Prime95 version 25.7 - Page 12

jrk · 2008-10-27, 18:55

Quote:

Originally Posted by starrynte

which did not work for me for some reason (remove checked assignments was checked), thus I had to unreserve it from prime95. anyhow, all is working fine now

I just noticed the same problem.

Exponents assigned to the computer "v4_computers" can't be unassigned.

S485122 · 2008-10-28, 13:23

I am puzzled by the performance of the 64 bits version of prime95 25,7with regards to P-1 factoring :

- there are the maximum memory problems that have been reported previously by several users.

- I am under the impression that it is much slower than v24,14 in certain parts for instance

Code:

[Oct 28 12:25] M42832379 stage 2 is 99.89% complete. Time: 686.8 s.
[Oct 28 12:37] M42832379 stage 2 is 100.00% complete. Time: 676.0 s.
[Oct 28 12:48] M42832379 stage 2 is 100.00% complete. Time: 676.8 s.
[Oct 28 12:59] M42832379 stage 2 is 100.00% complete. Time: 691.5 s.
[Oct 28 13:08] M42832379 stage 2 complete. 1862456 transforms. Time: 96732.2 s.

why is it spending more than 40 mn at 100% before starting the GCD stage ? Why do the GCDs after stage 1 and 2 take more than an hour and a half where v24,14 used about 15 mn ?

- the slowing up of the other instances running on the processor (LL tests) during stage 2 is much bigger than in 24.14 (more than 20% compared to some 7,5 %.)

One problem existed previously and is perhaps not program related, but OS or hardware related : on a quadcore at the end of P-1 stage 2 factorisation, sometimes all processes are slowed down beyong recognition. Stopping or even quitting and restarting the program does not help in such cases, only a reboot will. The problem appears much less frequently when doing lots of very small P-1 factoring jobs (on 2M or 5M exponents f.i.), those can, most of the time be done in succession whithout affecting the work of other cores.

Is someone else experiencing the same kind of problems ?

Jacob

Batalov · 2008-10-28, 18:53

Quote:

Originally Posted by S485122

...why is it spending more than 40 mn at 100% before starting the GCD stage ? ...
Jacob

This bug is old - http://www.mersenneforum.org/showpos...&postcount=173

It only affects the progress screen report, not the quality of computation.
(I.e. it is not like 100.5% progress in block Lanczos - there it means serious problems with the computation itself.)

--S

Shinzok · 2008-10-28, 20:46

Well I don't think I deleted anything....
The only thing I did was just copy the prime95.exe(25.7 version) over the old prime95.exe(25.6 version)

So how would I be able to resume the previous numbers I was working on... (Or atleast send my work that I did so someone else would be able to continue or something)
I ask this because it's almost a month of work (12h day)

-Ehm I also had to remove the following line from prime.txt (hope I didn't break anything important now)
V24OptionsConverted=1

OneOfMany · 2008-10-28, 23:24

Anyone notice the day/night times working backwards (or not at all)? I just installed the it and setup the client to use 512MB of RAM at night...keeping the 8 for day time when I'm here.

I notice the log message said it was using 512MB (switched to 1024 when I changed the night setting to 1024).

Running v25.7 build 3 (that was what one of the download sites gave me when I clicked the Windows link) on Vista 32 bit, 2GB RAM, default settings for install.

starrynte · 2008-10-29, 01:45

factoring does not start a helper thread...is this correct?
if it is, then is it more efficient on a dual core machine to run one factoring assignment on both cores for 2 hours a day, or one factoring assignment on each core for
1 hour a day each?

Uncwilly · 2008-10-29, 04:48

Quote:

Originally Posted by starrynte

factoring does not start a helper thread...is this correct?
if it is, then is it more efficient on a dual core machine to run one factoring assignment on both cores for 2 hours a day, or one factoring assignment on each core for
1 hour a day each?

I don't see why you want to limit it to 2 total core hours. But each core will be about 10% longer, if they are run at the same time. Remember that Prime95 runs in the background and should not effect the perforence for most users. If you want abolute max output, not knowing what else you do on your system: 2 single core hours > 1 double core hour.

YMMV

sdbardwick · 2008-10-29, 07:40

FYI, Prime95 v25.7.3 (Win XP32 SP3) notes this as being a single core hyperthreaded processor, rather than as dual-core non-hyperthreaded. It is on a low-rent NVidia chipset board, so perhaps the motherboard BIOS is also misreporting the hyperthreading status.

Prime95 seems to work fine, after you make appropriate adjustments to the threading; 1 worker assigned to each logical processor without helpers. Iteration times are identical to the previous installation of 2 discrete 24.14 instances.

Mini-Geek · 2008-10-29, 11:37

Quote:

Originally Posted by starrynte

factoring does not start a helper thread...is this correct?
if it is, then is it more efficient on a dual core machine to run one factoring assignment on both cores for 2 hours a day, or one factoring assignment on each core for
1 hour a day each?

Do you mean one assignment on both cores for 1 real hour a day vs one on each core for 1 real hour a day, or for the first part is it one assignment on both cores for 2 real hours a day. If it's the former, as Uncwilly said, "2 single core hours > 1 double core hour". If it's the latter, it'd be "2 double core hours > 1 double core hour".

JHagerson · 2008-10-29, 12:56

I'm sorry if I missed it, but I can't find where to tell Prime 95 25.7 what kind of work I want to do. I had a bunch of LMH assignments to go one additional bit in factor depth. Suddenly, I have many fewer exponents and I have committed to factor them all the way to LL depth. In older versions, there was a series of check boxes that allowed me to specify the work I wanted.

If the answer is "turn off PrimeNet and continue to report manually," I wish that would have been stated somewhere before I made the changes I already made. Is the FactorOverride option ignored? It appears that the ending factor depth is now listed with each assignment in worktodo.txt.

