Prime95 30.8 (big P-1 changes, see post #551)

[kriesel]

Now that v30.8b15 mprime / prime95 have been promoted to general availability at https://www.mersenne.org/download/, "beta" in the thread title no longer accurately reflects the status.

[Kaboom]

Originally Posted by LOBES

Long story short, my clients, on various machines, no longer send requests for CERT work. I am using Linux64,Prime95,v30.8,build 15.

I, too, haven't got any CERT work since upgrading to 30.8b15.

In my configuration I have the following lines related to CERT work:

$ grep Cert /var/lib/gimps/local.txt 
CertDailyRemainingLastUpdate=1654697164
CertDailyMBRemaining=40
CertDailyCPURemaining=10
CertWork=1
CertDailyCPULimit=10

[LOBES]

Originally Posted by Kaboom

I, too, haven't got any CERT work since upgrading to 30.8b15.

In my configuration I have the following lines related to CERT work:

$ grep Cert /var/lib/gimps/local.txt 
CertDailyRemainingLastUpdate=1654697164
CertDailyMBRemaining=40
CertDailyCPURemaining=10
CertWork=1
CertDailyCPULimit=10

Can you run a simple

./mprime -m

and see if it spits out Comm logs trying to get a cert?

[pepi37]

Since it is hot wave here I decide to run CRUS but on my Ryzen 7 3700x

AMD Ryzen 7 3700X 8-Core Processor
CPU speed: 4000.03 MHz, 8 cores
CPU features: 3DNow! Prefetch, SSE, SSE2, SSE4, AVX, AVX2, FMA
L1 cache size: 8x32 KB, L2 cache size: 8x512 KB, L3 cache size: 2x16 MB
L1 cache line size: 64 bytes, L2 cache line size: 64 bytes

I never has any problem with this CPU and he has done a lot of job: even my last own on base2 ( so I have Gerbicz test) to check it. I know no test is perfect, but still. No error at all.


Now I set 2 workers , 4 cores per test ( on base 395. k4) Every time worker 2 gives ok results, and every time worker 1 gives some kind of error.
But when I see this in log I stop all

[Worker #1 Jul 6 15:34:58] Possible hardware errors have occurred during the test! 4 ROUNDOFF > 0.4.
[Worker #1 Jul 6 15:34:58] Confidence in final result is poor.

CPU TEMP doesnot exceed 70°C, it is slightly OC ( from 3.8 to 4 Ghz) works as I say for years without problem.
Now I recheck all results (106 of them) from this batch on two other CPU to verify results and so far even "poor" residue match.


Log is attached (ziped)


Any idea?

[Prime95]

Originally Posted by pepi37

Any idea?

Just a guess. The errors are hardware related (the round off error is not barely over 0.4, it is way over 0.4).
Perhaps, 1 worker fits completely in the L3 cache and 2 workers does not. This would make 1 worker faster than 2 workers and thus more stressful on the FPU/CPU/caches.

[pepi37]

576k candidate x2 workers should or not fit in 1x16 MB of L3 cache, my cpu has 2 x 16MB
Since this CPU have one CCX with eight cores , I assume first worker will put candidate in first pat of L3 cache , and second worker will put in second part of L3 cache.
Now I rerun same tests on him also but now in config 4 workers ( 2 cores per workers) and it works perfectly.

I must do additional test to see can I use 2*4 combination on higher FFT size or I will be forced to use 1*8 combination on higher FFT candidate.

[Kaboom]

Originally Posted by LOBES

Can you run a simple

./mprime -m

and see if it spits out Comm logs trying to get a cert?

I set out to do that, with no results at first, but then, somehow, after changing log verbosity, it got a CERT assignment. No idea how I did it, though.

[M0CZY]

I've updated my old 32-bit Dell laptop running Windows 7 Enterprise to p95v308b15, but the program doesn't work. It won't start up; nothing happens.
Can you confirm that Windows 7 32-bit is still supported? The older p95v307b9, which I still have works fine.

[Prime95]

Originally Posted by M0CZY

I've updated my old 32-bit Dell laptop running Windows 7 Enterprise to p95v308b15, but the program doesn't work. It won't start up; nothing happens.
Can you confirm that Windows 7 32-bit is still supported? The older p95v307b9, which I still have works fine.

I can confirm it is *supposed* to work. I do not have a 32-bit Windows install to test on. Has anyone else tried 30.8b15 on Win32?

[S485122]

Originally Posted by M0CZY

I've updated my old 32-bit Dell laptop running Windows 7 Enterprise to p95v308b15, but the program doesn't work. It won't start up; nothing happens.
...

First of all verify in TaskManager that the progam is not running windowless.
Then open a command prompt window, navigate to the Prime95 folder, then launch Prime95. Maybe you will see a message...
If none of this gives any clue, you could try to look in the EventViewer...

[Kaboom]

Originally Posted by Kaboom

I set out to do that, with no results at first, but then, somehow, after changing log verbosity, it got a CERT assignment. No idea how I did it, though.

...and then nothing more.