Prime95 v30.4/30.5/30.6

[Prime95]

Originally Posted by lycorn

CPU: Intel i5-7400 (kaby Lake) @ 3GHz. 16 GB DDR4 2400 memory (dual channel - 2 x 8GB).

What are your day and night memory settings?

[lycorn]

Originally Posted by Prime95

What are your day and night memory settings?

12 GB (for both day and night).

Prime95, as far as I could see, don´t usually goes anywhere close to these values, though.
I observed the memory usage for some time noting that, for this size of exponents, the allocation was 4921 MB for the first and second worker entering stage 2 and 2247 MB for the third. When the fourth joined, the first and second were restarted with 3304 and 3227 MB.These values were kept for the rest of the 4 runs The total mem used would amount to 12,216 MB, slightly over the allowed value, but the usage reported by Task Manager never went above ~9.6 GB. This is a recurrent behaviour, I have observed it more than once. Although the values might fluctuate somehow (for example, the first worker entering S2 may restart twice, from 4921 MB to 3227 MB then 2247 MB), the general pattern holds.
But I have also observed Prime95 using 11.5GB total, for a brief period, with all workers in S2 That´s probably when there are several workers performing the modular inverse at the same time. So, all in all, I guess Prime95 is working as expected.
I´ll leave as it is for a couple of days to see how it does, and if another crash happens I´ll reduce the memory setting to 10 GB to see if it helps.
Unless, of course, you have some other suggestion or ask me to try something else.

[lycorn]

Originally Posted by LaurV

How much are the temperatures of the toy? (that's the most important info, which I didn't see in the post).

You didn´t see it in the post for a good reason. Temps are stable at 60 - 62 celsius, under full load, which totally harmless. I didn´t even bother to mention that.

Originally Posted by LaurV

The machine is old, so it may need some maintenance, you know, removing the dust clogs from the fans, re-seating of the CPU (change/reapply the thermal paste), etc. We do this yearly, or even every 6 months or so. You know, my grandma was virgin for a very long time, but suddenly she wasn't. Luckily for me, otherwise I won't be anymore, and who would post stupid things on mersenneforum? Haha.

The machine is two years old. I never saw the need to reseat the CPU and/or reapply new paste on a 2-year old machine. The heatsink is not clogged with dust either. Probably I´ll clean it this summer. Anyhow, the machine is certainly a lot younger than your grandma, so I reckon it´s still virgin...

Originally Posted by LaurV

If you do only ECM, it won't matter much anyhow.

Yes, that´s the case, for the foreseeable future.

[tha]

I put this in worktodo.txt:

[Worker #1]
Pminus1=N/A,1,2,15510977,-1,1500000,30000000
Pminus1=N/A,1,2,15511381,-1,1500000
Pminus1=N/A,1,2,15511403,-1,1500000,30000000

Which resulted in this output:

Your choice: [Main thread Jan 6 20:13] Starting worker.
[Work thread Jan 6 20:13] Worker starting
[Work thread Jan 6 20:13] Setting affinity to run worker on CPU core #1
[Work thread Jan 6 20:13] 
[Work thread Jan 6 20:13] P-1 on M15510977 with B1=1500000, B2=30000000
[Work thread Jan 6 20:13] Setting affinity to run helper thread 1 on CPU core #2
[Work thread Jan 6 20:13] Setting affinity to run helper thread 2 on CPU core #3
[Work thread Jan 6 20:13] Setting affinity to run helper thread 3 on CPU core #4
[Work thread Jan 6 20:13] Using FMA3 FFT length 800K, Pass1=320, Pass2=2560, clm=4, 4 threads
[Work thread Jan 6 20:13] Using 11057MB of memory.
[Work thread Jan 6 20:14] Stage 2 init complete. 110047 transforms. Time: 60.167 sec.
[Work thread Jan 6 20:14] M15510977 stage 2 is 43.88% complete.
[Work thread Jan 6 20:18] M15510977 stage 2 is 65.04% complete. Time: 265.576 sec.
[Work thread Jan 6 20:22] M15510977 stage 2 is 85.40% complete. Time: 255.297 sec.
[Work thread Jan 6 20:26] M15510977 stage 2 complete. 1358614 transforms. Time: 715.411 sec.
[Work thread Jan 6 20:26] Stage 2 GCD complete. Time: 2.544 sec.
[Work thread Jan 6 20:26] M15510977 completed P-1, B1=1500000, B2=30000000, E=12, Wi8: 9088F10D
[Comm thread Jan 6 20:26] Sending result to server: UID: Tha/Z-170, M15510977 completed P-1, B1=1500000, B2=30000000, E=12, Wi8: 9088F10D
[Comm thread Jan 6 20:26] 
[Work thread Jan 6 20:26] 
[Work thread Jan 6 20:26] Using minimum bound #1 of 30
[Work thread Jan 6 20:26] P-1 on M1 with B1=30, B2=15511381
[Work thread Jan 6 20:26] Cannot initialize FFT code, errcode=1008
[Work thread Jan 6 20:26] Worker stopped.
[Main thread Jan 6 20:26] Execution halted.
[Main thread Jan 6 20:26] Choose Test/Continue to restart.
[Comm thread Jan 6 20:26] PrimeNet success code with additional info:
[Comm thread Jan 6 20:26] CPU credit is 2.2200 GHz-days.
[Comm thread Jan 6 20:26] Done communicating with server.

