Why is this result invalid?

Dirk Sadowski
Dirk Sadowski
Joined: 26 Oct 09
Posts: 24
Credit: 103412693
RAC: 0
Topic 209888

Why is this result invalid?

https://einsteinathome.org/de/workunit/312645337

 

AFAIK, my PCs never got invalid results at Einstein@home.

 

The both wingmen are with Linux OS - because of this?

 

Thanks.

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5872
Credit: 117732158910
RAC: 34962638

Dirk Sadowski wrote:Why is

Dirk Sadowski wrote:

Why is this result invalid?

https://einsteinathome.org/de/workunit/312645337

Because it didn't agree closely enough with its 'quorum partner'.  A 'decider' was sent out and when it was returned, it was found to agree with the other result but not yours.  There is no easy way to know exactly what the discrepancy was.

Dirk Sadowski wrote:
AFAIK, my PCs never got invalid results at Einstein@home.

There's always a first time for everything - most people seem to experience the very slight risk of this happening to otherwise quite stable machines occasionally.

 

Dirk Sadowski wrote:
The both wingmen are with Linux OS - because of this?

Unlikely, I would think.  The Devs take a lot of care to ensure the small differences between all factors such as OS, hardware differences, etc., don't cause problems for the validation process.  Stats for this sort of thing are monitored all the time and action is taken to tweak the validator if there is any evidence for this.  There has been no comment about an issue like this for quite a while now, as far as I'm aware.

 

Cheers,
Gary.

archae86
archae86
Joined: 6 Dec 05
Posts: 3157
Credit: 7227721590
RAC: 1103094

I think that many machines

I think that many machines believed to be flawless get invalid results declared by Einstein on the current Gamma-ray pulsar binary search #1 on GPUs application at rates of the order of a few tenths of a percent.

Many machines that are overclocked enough to harm correctness get appreciably higher invalid rates. 

I don't believe the issues making up the baseline rate are well understood.  I think the baseline rate must be expected to be specific to application, and not general to the project as a whole.

 

Stick
Stick
Joined: 24 Feb 05
Posts: 790
Credit: 33133094
RAC: 1116

Got a new computer earlier

Got a new computer earlier this week and it seems to be having trouble with Binary Radio Pulsar Search tasks.  Tasks are finishing OK but a fair number (over 10%) are getting marked invalid.

Invalid Tasks for Computer 12571639

Just wondering if it's my computer or something else.

EDIT: Just noticed all my invalids are from a Beta app. - v1.34 (opencl-intel_gpu-Beta) windows_x86_64

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5872
Credit: 117732158910
RAC: 34962638

Stick wrote:... Tasks are

Stick wrote:
... Tasks are finishing OK but a fair number (over 10%) are getting marked invalid.

Hi Stick, Good to see you around again :-).

I run Linux on all my machines.  There is no Linux driver for Intel GPUs so I've never tried to crunch with one.  There are lots of complaints over quite a few years about Windows driver problems with earlier iGPUs.  At one stage the standard app was only allowed to get work if a known good driver was present.  As hardware and driver versions changed, this became too hard to keep updated so a beta app without driver restrictions was produced.  It's up to the user to find a version (if one exists) that will run without producing errors.  I don't know what the situation is with more recent hardware, but, from past experience, a 10% error rate sounds pretty good :-).

Stick wrote:
Just wondering if it's my computer or something else.

It's probably the driver.  At one point people were listing 'good' versions and 'bad' versions, but this was certainly for older hardware.  There were several threads around discussing this.  Two (quite lengthy ones) in particular that I recall are here and here.

 

Cheers,
Gary.

Stick
Stick
Joined: 24 Feb 05
Posts: 790
Credit: 33133094
RAC: 1116

Gary, Good to see/hear from

Gary,

Good to see/hear from you again, too.  Einstein has become so much of a set it/forget it project I haven't had or seen the need to be on the boards in a long while.  But your advice is as sound as ever.

My new laptop is still so new to me that I am still learning its idiosyncrasies.  Unfortunately, when I installed BOINC and Einstein on it, I just set it and forgot about. In the meantime, I've been checking to make sure the latest Windows updates, BIOS, drivers, etc. have been installed.  And I only discovered this issue last night.  I will double check the GPU driver.

I also checked my invalids again and they are rising significantly.  So I changed my project preferences to stop running Beta apps on this machine (temporarily).  If that helps, the likely suspect is the Beta app.

As I said earlier, I haven't been on the boards here much in a long while. In fact, not at all since the web-site was redesigned.  If it looks like the Beta app is the culprit here and my host can help getting the bugs out, I am willing to help.  But I am not exactly sure where on the boards to go for reporting. So I may need your assistance again.  If so, I will PM you.

Stick

 

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.