Prime95 version 26.6

[James Heinrich]

Originally Posted by Jwb52z

I don't think that any laptop on the market can have 16GB of RAM yet, so that's not even an option for me. I'll try both TF and P-1 in a few months.

Somewhat off-topic, but there are a number of "gaming"-targeted laptops that ship with 16GB of RAM, such as these examples from Asus and MSI. There's even some exotics (from companies like Eurocom) that offer 24GB models.

[Jwb52z]

Originally Posted by James Heinrich

Somewhat off-topic, but there are a number of "gaming"-targeted laptops that ship with 16GB of RAM, such as these examples from Asus and MSI. There's even some exotics (from companies like Eurocom) that offer 24GB models.

I don't want to get us off topic any further, so I'll just say "Wow" and move on from here.

[Bdot]

For P-1, I somehow think there was an option to save intermediate results in a way that you can re-run the same P-1 with higher B1 and B2, without the need to re-evaluate up to the previous B1 and B2. But I cannot find that in undoc.txt, or *txt ... is there a way to do that?

[KingKurly]

Originally Posted by Bdot

For P-1, I somehow think there was an option to save intermediate results in a way that you can re-run the same P-1 with higher B1 and B2, without the need to re-evaluate up to the previous B1 and B2. But I cannot find that in undoc.txt, or *txt ... is there a way to do that?

undoc.txt-By default P-1 work does not delete the save files when the work unit completes.
undoc.txt-This lets you run P-1 to a higher bound at a later date. You can force
undoc.txt-the program to delete save files by adding this line to prime.txt:
undoc.txt: KeepPminus1SaveFiles=0

So this would imply to me that KeepPminus1SaveFiles = 1 would be what you want.

[Bdot]

Originally Posted by KingKurly

So this would imply to me that KeepPminus1SaveFiles = 1 would be what you want.

Thanks KingKurly - I knew I had read that on some box. And I tricked myself: On my Dev-Box I usually only replace prime95.exe leaving all txt files untouched. No wonder my old undoc.txt did not have that KeepPminus1SaveFiles ...

Enhancement request for prime95: Add the version number to the txt files where possible/applicable. (I'll take that note for mfakto too ).

[BigBrother]

If I do P-1 factoring on a 1.5M number on my Core 2 Duo, it says:

Using Core2 type-3 FFT length 80K, Pass1=320, Pass2=256

When I run a number from this range on my i7, it says:

Using Pentium4 type-3 FFT length 80K, Pass1=320, Pass2=256

Is this because the Pentium4 code on 'small' FFTs is faster than Core2 code for an i7? My i7 is correctly identified by Prime95 and uses Core2 FFT code for LLs in the 54M range.

[Prime95]

Yes, the Pentium4 version benchmarked faster than the Core2 version on an i7

Is it normal to see periodic CPU temp variations with the In-Place torture test? I'm using a Intel i5-2500k.

[Dubslow]

I remember this has been discussed elsewhere, but I wanted to confirm that 26.6 does horribly mis-detect the clock speed of the Phenom X6 1055T Thuban. That runs at 2.8 GHz; it reported in the high 6.x GHz. No clue why.

[kladner]

Originally Posted by Dubslow

I remember this has been discussed elsewhere, but I wanted to confirm that 26.6 does horribly mis-detect the clock speed of the Phenom X6 1055T Thuban. That runs at 2.8 GHz; it reported in the high 6.x GHz. No clue why.

My 1090T @ 3.5GHz currently reports 10329.02MHz in 26.6.

[Bdot]

Despite my best efforts, mprime keeps unreserving my GPU-to-72 assignments.

prime.txt contains

UnreserveDays=365
MaxExponents=50
DaysOfWork=1
DaysBetweenCheckins=7

Still I see this during startup:

[Main thread Jan 11 13:45] Mersenne number primality test program version 26.5
[Main thread Jan 11 13:45] Optimizing for CPU architecture: Core i3/i5/i7, L2 cache size: 256 KB, L3 cache size: 12 MB
[Main thread Jan 11 13:45] Unable to detect some of the hyperthreaded logical CPUs.
[Main thread Jan 11 13:45] See AffinityScramble2 in undoc.txt.
[Main thread Jan 11 13:45] Starting workers.
[Comm thread Jan 11 13:45] Unreserving M45355897
[Worker #3 Jan 11 13:45] Waiting 10 seconds to stagger worker starts.
[Worker #2 Jan 11 13:45] Waiting 5 seconds to stagger worker starts.
[Worker #7 Jan 11 13:45] Waiting 30 seconds to stagger worker starts.
[Worker #5 Jan 11 13:45] Waiting 20 seconds to stagger worker starts.
[Worker #6 Jan 11 13:45] Waiting 25 seconds to stagger worker starts.
[Worker #1 Jan 11 13:45] Worker starting
[Worker #1 Jan 11 13:45] Setting affinity to run worker on logical CPU #0
[Worker #8 Jan 11 13:45] Waiting 35 seconds to stagger worker starts.
[Worker #4 Jan 11 13:45] Waiting 15 seconds to stagger worker starts.
[Comm thread Jan 11 13:45] Unreserving M45355487

My worktodo now has 30 exponents (including 10 P-1 ones).
This leaves one worker with only the current 45M exponent, which is already 65% complete,~ 9 days to go.

How can I prevent this unreserving?

I had added stuff via worktodo.add, but did not use the worker numbers as a had intended. As the resulting worktodo.txt was not to my liking, I stopped mprime, fixed worktodo.txt and restarted mprime. That's when it unreserved some of the new expos. Is it possible that mprime discovered that my wrong adding of work exceeded some limit (but no way 365 days, it was 5 expos at most for one worker). But it did not immediately unreserve it but saved that for the next restart/reporting in? Even though the work has been rebalanced in between?

I'd like a switch to completely disable unreserving ...