a certain number of 'cobblestones' (completed floating point operations), they should be comparable. But I agree, they're not.

That somewhat arcane point becomes more significant when statistics sites, and BOINC itself, back-calculate the floating point performance of a project from the number of credits awarded.

Hi, Credits cannot be compared from project to project. And even though GPU tasks give so much credit here at GPUGrid there is still substantial scientific weight and benefit to GPUGrid's CPU Work Units even though they don't give as much credit. Please keep this in mind.

I don't give a damn about credits. But all my CPU tasks fail miserably.
Tullio