mersenneforum.org GPU to 72 status...
 Register FAQ Search Today's Posts Mark Forums Read

2019-02-18, 18:08   #4291
TheJudger

"Oliver"
Mar 2005
Germany

100010101102 Posts

Quote:
 Originally Posted by Chuck Is there any preference whether I do TF spanning multiple bitlevels all at once or split it into stages?
Once the runtime per class is lets say something like 0.5s it is more efficient to do it step by step (read: stages or just on bitlevel at a time).

2019-02-18, 18:54   #4292
chalsall
If I May

"Chris Halsall"
Sep 2002

222728 Posts

Quote:
 Originally Posted by Chuck Is there any preference whether I do TF spanning multiple bitlevels all at once or split it into stages?
If I understand your question correctly...

We currently have a bit of a temporal buffer, but going to 77 "bits" is most desired at the moment. Working candidates currently at 76 takes the least amount of effort.

Going to 77 takes quite a bit of effort, and fewer factors will be found per "wall clock" time compared to TF'ing to lower bit levels, but we're trying to stay ahead of the Primenet P-1'ers (who are staying ahead of the Primenet LL'ers).

But, as always, do what you enjoy doing. And thanks for all the work!

 2019-02-19, 01:12 #4293 Chuck     May 2011 Orange Park, FL 22·7·31 Posts I was asking since "Let GPU72 decide" is giving me worktodo lines Code: Factor=2019-2-18,92116987,74,76 Factor=2019-2-18,92116993,74,76 Factor=2019-2-18,92117009,74,76 Factor=2019-2-18,92117029,74,76 Factor=2019-2-18,92117041,74,76 Factor=2019-2-18,92117173,74,76 Factor=2019-2-18,92117177,74,76 Factor=2019-2-18,92117183,74,76 Factor=2019-2-18,92117191,74,76 should I set stages=1 and let it run each bitlevel separately. In any case I am going to stop as soon as a factor is found.
 2019-04-21, 19:15 #4294 chalsall If I May     "Chris Halsall" Sep 2002 Barbados 100100101110102 Posts GPU72 is about to run out of "to 73 bits" work... Hey All. Just so everyone knows, within the next few hours GPU72 is going to run out of work to take from 72 to 73 "bits", in 94M. There is a little bit of this still left in 95M and above, but someone with quite a bit of firepower has been working this "off the books". In fact, they've been a little impolite, and have "stepped on toes" in 94M. Not nice, but not much I can do about it. So, shortly, anyone who pledges to "only" 73 bits will be soon given work to 74 bits instead. Initially in 92M, and then working upwards.
2019-04-22, 19:29   #4295
Thecmaster

"Arvid Björklin"
Apr 2016
Pitea, Sweden

73 Posts

Quote:
 Originally Posted by chalsall Hey All. Just so everyone knows, within the next few hours GPU72 is going to run out of work to take from 72 to 73 "bits", in 94M. There is a little bit of this still left in 95M and above, but someone with quite a bit of firepower has been working this "off the books". In fact, they've been a little impolite, and have "stepped on toes" in 94M. Not nice, but not much I can do about it. So, shortly, anyone who pledges to "only" 73 bits will be soon given work to 74 bits instead. Initially in 92M, and then working upwards.
I know you have the mindset to do whatever you like doing but maybe write a little guideline we can try to follow. Something like. 72-73 maximum 30 exponents / day. 73-74 and 74-75 maximum 50 /day and 75-77 unlimited.

Or something else that fits all of us so all with low end gpu have ok work to do and don't need to do a half exponent/day or so.

2019-04-22, 19:47   #4296
chalsall
If I May

"Chris Halsall"
Sep 2002

2·3·1,567 Posts

Quote:
 Originally Posted by Thecmaster Or something else that fits all of us so all with low end gpu have ok work to do and don't need to do a half exponent/day or so.
I understand and appreciate that not everyone has a load of GPUs to work on TF'ing to the optimal depth. I myself don't (not to mention electricity is stupidly expensive here in Barbados).

The best advice I can give is to work as deeply as you feel comfortable, while still taking enjoyment from the factors found per wall-clock time.

As I've said before, GPU72 was created to help the GIMPS effort. There is still work available directly from Primenet to lower bit levels, but the results won't be of any real use for many years.

I hope that makes sense, and speaks to your question.

2019-04-22, 21:24   #4297
Thecmaster

"Arvid Björklin"
Apr 2016
Pitea, Sweden

73 Posts

Quote:
 Originally Posted by chalsall I understand and appreciate that not everyone has a load of GPUs to work on TF'ing to the optimal depth. I myself don't (not to mention electricity is stupidly expensive here in Barbados). The best advice I can give is to work as deeply as you feel comfortable, while still taking enjoyment from the factors found per wall-clock time. As I've said before, GPU72 was created to help the GIMPS effort. There is still work available directly from Primenet to lower bit levels, but the results won't be of any real use for many years. I hope that makes sense, and speaks to your question.
I didn't got hit by this. I'm doing 74-75 bit in 92M range. I'm finishing 8 assignment per day right now and thay feels like a good tradeoff wall clock time vs factor found. So my thoughts and ideas was for those who's doing tf with slower gpu and still wants to do enough exponents per day to have a chances to find factors.

And. It was just thoughts and ideas so don't feel like you have to change anything.

😊😊😊😊😊😊😊

2019-04-22, 21:33   #4298
James Heinrich

"James Heinrich"
May 2004
ex-Northern Ontario

CAA16 Posts

Quote:
 Originally Posted by Thecmaster I'm finishing 8 assignment per day right now and thay feels like a good tradeoff wall clock time vs factor found.
In the end it's all about what makes each user happy / motivates them. Personally I churn about 23000 candidates (for ~300 factors) per day, and that makes me happy (but with the knowledge that my efforts likely won't contribute to finding any Mersenne primes in my lifetime).

2019-04-22, 23:16   #4299
chalsall
If I May

"Chris Halsall"
Sep 2002

2·3·1,567 Posts

Quote:
 Originally Posted by James Heinrich In the end it's all about what makes each user happy / motivates them.
Indeed. And further, the administrators and developers can enjoy finding "edge cases" while the users test the envelope.

This is actually rather important. It lets us test things, to see if we've made any mistakes for our "real work".

 2019-05-04, 07:56 #4300 petrw1 1976 Toyota Corona years forever!     "Wayne" Nov 2006 Saskatchewan, Canada 10001101100112 Posts 100M Total work completed just surpassed 100M GhzDays. That is 3 Curtis Cooper lifetime's. Over 72,407 factors GhzDays saved: 32,325,356 and counting... 1 CC lifetime.
 2019-05-13, 22:33 #4301 ixfd64 Bemusing Prompter     "Danny" Dec 2002 California 34·29 Posts I noticed GPU to 72 is no longer giving assignments to 73 bits. Any idea when 73-bit assignments will be available again?

 Similar Threads Thread Thread Starter Forum Replies Last Post Primeinator Operation Billion Digits 5 2011-12-06 02:35 1997rj7 Lone Mersenne Hunters 27 2008-09-29 13:52 Uncwilly Operation Billion Digits 22 2005-10-25 14:05 paulunderwood 3*2^n-1 Search 2 2005-03-13 17:03 1997rj7 Lone Mersenne Hunters 25 2004-06-18 16:46

All times are UTC. The time now is 07:58.

Mon Jan 18 07:58:57 UTC 2021 up 46 days, 4:10, 0 users, load averages: 2.50, 2.17, 2.07