GPU to 72 status... - Page 172

kracker · 2013-02-08, 00:05

Quote:

Originally Posted by chalsall

No.

Because a DC candidate only needs about one "bit" less of TFing than a LL candidate in about three years or so there will be no DCTF left to do.

Ok, I see what you mean.

petrw1 · 2013-02-08, 00:15

Quote:

Originally Posted by ckdo

Your link is broken. Anyway, the ETAs are of the "if everyone does only this" type.

Ah, missed that one little factoid...I assumed (yeah, I know) he was basing it on progress for each work type

Aramis Wyler · 2013-02-08, 19:11

FWIW, I think if we are trusting James' latest chart, the 'go deeper' bit will happen before we complete the current set. I am not sure what that graph calculates 'complete' as, (72 bits for 60m range? 73?) but that last graph from James shows the 60M range break even point at > 74, so I figure at some point the 'Estimated Days' graph will be recalibrated. I will be surprised if it takes 3 years to do so.

As an aside, this is the first time I've seen someone mention about the DC ending. I had looked into that some time ago and while I wasn't 100% on at exactly which M range it would end on (I was guessing 56) it did seem certain it would end. I simply couldn't calc how long that would take. It is good to have that knowledge in the public forum.

chalsall · 2013-02-08, 19:28

Quote:

Originally Posted by Aramis Wyler

FWIW, I think if we are trusting James' latest chart, the 'go deeper' bit will happen before we complete the current set. I am not sure what that graph calculates 'complete' as, (72 bits for 60m range? 73?) but that last graph from James shows the 60M range break even point at > 64, so I figure at some point the 'Estimated Days' graph will be recalibrated. I will be surprised if it takes 3 years to do so.

All true. But please keep in mind that even before the "big news" we were only just keeping ahead (and slowly building a buffer in front) of the LL "wave" only going to 73, rather than the 74 where we should have been.

I'm monitoring the assignments being issued by Primenet currently, but there's a chance we'll have to start releasing candidates at only 72 for a while. If that happens, we'll still have about an 80% chance of having another opporutnity to bring them back in for processing to 73 after the post-news "surge" finishes. Only about 20% of all LL assignments actually complete, and that will probably be even lower for what is being assigned at the moment.

Aramis Wyler · 2013-02-08, 19:48

I like that this is a fluid and carefully monitored thing.

I think it would be better if we had a defined target and a defined calulation for how close we were going to attempt to get to target (eg, if the target for 60M is 74 bits, we would relase it at 72 (target-2) if we were < 2 weeks ahead of the wave, or 73 (Target-1) if we were only 1 month ahead of the wave, and would keep it till 74 if we were > 30 days head.

These timespans are arbitrary that I'm making up - it's just to describe the concept of the formula I'm discussing, which in this case would be [Release Bit Depth] = [Target Bit Depth] - trunc([Lead Weeks]/2) where leadweeks < 5.

The point is, I understand that's a bit simplistic and so am glad it's actually monitored by a person with judgement.

ALSO: I think it would be pretty cool to have like, a DEFCON 2 OR DEFCON 1 theme on the page if our leadtime was low enough that we were reducing our release point by necesity.

ALSO ALSO: If I get my old card running in parallell with my new card I'll set the old one to do a bitdepth of just 73. My current one is (currently) on 74.

chalsall · 2013-02-08, 20:08

Quote:

Originally Posted by Aramis Wyler

ALSO: I think it would be pretty cool to have like, a DEFCON 2 OR DEFCON 1 theme on the page if our leadtime was low enough that we were reducing our release point by necesity.

LOL... In that case, ALL HANDS ON DECK!!! WE'RE CURRENTLY AT DEFCON 1!!!

Quote:

Originally Posted by Aramis Wyler

ALSO ALSO: If I get my old card running in parallell with my new card I'll set the old one to do a bitdepth of just 73. My current one is (currently) on 74.

Cool, Thanks.

Every cycle available from everyone could really be used at LLTFing right now, even if it's only going from 70 -> 71. I'm going to move my 560 to LLTFing as well (when it finishes its current "real-work" assignment).

kladner · 2013-02-08, 20:53

Quote:

Originally Posted by chalsall

LOL... In that case, ALL HANDS ON DECK!!! WE'RE CURRENTLY AT DEFCON 1!!!

