[QUOTE=James Heinrich;128976]I think v5 server ran out of TF work? I'm getting ECM work instead.[/QUOTE]Hmm... I've been running into this (type of) issue over the last little while -- the server doesn't seem to be registering the worktype-per-thread preferences as set in the Prime95 client (at least not all the time). For example, I tried setting thread #1 to P-1 and threads 2/3/4 to TF, but then all threads got assigned P-1 work, and [url]http://v5www.mersenne.org/editcpu/[/url] showed all threads having a P-1 preference. I've set the preference from the webpage now and it seems to stick, so I assume it's just a client-server communication problem?

[QUOTE=James Heinrich;130337]I've set the preference from the webpage now and it seems to stick, so I assume it's just a client-server communication problem?[/QUOTE]There goes that theory. After setting it on the website this morning, I come home to find all 4 threads have P-1 work queued up. Something's not behaving :sad:

[QUOTE=James Heinrich;130421]There goes that theory. After setting it on the website this morning, I come home to find all 4 threads have P-1 work queued up. Something's not behaving :sad:[/QUOTE]

Email the prime.log file

I just fixed a server bug. If you changed the work preference for the first worker thread, the server thought you were changing the work preference for all worker threads.

Hopefully that explains your problem.

[QUOTE=Prime95;130519]I just fixed a server bug... Hopefully that explains your problem.[/QUOTE]That sounds exactly like what I was experiencing. Thanks. I'll let you know if it reoccurs.

I was working on the get assignment server code. I broke it briefly and probably screwed a few clients.

If you have any trial factoring assignments where the "how-far-to-factor-to" value is 80 bits in worktodo.txt, then edit the worktodo.txt and change the 80 to something reasonable like 59.

Also note that v5 manual assignments and assignment extension forms are now operational. Manual results reporting is much harder and not yet implemented.

[QUOTE=Prime95;130848]...probably screwed a few clients.[/QUOTE]Ouch! :shock:
Even assuming a relatively fast CPU (my Core2 @ 3.5GHz) it should take roughly 2.8 [i]years[/i] to TF to 2^80 (assuming my calculations are right).

BTW: the Prime95 (25.6-32) client doesn't seem to calculate ETA properly with such large TF target: I changed one current assignment from "to 2^67" to "to 2^80" and it estimated I would be done about 3 hours in the past :ermm:

[QUOTE=Prime95;130848]If you have any trial factoring assignments where the "how-far-to-factor-to" value is 80 bits in worktodo.txt, then edit the worktodo.txt and change the 80 to something reasonable like 59.[/QUOTE]

Slightly un-related, I noticed that the factoring limits went up.

M4742xxxx - 68
M4752xxxx - 69

[QUOTE=Uncwilly;131153]Slightly un-related, I noticed that the factoring limits went up.[/QUOTE]I believe those would be post-Pminus1 trialfactoring. From what I remember, George said that the new approach is to TF up to ~67 bits (in the current active range anyways), then do P-1, then continue TF another 2 bit depths to the final level (currently ~2^69). My machine is currently working in the ~56M range for regular TF, and ~46M range for P-1.

[QUOTE=Uncwilly;131153]Slightly un-related, I noticed that the factoring limits went up.

M4742xxxx - 68
M4752xxxx - 69[/QUOTE]Yep - the boundary between 68 & 69 bits is 47450000, since 24.12. Details [URL="http://www.mersenneforum.org/showthread.php?t=4213"]here[/URL].