V25.7 TF estimate way out ... or am I?

[petrw1]

From my current Worker Window.

[Nov 3 18:36] Worker starting
[Nov 3 18:36] Resuming trial factoring of M50324467 to 2^69
[Nov 3 18:36] Trial factoring M50324467 to 2^69 is 25.00% complete.
. . .
[Nov 6 20:37] Trial factoring M50324467 to 2^69 is 87.02% complete. Time: 3599.821 sec.
[Nov 6 21:37] Trial factoring M50324467 to 2^69 is 87.86% complete. Time: 3598.925 sec.
[Nov 6 22:37] Trial factoring M50324467 to 2^69 is 88.70% complete. Time: 3598.344 sec.

It's only a PIII 866 Mhz so it is doing 0.84% per hour.
At this rate I would estimate about 16 more hours.

However the status window gives a completion date of Tues Nov 11 at 7:30 AM.

I already checked my computer clock. It does say Nov 6 today as is evident in the last update above.

Am I reading this wrong OR is this a bug?

[James Heinrich]

Possibly the ETA is based on factoring up to 2^70:
11.3% (2^69) / 0.84%/hr = 13.45hr
100% (2^70) / 0.42%/hr = 238hr
251hrs = 10.5 days... hmm, no, that's not right either...

PrimeNet shows you have the exponent signed out for LL testing, not just TF... so your TF will have to complete up to whatever level (2^69?), then P-1, then LL... are you sure it's 11-Nov-2008? I wouldn't be surprised if P-1+LL could take ~370 days...

[petrw1]

Originally Posted by James Heinrich

Possibly the ETA is based on factoring up to 2^70:
11.3% (2^69) / 0.84%/hr = 13.45hr
100% (2^70) / 0.42%/hr = 238hr
251hrs = 10.5 days... hmm, no, that's not right either...

PrimeNet shows you have the exponent signed out for LL testing, not just TF... so your TF will have to complete up to whatever level (2^69?), then P-1, then LL... are you sure it's 11-Nov-2008? I wouldn't be surprised if P-1+LL could take ~370 days...

This is the line in my assignment report:

v4_computers 0 50324467  LL 0.00% 2008-10-27 17:27 11 2008-10-29 00:38 2008-10-30 00:38 2008-11-07 00:38 0

It says LL but then all assignments from v4 came across as LL for me, even the TF ones.

and it says 2008-11-07

The report that said Nov 11 is the Test ... Status ... dialog right from the PC itself.

[petrw1]

Well it finished today as I expected ... so I guess my own status report was wrong.

[starrynte]

sometimes i find prime95 has inaccurate estimates. for instance, once an exponent was estimated by me to finish at 9:10 on a certain day. it was 99% done. initially, prime95 estimated finishing in two days. when it was 99.8% done, it estimated 10:10. 99.9%, 9:30. and as it got closer to the completion it kept revising until it finished at 9:09 which was close to my original estimate...i'm also curious to how prime95 calculates completion dates, since some people like me don't leave their computer on 24/7

[Mini-Geek]

Originally Posted by starrynte

sometimes i find prime95 has inaccurate estimates. for instance, once an exponent was estimated by me to finish at 9:10 on a certain day. it was 99% done. initially, prime95 estimated finishing in two days. when it was 99.8% done, it estimated 10:10. 99.9%, 9:30. and as it got closer to the completion it kept revising until it finished at 9:09 which was close to my original estimate...i'm also curious to how prime95 calculates completion dates, since some people like me don't leave their computer on 24/7

Your observation and your question are more related than you probably realize. When Prime95 knows via the setting, or sees from its progress, that it's only on for, say, 16 hours a day, it adjusts the estimate by 16/24 to make it correct over a long period of time (i.e. it assumes you average 16/24th your max speed). As you noticed, this isn't accurate for the short term. There's no simple way to make it work better (although if you want you could set the hours to the normal amount all the time except that close to finishing when you put it to 24).