Heads-up: BRP4 might run briefly out of work

Link
Link
Joined: 15 Mar 20
Posts: 121
Credit: 9177934
RAC: 47812

Ah, OK. Than it will be BRP4

Ah, OK. Than it will be BRP4 without the G for my Android device in the future, let's hope the returned results won't be "torpedoed" by too many Intel-GPUs like in the past (that's why I moved it to BRP4G, there I got 100% validation rate against the x86 CPU applications).

.

Oliver Behnke
Oliver Behnke
Moderator
Administrator
Joined: 4 Sep 07
Posts: 982
Credit: 25170813
RAC: 0

Crossing my fingers. It's

Crossing my fingers. It's always a fine line between maximum cross-validation and algorithmic accuracy.

Thanks!

Einstein@Home Project

Link
Link
Joined: 15 Mar 20
Posts: 121
Credit: 9177934
RAC: 47812

Inconclusives piling up

Inconclusives piling up already. Any chance to return to this pretty well working solution, now that you don't use the "G" for x86 CPUs (eventually with standard sized WUs if that works better)? AFAICT, also the v1.61 application validates well against Intel GPU but not so well against Android, so should go to this group as well. The invalid rate for BRP4 is still around 20%, splitting it into two groups should eliminate a lot of these "invalids" and with that save lots of computation time. Basically same like you made a separate group for Apple.

.

Oliver Behnke
Oliver Behnke
Moderator
Administrator
Joined: 4 Sep 07
Posts: 982
Credit: 25170813
RAC: 0

Darn, unfortunately I

Darn, unfortunately I likewise have to quote myself. Doing this isn't trivial, unfortunately, and we're currently not at full strength. Please bare with us for the next two weeks :-/

Oliver

Einstein@Home Project

astro-marwil
astro-marwil
Joined: 28 May 05
Posts: 531
Credit: 626266543
RAC: 1097539

Hallo! Will it be possible

Hallo!

Will it be possible to arrange crunching also with GPUs from AMD, Intel and NVIDIA under BRP4A control? BRP4G performed very similar well as BRP4A. But probably the performances of the servers of BRP4A are too low for this.

Kind regards and happy crunching

Martin

Oliver Behnke
Oliver Behnke
Moderator
Administrator
Joined: 4 Sep 07
Posts: 982
Credit: 25170813
RAC: 0

We'll look into this when we

We'll look into this when we reconsider the BRP4 searches as soon as we're back to normal operations. The compute cycles you thankfully donate should always be used in the best way possible, which is why we've to assign the available resources by scientific value.

Thanks

Einstein@Home Project

Link
Link
Joined: 15 Mar 20
Posts: 121
Credit: 9177934
RAC: 47812

This is pointless, already

This is pointless, already 50% invalid and will likely end up at around 80-90% as every returend result becomes inconclusive. Too many Intel GPUs and "v1.61 aarch64-unknown-linux-gnu". The only valid WU so far was one with a wingman running a custom CUDA application on NVIDIA Tegra X1 (and that computer also seems to have over 50% invalid results). BRP4A doesn't work either. Moving the phone to Moo! Wrapper until this is fixed.

.

Sid
Sid
Joined: 17 Oct 10
Posts: 164
Credit: 963217145
RAC: 380529

Link написал: This is

Link wrote:

This is pointless, already 50% invalid and will likely end up at around 80-90% as every returend result becomes inconclusive. Too many Intel GPUs and "v1.61 aarch64-unknown-linux-gnu". The only valid WU so far was one with a wingman running a custom CUDA application on NVIDIA Tegra X1 (and that computer also seems to have over 50% invalid results). BRP4A doesn't work either. Moving the phone to Moo! Wrapper until this is fixed.

That probably was me however my error rate is a bit more thet 30%. Makes no sense to proceed.

PorkyPies
PorkyPies
Joined: 27 Apr 16
Posts: 199
Credit: 33740004
RAC: 6814

For my Pi Cluster I have

For my Pi Cluster I have around a 5 to 6% rate of invalids.

Of the invalids I looked at they aren't all Intel GPU, a lot are Android not validating. Looks like v1.61 and v1.47 don't like validating against each other. The v1.51 (android) seems to match the v1.70 (intel) app and the v1.61 is declared invalid.

Maybe they need to compare like versions with each other.

mikey
mikey
Joined: 22 Jan 05
Posts: 12640
Credit: 1839026286
RAC: 5469

PorkyPies wrote: For my Pi

PorkyPies wrote:

For my Pi Cluster I have around a 5 to 6% rate of invalids.

Of the invalids I looked at they aren't all Intel GPU, a lot are Android not validating. Looks like v1.61 and v1.47 don't like validating against each other. The v1.51 (android) seems to match the v1.70 (intel) app and the v1.61 is declared invalid.

Maybe they need to compare like versions with each other.

Mine are similar: 

 

Comment viewing options

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