Prime95 v30.3

[Chuck]

I got a good result from this CERT and the result was successfully uploaded; Ben's original PRP now is highlighted in green.

But on my screen it says MD5 starting value does not match and keeps retrying the same CERT.

[pinhodecarlos]

Hi Chuck, yesterday I got the same issue but on my side the client managed to abort the task on first attempt and now I’m doing CERT work. Haven’t done anything, just let the client manage by itself. The only tweaks I have on my setup is 100% request time for CERT, unlimited bandwidth.
A side note, what CPU are you using, my old machine takes 7 days to LL DC at 50M exponents and 2 hours for CERT at 98M.

[retina]

Can I ask why CERT work is considered enough of a priority to interrupt the usual PRP? It makes the estimated completion times completely useless, and introduces extra overhead in saving and reloading so often.

Is this a server resource limitation? Or are we just trying to give the original tester a warm fuzzy when the CERT is validated quickly?

We used to leave confirmation of LL tests for many years. Why the sudden haste now?

[Batalov]

Perhaps they purge the savefile from the server once the verification is complete?
It is a big chunk of disk to keep for many years.

[retina]

Originally Posted by Batalov

Perhaps they purge the savefile from the server once the verification is complete?
It is a big chunk of disk to keep for many years.

Sure, many years is difficult to manage. But can it wait until the current PRP is completed and then work on any pending CERTs?

[Chuck]

Originally Posted by pinhodecarlos

Hi Chuck, yesterday I got the same issue but on my side the client managed to abort the task on first attempt and now I’m doing CERT work. Haven’t done anything, just let the client manage by itself. The only tweaks I have on my setup is 100% request time for CERT, unlimited bandwidth.
A side note, what CPU are you using, my old machine takes 7 days to LL DC at 50M exponents and 2 hours for CERT at 98M.

I just deleted the CERT from my worktodo file since it had already completed successfully.

Using an Intel I7-7800x; takes about 30 hours to do a DC and 30 minutes for a CERT.

[storm5510]

If we retired LL and LL-DC now, the amount of work needing done would remain large. LL-DC's would need to pass through PRP as a first-time test to get to a certification. Any way you want to slice it, the time required could be years. However, it would all eventually catch up as PRP-DC no longer exists.

[S485122]

Originally Posted by retina

Can I ask why CERT work is considered enough of a priority to interrupt the usual PRP? It makes the estimated completion times completely useless, and introduces extra overhead in saving and reloading so often.

Is this a server resource limitation? Or are we just trying to give the original tester a warm fuzzy when the CERT is validated quickly?

We used to leave confirmation of LL tests for many years. Why the sudden haste now?

Good posting IMHO.

Jacob

[Prime95]

Originally Posted by retina

Can I ask why CERT work is considered enough of a priority to interrupt the usual PRP? It makes the estimated completion times completely useless, and introduces extra overhead in saving and reloading so often.

Is this a server resource limitation? Or are we just trying to give the original tester a warm fuzzy when the CERT is validated quickly?

Yes, it is a server resource issue. I don't want to be in the position of having tens of thousands of large CERT files on the server.

If the interruptions are annoying, set the daily download limit to 12MB and you'll get one Ben Delo cert per day (or several PRP-CF certs). You can also set options to only get wavefront CERTs (see undoc.txt: exponent range of 80M and up should do) if you find the PRP-CF interruptions particularly annoying.

Not coded yet on the server, but I'm thinking of a 5-day expiration for CERTs.

[S485122]

Originally Posted by Prime95

...
(or several PRP-CF certs)
...

Why should people get involved in CF work ?
When looking at what is done, it seems not very well organised : people certifying many times that a CF is not prime.
CF has little to do with GIMPS. I do understand that some want to find factors (even if one project is setting aims that are not rational.) No problem with that (their kit, their power, their strife...) But why should "ordinary" users of PrimeNet be forced, by default, to participate in that work ?

Jacob

[Prime95]

Originally Posted by S485122

Why should people get involved in CF work ?
When looking at what is done, it seems not very well organised : people certifying many times that a CF is not prime.
CF has little to do with GIMPS. I do understand that some want to find factors (even if one project is setting aims that are not rational.) No problem with that (their kit, their power, their strife...) But why should "ordinary" users of PrimeNet be forced, by default, to participate in that work ?

Fair question. GIMPS main goal is finding new Mersenne primes, but it also is a learn all we can about Mersenne numbers. In that regard, PRP-CF is GIMPS related, but I agree its a niche project.

What do others think? I could easily change the default to only cert exponents over 50M unless you've signed up for PRP-CF work. It would be more work, but I could also try handing out PRP-CF certs in batches of 5 or 10 for less interruptions.