processed dc and tc posts

[Uncwilly]

The list has been updated. There are 6 new suspect results in the Cat 2-3 range.

[Uncwilly]

Originally Posted by Mark Rose

Requesting a triple check of 58255163. This machine was recently returned to service and may have issues.

Sorry, you did not match.

[ATH]

Ready for a TC but will probably be handed out randomly quickly: 53988203

[Runtime Error]

Originally Posted by ATH

Ready for a TC but will probably be handed out randomly quickly: 53988203

I got it. Running as LL. Should be done Monday.

[S485122]

Originally Posted by ATH

Ready for a TC but will probably be handed out randomly quickly: 53988203

You should have specified that since you ran with a shift of 0, the verification should be done with some software that does offer a shift.

Jacob

[jas]

Originally Posted by Uncwilly

DC's on suspect single LL's should no longer be done.

Any single suspect LL should be redone with v30 as PRP-VDF. The balance between the DC and higher than usual TC rate vs running a fresh a PRP & cert lean toward the PRP path to require fewer cycles long term. The error checking is better on PRP.

TC-LLs still make sense to clean up as LL (the required quad rate is low enough).

Why did my mprime v30 client do a LL DC on M54604643? Shouldn't it have done a PRP-VDF instead? If my client had done a PRP-VDF, someone else could have verified the CERT, and the bogus LL result be discarded. Or am I misunderstanding something? Are we doing LL DC's just to discover old faulty machines?



/Simon

[ATH]

If you choose "double-check prime test" in the menu or add lines like Test=<exponent>,74,1 v30.3 will still run LL DC. The only way to run PRP tests in the LL DC range is reserving them as LL DC and then change the worktodo to PRP=<AID>,1,2,<exponent>,-1.
I have not tested if the server will actually accept this, so it might reject it.

There have been discussions if we should run LL DC or PRP Cert on those. If we run PRP Cert we are kind of "throwing" all those peoples LL results out and nullifying all that work

PRP Cert is slightly more efficient but it is only 1%-2% of LL tests on average that is wrong. I don't remember the exact number. Personally I feel like a double checked LL test is slightly "better". I trust that 100% but I "only" trust PRP Cert like 99.99% because I do not understand the proof behind it and how hard it is to fake.

[Uncwilly]

If the first time LL is marked "Suspect" then the chance of it having an error is much higher than the 1-2% that ATH mentioned (which is average overall). Right now most suspects that are the sole check have been already dealt with. If you have a number with one and ask, one of us will generally give you an LL. A random user or an anon user, we won't tend to do it. I am doing one to potentially help someone.

Right now the server is assigning DC's that match the first time check type.

[Jan S]

Doublecheck=54319141,74,1
Doublecheck=54600037,74,1

I took and registered.

[Runtime Error]

Originally Posted by ATH

If you choose "double-check prime test" in the menu or add lines like Test=<exponent>,74,1 v30.3 will still run LL DC. The only way to run PRP tests in the LL DC range is reserving them as LL DC and then change the worktodo to PRP=<AID>,1,2,<exponent>,-1.
I have not tested if the server will actually accept this, so it might reject it.

I've done a bunch of manual testing DCs where I requested it as LL and changed it to PRP myself. The server accepts the manual testing results, but throws a "invalid result type, accepting results but not deleting assignment" error... But it actually does indeed delete the assignment.


Edit: I actually had a couple to turn in just now. Here is a screenshot of the error. 56645269, 56645383, 56645453 have been removed from my assignments. (Proof files have yet to upload.)

[Mark Rose]

Took everything below 86M without a suspect result.

Also, 59167019 is likely bad but someone else took it.