GPU to 72 status...

[flashjh]

Originally Posted by Old man PrimeNet

The fix was to block a range of IP addresses. Found someone passing high-volume trial-factoring work on many nodes. See my #1 posting Oct 2008. Need to play nice or don't play at all.

Does anyone know who got blocked in reference to this post?

[chalsall]

Originally Posted by flashjh

Does anyone know who got blocked in reference to this post?

I personally don't. It wasn't GPU72 -- since we're now mostly doing ->73 work our aggregate submissions have decreased by about half (as expected).

[ckdo]

Originally Posted by ckdo

We have just finished DCTF on exponents below 26M (for the time being).

We have now finished DCTF on all unverified candidates below 32M.

[chalsall]

Originally Posted by ckdo

We have now finished DCTF on all unverified candidates below 32M.

Yup. Here's the relevant report for anyone interested.

Also, so everyone knows, we're currently 436 days ahead of the DC wave-front.

Thanks for all the cycles everyone!

[nucleon]

Originally Posted by petrw1

Over 100,000 total P-1 work done.
Craig Meyers: Over 500,000 work done; over 1,000 factors found.

Over 200,000 saved now.

Factor and ye shall be saved :)

-- Craig

[lycorn]

Originally Posted by chalsall

Also, so everyone knows, we're currently 436 days ahead of the DC wave-front.

Wow! That seems to indicate we should start putting more effort on LL TF, isn´t it?

BTW, how far ahead are we from LL wavefront (in case we actually are...)

[chalsall]

Originally Posted by lycorn

Wow! That seems to indicate we should start putting more effort on LL TF, isn´t it?

That's really already happening. Our biggest DCTF worker "to completion", bcp19, is splitting his work between DCTF and LLTF in a nearly perfect ratio to maintain the lead ahead of 131 DC completions a day.

Originally Posted by lycorn

BTW, how far ahead are we from LL wavefront (in case we actually are...)

Approximately 37 days ahead. So, definitely, anyone wondering what work type they should do, LLTF is higher priority. Or, LL P-1.

My hope is we'll have everything cleared out below 60M right around the one year anniversary of GPU72 -- November 1st. Then we can consider coming back down, and taking some of 56M to 73 as the candidates become available.

[kladner]

I just TF'd this one to 73. I see below that it is "AssignedLL testing to "ANONYMOUS" on 2012-04-26".

58002047No factors below2^73 
P-1B1=605000, B2=11797500 
AssignedLL testing to "ANONYMOUS" on 2012-04-26 
Historyno factor from 2^63 to 2^64 by "Torben_Kock" on 2008-10-25 
Historyno factor from 2^64 to 2^65 by "Richard Kint" on 2008-10-26 
Historyno factor from 2^65 to 2^66 by "Sturle Sunde" on 2008-10-26 
Historyno factor from 2^66 to 2^67 by "Sturle Sunde" on 2008-10-26 
Historyno factor from 2^67 to 2^68 by "DoubleAgent" on 2009-12-04 
Historyno factor for M58002047 from 2^68 to 2^71 [mfaktc 0.17-Win barrett79_mul32] by "rduerr" on 2011-10-02 
HistoryB1=605000, B2=11797500 by "Abdullah Baig" on 2012-04-24 
Historyno factor for M58002047 from 2^71 to 2^73 [mfaktc 0.19 barrett76_mul32] by "ktony" on 2012-09-19

Is this normal? It worked out OK, in that it still has no known factors. But I'm curious as to how this came about.

[chalsall]

Originally Posted by kladner

Is this normal? It worked out OK, in that it still has no known factors. But I'm curious as to how this came about.

I have a spider which looks for old assignments to ANONYMOUS which appear to be abandoned, and attempts to reserve them. I'll have to look at the code to see why it thought it owned this assignment. I'll disable this spider until I figure it out.

[petrw1]

Originally Posted by chalsall

Approximately 37 days ahead. So, definitely, anyone wondering what work type they should do, LLTF is higher priority. Or, LL P-1.

My hope is we'll have everything cleared out below 60M right around the one year anniversary of GPU72 -- November 1st. Then we can consider coming back down, and taking some of 56M to 73 as the candidates become available.

I don't know for sure but it appears the LL wavefront is close to 60M or at least there are assignments over 60M.
Is gpu72 now far enough into the 60s to ensure no more LL is handed out without deep TF?

In other words is it mostly now just going on ahead more and more in the future?

[chalsall]

Originally Posted by petrw1

I don't know for sure but it appears the LL wavefront is close to 60M or at least there are assignments over 60M.
Is gpu72 now far enough into the 60s to ensure no more LL is handed out without deep TF?

To the best of my knowledge, Primenet has never handed out a LL assignment above 60M without the user specifically requesting it. Additionally, I would argue the wave-front should be considered to be below 59M at the moment as even late in the day (UTC) only occasionally does Primenet hand out assignments in 58M.

Originally Posted by petrw1

In other words is it mostly now just going on ahead more and more in the future?

Pretty much.

Right now we're finishing off 58M to 73 -- 7,127 to go. Almost all of the candidates reserved by the system in 57M are being (or soon will be) P-1'ed. Every day we get a few candidates to TF in 57M, but not many; right now only 217 are being TFed there.

According to our rolling 30 day average TF production, we should be able to clear out everything below 60M to the appropriate TF level in about 45 days. Then, like I said, we can consider coming back down and doing some additional TFing in the 56M (and possibly 55M) range. James' analysis shows that TFing to 73 is "profitable" down to about 55M, but only after we're comfortably ahead of the wave.

Others, of course, may simply prefer to march ahead into the 60Ms. Or, start using their GPUs for LLing.