It also wiped out most of the worktodo.txt file! It put this in there:

Pminus1=N/A,1,2,1,-1,2,15511381,1500000

[petrw1]

And saw this in Status.

Maybe it can't handle having no B2?

[ET_]

I started a ECM range on F19 with mprime v.29.8. Restarted it on stage 1, with v30.4b5, 14GB of RAM assigned to each of the 2 workers, and got this message:

ECM on F19: curve #1 with s=7741023751586926, B1=44000000, B2=TBD
...
...
[Worker #1 Jan 7 00:52] Stage 1 complete. 348387014 transforms, 1 modular inverses. Time: 0.088 sec.
[Worker #1 Jan 7 00:52] Optimal B2 is 189*B1 = 8316000000.
[Worker #1 Jan 7 00:52] D: 6930, relative primes: 17543, stage 2 primes: 41752603, pair%=92.77

Now I am happily crunching on Stage2

BTW, switching from Stage1 to Stage2 of F19 I had theis strange behavior:

[axn]

Build 5 also suffers from Infinite loop bug. This is linux, btw.

The setup is:
Memory = 57g
Each of the 6 worker is restricted to 9.5g. mprime however uses 9966m actual per worker (instead of the 9728m allowed) . Somehow when all 6 workers entered stage 2 together, I got the infinite loop for one of the worker (it just loops printing "Available memory is xxxx").

Had to quit and change total memory to 9966*6= 59796m (without changing per-worker allocation). Now seems to be working fine. Let's see if this recurs.

[petrw1]

After a week running 30.4 b3 without issue it "disappeared" this morning.

I just upgraded to this b5. Keep you posted.

[petrw1]

Originally Posted by petrw1

After a week running 30.4 b3 without issue it "disappeared" this morning.

I just upgraded to this b5. Keep you posted.

On restart it backed up and redid some work and found the same factor for the same exponent LOL.
Second was not needed of course.

[lycorn]

Originally Posted by petrw1

After a week running 30.4 b3 without issue it "disappeared" this morning.

I just upgraded to this b5. Keep you posted.

Welcome to the club. See posts #3 and 19 on this thread.

[garo]

This is a tricky one for you George. I am running build 30.3b2 but no reason why the same issue (won't call it a bug) doesn't exist in this version.


Kaspersky was bugging me so I said yes to some security fixes it wanted to do. I'm running Prime95 from the Program Files folder in Windows 7. Always have. I know it's frowned upon now but old habits die hard. So Prime95 stopped being able to write to the folder because Kaspersky had messed wth the ownership. I'm also running as admin so basically security nightmare. So far so good. But now when Prime95 runs the Jacobi benchmarks it is unable to write a checkpoint so forgets about progress and restarts from the last checkpoint written on disk. This seems undesirable in my opinion. If for whatever reason the disk becomes temporarily unavailable, running the Jacobi benchmarks can wipe all progress.