PRP triple checks?

[GP2]

PRP triple check needed.

This should be done with mprime/Prime95 rather than gpuOwL because it needs to be residue type 1.

PRPDC=1,2,86670349,-1,76,0,3,1

[S485122]

Originally Posted by GP2

PRP triple check needed.
...

This is not supposed to happen. Both PRP's have been done this year, I suppose with mature software.

Some people even suggested that with PRP, except for the possibility of cheating, there was no need for double checks.

I think the run that yielded the wrong result should be thoroughly investigated.

Jacob

[endless mike]

Originally Posted by GP2

PRP triple check needed.

This should be done with mprime/Prime95 rather than gpuOwL because it needs to be residue type 1.

PRPDC=1,2,86670349,-1,76,0,3,1

I don’t think that’s true about not using gpuowl. I’m still on v3.8 (old I know) but all my results are type 1.

[Prime95]

Originally Posted by S485122

This is not supposed to happen. Both PRP's have been done this year, I suppose with mature software.

True, this isn't supposed to happen. Simon's machine is the trouble-maker. Fortunately, this led to a beefed up implementation of Gerbicz error checking in v29.6.

[kriesel]

Originally Posted by endless mike

I’m still on v3.8 (old I know) but all my results are type 1.

Smart. In some cases (exponent, gpu model, OS) V3.8 is the fastest available suitable version of gpuowl. See the pdf attachments at https://www.mersenneforum.org/showpo...35&postcount=2

Originally Posted by Prime95

Simon's machine is the trouble-maker. Fortunately, this led to a beefed up implementation of Gerbicz error checking in v29.6.

A "canary in the coal mine" here and there is a good thing, alerting to issues that might not otherwise be found in time.