GPU to 72 status... - Page 98

chalsall · 2012-04-06, 01:04

Quote:

Originally Posted by nucleon

I selected oldest exponent and it allocated based on lowest TF level. I can't rule out user error. I went back button in browser and it showed oldest.

The "select" was sorting by lowest TF level first, and then age. Changed to be solely dependent upon age.

nucleon · 2012-04-06, 01:37

Quote:

Originally Posted by chalsall

The "select" was sorting by lowest TF level first, and then age. Changed to be solely dependent upon age.

Cheers.

bcp19 · 2012-04-06, 14:50

On the View Assignments page, the table at the top shows the overall average from start to date, which for the most part is fine, as LL/P-1 take a lot of time compared to TF. On TF though, if someone makes a system change, it takes a long time to get close to current capability, and mine shows ~1/2 of my current. Maybe it should show a 10 or 15 day average?

Dubslow · 2012-04-06, 19:43

Heh, regarding my P-1 factors, I apparently found two back-to-back overnight, but of course they were both relatively small (76.x and 81.x).

Edit:

http://mersenne.info/exponent_status_line_graph/1/0/
See the bump?

Radikalinsky · 2012-04-06, 20:36

Quote:

Originally Posted by chalsall

And as some might have noticed, we're now effectively out of all work except 71 -> 72.

That's a pity, because 2 of my GPUs are slow and old, with only Compute Capability 1.1. Therefore, they get ineffective above 71. Unfortunately, I forgot how ineffective.

At the moment a few exponents are available below 71, but I get this nasty
* Note: Minumum is 72.
Their contribution is about 25 GHzDays/day, or it was.
I think it was useful to factor these exponents up to 71, and then someone else takes them to 72 or 73.

I could pledge to factor them to 72 and factor them to only 71.
Or I could switch to the DC range, which is far ahead of the "wave", or to CUDALucas.

Do you have better suggestions?

Thanks,
Rad

Dubslow · 2012-04-06, 20:39

You might consider DCTF, they're at lower bit levels.

Radikalinsky · 2012-04-06, 20:44

DCTF is not very useful, because it is so far ahead.
I could do the part up to 71 in the slow GPUs, and the remainder to 72 or 73 in a fast new GPU, but that is complicated.
Perhaps CUDALucas is the most useful option.

chalsall · 2012-04-06, 21:12

Quote:

Originally Posted by Radikalinsky

DCTF is not very useful, because it is so far ahead.
I could do the part up to 71 in the slow GPUs, and the remainder to 72 or 73 in a fast new GPU, but that is complicated.

I understand your concern. And I only did the "Minumum pledge is 72" after effectively everything below 60M was already taken to 71. Yes, a few below 71 trickle out every day, but it is a very small number and I want to see them processed in one assignment cycle (or two, if P-1 is needed).

Quote:

Originally Posted by Radikalinsky

Perhaps CUDALucas is the most useful option.

Unfortunately your cards can't run CUDALucas.

One thing I could do for people like you who have cards like you have is to bring in a few candidates at 60M which are already at 70, and make them available to only go to 71.

They're well ahead of the "wave", but the work would be more beneficial to the project than DCTF (in my opinion).

Thoughts?

Radikalinsky · 2012-04-06, 21:26

Quote:

One thing I could do for people like you who have cards like you have is to bring in a few candidates at 60M which are already at 70, and make them available to only go to 71.

Thank you, this sounds more useful than DCTF.

chalsall · 2012-04-06, 21:42

Quote:

Originally Posted by Radikalinsky

Thank you, this sounds more useful than DCTF.

OK. A quick hack before I run out for a "Sundowner"...

If you go to the LLTF form and choose the new option "I have a slow card!!!" it will ignore the range and pledge fields, and issue you work between 60M and 61M to TF from 70 to 71.

Dubslow · 2012-04-08, 02:30

Quote:

Originally Posted by chalsall

No. We've got a large batch of TF results without P-1 expected to complete shortly, so I threw 600 back to PrimeNet to hand out as it feels is appropriate.

Hmm... now we're down to 63...

2012-04-06, 19:43	#1071
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 3·29·83 Posts	Heh, regarding my P-1 factors, I apparently found two back-to-back overnight, but of course they were both relatively small (76.x and 81.x). Edit: http://mersenne.info/exponent_status_line_graph/1/0/ See the bump? Last fiddled with by Dubslow on 2012-04-06 at 20:15

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Status	Primeinator	Operation Billion Digits	5	2011-12-06 02:35
62 bit status	1997rj7	Lone Mersenne Hunters	27	2008-09-29 13:52
OBD Status	Uncwilly	Operation Billion Digits	22	2005-10-25 14:05
1-2M LLR status	paulunderwood	3*2^n-1 Search	2	2005-03-13 17:03
Status of 26.0M - 26.5M	1997rj7	Lone Mersenne Hunters	25	2004-06-18 16:46

2012-04-06, 14:50	#1070
bcp19 Oct 2011 1010100111₂ Posts	On the View Assignments page, the table at the top shows the overall average from start to date, which for the most part is fine, as LL/P-1 take a lot of time compared to TF. On TF though, if someone makes a system change, it takes a long time to get close to current capability, and mine shows ~1/2 of my current. Maybe it should show a 10 or 15 day average?

2012-04-06, 20:39	#1073
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 3×29×83 Posts	You might consider DCTF, they're at lower bit levels.

2012-04-06, 20:44	#1074
Radikalinsky Mar 2011 Germany 1011₂ Posts	DCTF is not very useful, because it is so far ahead. I could do the part up to 71 in the slow GPUs, and the remainder to 72 or 73 in a fast new GPU, but that is complicated. Perhaps CUDALucas is the most useful option.