GPU to 72 status...

[Dubslow]

Observation concerning why Spidey might have been assigned double checks before 45M:

Thresholds for first-time LL assignments
Force double-checks if CPU reliability less than	0.7
and CPU confidence level is greater than or equal to	2.0

So an LL doesn't necessarily need a nonzero error code to have a forced double check.

[ckdo]

Originally Posted by chalsall

Up to a maximum of 2000 assignments...

Bummer. 2000 assignments being only 7 days worth probably doesn't justify being exempted, I guess.

[chalsall]

Originally Posted by ckdo

Bummer. 2000 assignments being only 7 days worth probably doesn't justify being exempted, I guess.

On the other hand... 2000 assignments from 67 to 69, rather than only to 68, is 21 days worth.... (Hint, hint, hint... )

Again, I don't have a problem with anyone only going one bit level. But I don't think it's fair that only a few have the chance to do so....

[oswald]

Originally Posted by chalsall

Again, I don't have a problem with anyone only going one bit level. But I don't think it's fair that only a few have the chance to do so....

I can go

67 to 72
OR
70 to 72.

Which works out better for everyone?

[kladner]

My current practice is to try to get assignments as low-factored as possible, and take them to 72. That says nothing about what others do. Mostly this has meant doing 70-72 lately, and seeing very few factors found. :(

"Sometimes the magic works. Sometimes it doesn't."

EDIT: The 69-72 runs with stages disabled I have right now take about 4.5 hours. These are 56-58+M exponents. I'm running 2 at a time on a GTX 460. This seems to come out to about 10.5 completions per day.

[chalsall]

Originally Posted by kladner

My current practice is to try to get assignments as low-factored as possible, and take them to 72. That says nothing about what others do. Mostly this has meant doing 70-72 lately, and seeing very few factors found. :(

"Sometimes the magic works. Sometimes it doesn't."

Thank you for going to 72, and I'm sorry to hear that you are preceiving that you're not finding many factors. However, I did an analysis of your results vs the overall stats, and you're pretty close to what you should be seeing.

From the empirical data on the Trial Factoring Cost page, there is approximately a 1.03% chance of finding a factor going from 68 to 69, 1.01% from 69 to 70, 1.00% to 71, and 0.93% to 72. Note that this is a summary of results over the entire LL range.

Based on the work you've done, you should have found approximately 24.1 factors. You've actually found 23.

[James Heinrich]

Originally Posted by chalsall

Based on the work you've done, you should have found approximately 24.1 factors. You've actually found 23.

I'm sure that's a tidbit of interesting for many people -- could you put that on the user stats page (expected vs found factors; by bit range and overall)?

Of course, if you do that there's bound to be somebody complaining that they're missing some factors because they're below average.

[kladner]

Originally Posted by chalsall

Thank you for going to 72, and I'm sorry to hear that you are preceiving that you're not finding many factors. However, I did an analysis of your results vs the overall stats, and you're pretty close to what you should be seeing.

From the empirical data on the Trial Factoring Cost page, there is approximately a 1.03% chance of finding a factor going from 68 to 69, 1.01% from 69 to 70, 1.00% to 71, and 0.93% to 72. Note that this is a summary of results over the entire LL range.

Based on the work you've done, you should have found approximately 24.1 factors. You've actually found 23.

Thanks! It's good to know the odds. I should not make a fuss over it, anyway. I have had far more fruitful periods to arrive at those averages. That should suffice. It all serves the overall project.

(It did seem that the DC range had a lot more factors lurking. But that leaves aside the fact that I ran through a lot more exponents in ~67-69 range.)

[chalsall]

Originally Posted by James Heinrich

I'm sure that's a tidbit of interesting for many people -- could you put that on the user stats page (expected vs found factors; by bit range and overall)?

I knew I shouldn't have opened my mouth...

I would be hesitant to do that for each bit range for each user's individual attempts, because the statistical data is quite noisy. Please see a new report I just created, Factor Found Percentages to see what I mean. (And, yes, I'll do a similar report for P-1 percentages as well.)

However, I agree such statistics based on overall percentages for each bit level and meta range (DC and LL) would be interesting. Added to my To Do list... (Sigh... It seems the more work I do on the system, the longer (rather than shorter) the list becomes... )

[LaurV]

Originally Posted by ckdo

My "View Assignments" page states
You currently have 4568 Trial Factoring Assignments totaling 4110.091 GHz Days of work[...]
at the top vs.
4297 Assignments.
at the bottom. The latter is correct; I'm not sure about the GHzd total.

For me they NEVER matched, but they ALWAYS were correct. Both of them. Remark: the first number is only the TF assignments, but the last is the TOTAL. It is normal to be different (however, not normal that the first being bigger then the second). The first one is used to estimate the time needed to work them out. I think you should not change them.

[ckdo]

I had (and have) 38 non-TF assignments. The upper count simply wasn't updating any more.