Asking for TF, getting DC

lukerichards · 2020-03-01, 11:50

Hi,

I have a computer that is only ever on occasionally, but when it is on it's usually on for a good few hours (say 3-6 hours) at a time.

It has access to 16 cores and I've set P95 to run 16 workers running TF on each.

Worker 1 collects a TF job, but PrimeNet is assigning LL DC work. Is there any way to override this?

Worker 1 had 14 jobs queued up so for now I've gone into worktodo.txt and reallocated 13 of them manually to workers 2-14, but P95 has just appended the LL DC job to do after that TF is done.

I don't want to do LL work on here - a job takes far too long on a computer that is only running for, say, 12-24 hours a month.

ATH · 2020-03-01, 12:28

Probably because that computer never did a DC and your account is set to "one matching double-check yearly" for each computer:

https://www.mersenne.org/thresholds/

lukerichards · 2020-03-01, 12:30

Update:

I logged into my mersenne.org account and found that Workers 2-14 were set for "Whatever makes most sense", despite that not being the setting in Prime95.

Having manually changed each core on mersenne.org to be doing TF work, closed P95 and reopened it, we're now up and running as expected.

chris2be8 · 2020-03-01, 16:31

TF on a CPU is not much use since GPUs do TF much faster than CPUs do. Has it got enough RAM to do P-1 tests? That would be more use to the project.

Chris

kriesel · 2020-03-01, 17:07

How many TF GhzD/day can the cpu in question provide? A good gpu can do 1000 to 3000, 24/7. (GTX1080, 1650 are ~1000; RTX2080 Super ~3300. The best use of a cpu core for TF is to keep a good gpu well fed, or several. https://www.mersenne.ca/mfaktc.php https://www.mersenne.ca/throughput.p...2288&mhz2=2700)

I'll occasionally run cpu TF, but only for test purposes, or the very rare foray into very high exponents that available gpu software like mfaktx can't handle.

2020-03-01, 11:50	#1
lukerichards "Luke Richards" Jan 2018 Birmingham, UK 2⁵·3² Posts	Asking for TF, getting DC Hi, I have a computer that is only ever on occasionally, but when it is on it's usually on for a good few hours (say 3-6 hours) at a time. It has access to 16 cores and I've set P95 to run 16 workers running TF on each. Worker 1 collects a TF job, but PrimeNet is assigning LL DC work. Is there any way to override this? Worker 1 had 14 jobs queued up so for now I've gone into worktodo.txt and reallocated 13 of them manually to workers 2-14, but P95 has just appended the LL DC job to do after that TF is done. I don't want to do LL work on here - a job takes far too long on a computer that is only running for, say, 12-24 hours a month.

2020-03-01, 17:07	#5
kriesel "TF79LL86GIMPS96gpu17" Mar 2017 US midwest 4,903 Posts	TF is a waste of a good cpu (except...) How many TF GhzD/day can the cpu in question provide? A good gpu can do 1000 to 3000, 24/7. (GTX1080, 1650 are ~1000; RTX2080 Super ~3300. The best use of a cpu core for TF is to keep a good gpu well fed, or several. https://www.mersenne.ca/mfaktc.php https://www.mersenne.ca/throughput.p...2288&mhz2=2700) I'll occasionally run cpu TF, but only for test purposes, or the very rare foray into very high exponents that available gpu software like mfaktx can't handle.

2020-03-01, 12:28	#2
ATH Einyen Dec 2003 Denmark 7×433 Posts	Probably because that computer never did a DC and your account is set to "one matching double-check yearly" for each computer: https://www.mersenne.org/thresholds/

2020-03-01, 12:30	#3
lukerichards "Luke Richards" Jan 2018 Birmingham, UK 2⁵×3² Posts	Update: I logged into my mersenne.org account and found that Workers 2-14 were set for "Whatever makes most sense", despite that not being the setting in Prime95. Having manually changed each core on mersenne.org to be doing TF work, closed P95 and reopened it, we're now up and running as expected.

2020-03-01, 16:31	#4
chris2be8 Sep 2009 3·23·29 Posts	TF on a CPU is not much use since GPUs do TF much faster than CPUs do. Has it got enough RAM to do P-1 tests? That would be more use to the project. Chris