New work types for GPU72?

[KyleAskine]

I agree - we need to do away with work saved in this case.

I think it is a bit of a silly metric anyway - it incentivizes people to take stuff off the top of the range, which isn't really what we want anyway.

The real metric should be 'candidates released to Primenet' or something like that.

[nucleon]

Originally Posted by Xyzzy

Just a question:

If we start factoring real big numbers isn't this going to greatly affect our "work saved" metric?

That is the metric that is the most important, right?

heheh that's what I _was_ doing prior to gpu272. Check out the 380M range. It was sensationally awesome getting into a range hardly touched. I could do great swaths of depth and rack up all sorts of good stats.

But then someone made the point - the project may never reach there with first time tests. If you think that cpus will hit a limit soon, then the great exponential improvements to get to the point where LL completes in a feasible time will never eventuate.

-- Craig

[davieddy]

Sigh...

Oh Ye of little faith.

D

[Uncwilly]

Originally Posted by nucleon

But then someone made the point - the project may never reach there with first time tests. If you think that cpus will hit a limit soon, then the great exponential improvements to get to the point where LL completes in a feasible time will never eventuate.

The main reason I suggested this and would like the help in this range is that there are about 1400 exponents assigned to LL at any given time. Many were or are poorly TF'ed. It would be nice to have a bunch of candidates available ahead of the curve. People are wasting CPU years on some of these exponents.

If it wasn't for the EFF prize, the range would be less interesting.

[chalsall]

Originally Posted by KyleAskine

I agree - we need to do away with work saved in this case.

I think it is a bit of a silly metric anyway - it incentivizes people to take stuff off the top of the range, which isn't really what we want anyway.

The real metric should be 'candidates released to Primenet' or something like that.

I don't entirely disagree -- and to be perfectly honest that metric was created in order to entice Xyzzy and Craig back to the LLTF range.

But I don't think the metric is entirely meaningless either, so long as it's used for candidates which are to be LL/DCed in the reasonable future. But if we were to offer and track LMH and/or 332M work, we would have to figure out some type of scaling factor. Perhaps something like Work Saved / Years Away.

[firejuggler]

M3321935017 takes 91 630.041 Ghz/day to LL once

M60082823 takes 133.476 Ghz/day to LL once ( exemple of reasonnable work -one month?)

ratio is 686.5
Supposing that Moore law keep working for as long as needed, ( double computer output each 2 year) that would put the reasonnable range in 18 to 20 years.
Also, one vote for the P-1 on DC

[kracker]

Originally Posted by firejuggler

Also, one vote for the P-1 on DC

+1

[chalsall]

Originally Posted by firejuggler

Also, one vote for the P-1 on DC

OK, this work type has now been made available from GPU72.

Although keep in mind it will take some time before there are many candidates which have had the required TFing done. As it happens, there were 26 which were already there.

[firejuggler]

As soon as I finish my 100~ 8.8M and my 9 56M P-1 i'll work on the DC P-1

[kracker]

Thanks! :)

http://gpu72.com/account/getassignments/
*clicky*
http://gpu72.com/account/getassignments/p-1/

[firejuggler]

nope^^ http://gpu72.com/account/getassignments/dcp-1/