Thank you for your help.

OneOfMany · 2008-10-29, 16:23

Yep. No change in memory usage (at least not mentioned in the logs) at the times listed by default in the day/night switcher.

Also after running over night, this morning I saw a few "other workers are using lots of memory now" and mem usage for Prime95 was jumping between 126MB and 50MB. Back and forth. I assume this isn't right...

2008-10-28, 13:23	#123
S485122 Sep 2006 Brussels, Belgium 2×3×281 Posts	Prime95 v25.7 64 bits and P-1 factoring I am puzzled by the performance of the 64 bits version of prime95 25,7with regards to P-1 factoring : - there are the maximum memory problems that have been reported previously by several users. - I am under the impression that it is much slower than v24,14 in certain parts for instance Code: [Oct 28 12:25] M42832379 stage 2 is 99.89% complete. Time: 686.8 s. [Oct 28 12:37] M42832379 stage 2 is 100.00% complete. Time: 676.0 s. [Oct 28 12:48] M42832379 stage 2 is 100.00% complete. Time: 676.8 s. [Oct 28 12:59] M42832379 stage 2 is 100.00% complete. Time: 691.5 s. [Oct 28 13:08] M42832379 stage 2 complete. 1862456 transforms. Time: 96732.2 s. why is it spending more than 40 mn at 100% before starting the GCD stage ? Why do the GCDs after stage 1 and 2 take more than an hour and a half where v24,14 used about 15 mn ? - the slowing up of the other instances running on the processor (LL tests) during stage 2 is much bigger than in 24.14 (more than 20% compared to some 7,5 %.) One problem existed previously and is perhaps not program related, but OS or hardware related : on a quadcore at the end of P-1 stage 2 factorisation, sometimes all processes are slowed down beyong recognition. Stopping or even quitting and restarting the program does not help in such cases, only a reboot will. The problem appears much less frequently when doing lots of very small P-1 factoring jobs (on 2M or 5M exponents f.i.), those can, most of the time be done in succession whithout affecting the work of other cores. Is someone else experiencing the same kind of problems ? Jacob Last fiddled with by S485122 on 2008-10-28 at 13:26 Reason: last sentence is a question, hence the question mark.

2008-10-28, 20:46	#125
Shinzok Oct 2008 Zevenbergen, NL 6₁₆ Posts	Well I don't think I deleted anything.... The only thing I did was just copy the prime95.exe(25.7 version) over the old prime95.exe(25.6 version) So how would I be able to resume the previous numbers I was working on... (Or atleast send my work that I did so someone else would be able to continue or something) I ask this because it's almost a month of work (12h day) -Ehm I also had to remove the following line from prime.txt (hope I didn't break anything important now) V24OptionsConverted=1 Last fiddled with by Shinzok on 2008-10-28 at 20:48 Reason: forgot to mention something that might be of importance

2008-10-29, 07:40	#129
sdbardwick Aug 2002 North San Diego County 5·137 Posts	Intel E2180 FYI, Prime95 v25.7.3 (Win XP32 SP3) notes this as being a single core hyperthreaded processor, rather than as dual-core non-hyperthreaded. It is on a low-rent NVidia chipset board, so perhaps the motherboard BIOS is also misreporting the hyperthreading status. Prime95 seems to work fine, after you make appropriate adjustments to the threading; 1 worker assigned to each logical processor without helpers. Iteration times are identical to the previous installation of 2 discrete 24.14 instances.

2008-10-29, 12:56	#131
JHagerson May 2005 Naperville, IL, USA 11000100₂ Posts	Specifying type of work to do... I'm sorry if I missed it, but I can't find where to tell Prime 95 25.7 what kind of work I want to do. I had a bunch of LMH assignments to go one additional bit in factor depth. Suddenly, I have many fewer exponents and I have committed to factor them all the way to LL depth. In older versions, there was a series of check boxes that allowed me to specify the work I wanted. If the answer is "turn off PrimeNet and continue to report manually," I wish that would have been stated somewhere before I made the changes I already made. Is the FactorOverride option ignored? It appears that the ending factor depth is now listed with each assignment in worktodo.txt. Thank you for your help.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Prime95 version 27.3	Prime95	Software	148	2012-03-18 19:24
Prime95 version 26.3	Prime95	Software	76	2010-12-11 00:11
Prime95 version 25.5	Prime95	PrimeNet	369	2008-02-26 05:21
Prime95 version 25.4	Prime95	PrimeNet	143	2007-09-24 21:01
When the next prime95 version ?	pacionet	Software	74	2006-12-07 20:30

2008-10-28, 23:24	#126
OneOfMany Sep 2008 5 Posts	Anyone notice the day/night times working backwards (or not at all)? I just installed the it and setup the client to use 512MB of RAM at night...keeping the 8 for day time when I'm here. I notice the log message said it was using 512MB (switched to 1024 when I changed the night setting to 1024). Running v25.7 build 3 (that was what one of the download sites gave me when I clicked the Windows link) on Vista 32 bit, 2GB RAM, default settings for install.

2008-10-29, 01:45	#127
starrynte Oct 2008 California 2²·59 Posts	factoring does not start a helper thread...is this correct? if it is, then is it more efficient on a dual core machine to run one factoring assignment on both cores for 2 hours a day, or one factoring assignment on each core for 1 hour a day each?

2008-10-29, 16:23	#132
OneOfMany Sep 2008 5 Posts	Yep. No change in memory usage (at least not mentioned in the logs) at the times listed by default in the day/night switcher. Also after running over night, this morning I saw a few "other workers are using lots of memory now" and mem usage for Prime95 was jumping between 126MB and 50MB. Back and forth. I assume this isn't right...