Every cycle available from everyone could really be used at LLTFing right now, even if it's only going from 70 -> 71. I'm going to move my 560 to LLTFing as well (when it finishes its current "real-work" assignment).

That being the case, let me say that I have some DCTF in queue, but not started. I would be glad to throw them back into the GPU72 pool and let MISFIT fetch only LLTF.

EDIT: What say you, Captain Halsall?

chalsall · 2013-02-08, 21:01

Quote:

Originally Posted by kladner

EDIT: What say you, Captain Halsall?

LOL... As always, people are free to do whatever they want. However, it would be best for GIMPS overall if all cycles went to LLTFing for at least the next month.

As Admiral Woltman himself signed off on, even if a few 32Ms (already at 70 "bits") are assigned before we eventually take all of 31M to 70, it won't be a long-term detriment to the project.

kladner · 2013-02-08, 21:03

Aye aye, Sir!

garo · 2013-02-09, 00:11

Yup we need LLTF on GPUs and P-1 on CPUs. I'm currently suspending DCs and doing P-1 on all four cores of my home rog now that I got extra memory last week.

Redarm · 2013-02-09, 15:53

The range of 65M - 66M seems to be locked...
could anybody verify this?

2013-02-08, 19:11	#1884
Aramis Wyler "Bill Staffen" Jan 2013 Pittsburgh, PA, USA 2³·53 Posts	FWIW, I think if we are trusting James' latest chart, the 'go deeper' bit will happen before we complete the current set. I am not sure what that graph calculates 'complete' as, (72 bits for 60m range? 73?) but that last graph from James shows the 60M range break even point at > 74, so I figure at some point the 'Estimated Days' graph will be recalibrated. I will be surprised if it takes 3 years to do so. As an aside, this is the first time I've seen someone mention about the DC ending. I had looked into that some time ago and while I wasn't 100% on at exactly which M range it would end on (I was guessing 56) it did seem certain it would end. I simply couldn't calc how long that would take. It is good to have that knowledge in the public forum. Last fiddled with by Aramis Wyler on 2013-02-08 at 19:34

2013-02-08, 19:48	#1886
Aramis Wyler "Bill Staffen" Jan 2013 Pittsburgh, PA, USA 650₈ Posts	I like that this is a fluid and carefully monitored thing. I think it would be better if we had a defined target and a defined calulation for how close we were going to attempt to get to target (eg, if the target for 60M is 74 bits, we would relase it at 72 (target-2) if we were < 2 weeks ahead of the wave, or 73 (Target-1) if we were only 1 month ahead of the wave, and would keep it till 74 if we were > 30 days head. These timespans are arbitrary that I'm making up - it's just to describe the concept of the formula I'm discussing, which in this case would be [Release Bit Depth] = [Target Bit Depth] - trunc([Lead Weeks]/2) where leadweeks < 5. The point is, I understand that's a bit simplistic and so am glad it's actually monitored by a person with judgement. ALSO: I think it would be pretty cool to have like, a DEFCON 2 OR DEFCON 1 theme on the page if our leadtime was low enough that we were reducing our release point by necesity. ALSO ALSO: If I get my old card running in parallell with my new card I'll set the old one to do a bitdepth of just 73. My current one is (currently) on 74. Last fiddled with by Aramis Wyler on 2013-02-08 at 19:52

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

2013-02-08, 21:03	#1890
kladner "Kieren" Jul 2011 In My Own Galaxy! 2×3×1,693 Posts	Aye aye, Sir!

2013-02-09, 00:11	#1891
garo Aug 2002 Termonfeckin, IE 2²×691 Posts	Yup we need LLTF on GPUs and P-1 on CPUs. I'm currently suspending DCs and doing P-1 on all four cores of my home rog now that I got extra memory last week.

2013-02-09, 15:53	#1892
Redarm Apr 2012 Berlin Germany 33₁₆ Posts	The range of 65M - 66M seems to be locked... could anybody verify this?