V24.12 Release 1 Crashing on P4 M

[njcroquet1]

I updated all of my desktop P4's the other day to V24.12 Rel 1 and have seen 6-8% improvements on all machines over V24.6 or 23.8. However, my P4 M laptop, after running for 3 or 4 days at .083 (it was .093 with V24.6) will only run for 3-4 minutes before shutting down by itself.
Pausing Prime95 at bootup stops the problem and the machine seems fine.
I have temporarily gone back to V24.6 until I figure this out.

Machine specs: Latitude 840, P4 M @2.2GHz, 512 Mb RAM, Win XP Pro

Anyone have any ideas?

Thanks.......Greg

[Prime95]

Can you give more info? Any error messages on screen? in results.txt? in prime.log? If you turn on manual communication does it work (but fail when you force a communication to primenet?) Were you TFing, P-1, or LL testing? What exponent?

[Dresdenboy]

Could that be caused by the temperatures (both ambient and inside the laptop)? A P4 M is not a P-M. And it dissipates more heat. Switching to the newer client (and maybe a higher ambient temperature) could just increase the amount of heat to a level, where the mainboard decides to switch off the computer. Lower iteration times also mean, that the same work is done in less time.

[Cruelty]

I guess that any laptop (especially those equiped with P4)should be ready for high temperatures, and should not shutdown - besides, P4 utilize "thermal throttling" which should help lower the temperature if it becomes critical...

[njcroquet1]

I am currently running LL on M34563479 (at iteration 29725000). No screen messages were visible as the screen just goes black and shutdown occurs.

The last entry in results.txt is from the benchmarks I ran when I first upgraded to V24.12. As you can see, it does not recognize the correct CPU speed:

Mobile Intel(R) Pentium(R) 4 - M CPU 2.20GHz
CPU speed: 1196.45 MHz
CPU features: RDTSC, CMOV, Prefetch, MMX, SSE, SSE2
L1 cache size: 8 KB
L2 cache size: 512 KB
L1 cache line size: 64 bytes
L2 cache line size: 128 bytes
TLBS: 64
Prime95 32-bit version 24.12, RdtscTiming=1
Best time for 512K FFT length: 22.444 ms.
Best time for 640K FFT length: 29.443 ms.
Best time for 768K FFT length: 35.707 ms.
Best time for 896K FFT length: 41.630 ms.
Best time for 1024K FFT length: 47.098 ms.
Best time for 1280K FFT length: 59.351 ms.
Best time for 1536K FFT length: 172.016 ms.
Best time for 1792K FFT length: 202.530 ms.
Best time for 2048K FFT length: 226.745 ms.
Best time for 2560K FFT length: 295.806 ms.
Best time for 3072K FFT length: 375.575 ms.
Best time for 3584K FFT length: 444.537 ms.
Best time for 4096K FFT length: 504.569 ms.
Best time for 58 bit trial factors: 35.712 ms.
Best time for 59 bit trial factors: 36.273 ms.
Best time for 60 bit trial factors: 35.650 ms.
Best time for 61 bit trial factors: 35.692 ms.
Best time for 62 bit trial factors: 38.183 ms.
Best time for 63 bit trial factors: 38.353 ms.
Best time for 64 bit trial factors: 46.937 ms.
Best time for 65 bit trial factors: 46.852 ms.
Best time for 66 bit trial factors: 46.869 ms.
Best time for 67 bit trial factors: 47.012 ms.

The last entry in Prime.log is:

[Thu Jun 23 18:40:16 2005 - ver 24.12]
Updating computer information on the server
Sending expected completion date for M34563479: Jun 30 2005
Sending expected completion date for M34639027: Aug 11 2005

V24.6 has been running fine for 12 hours now at .092.

[Cruelty]

Originally Posted by njcroquet1

The last entry in results.txt is from the benchmarks I ran when I first upgraded to V24.12. As you can see, it does not recognize the correct CPU speed:

Mobile Intel(R) Pentium(R) 4 - M CPU 2.20GHz
CPU speed: 1196.45 MHz

This is mobile CPU, with "speed-step" technology, therefore it can lower it's operating frequency according to predefined profiles (especially when running on battery).

[njcroquet1]

This machine always runs on AC power.

I did find an interesting event log item under Administrative Tools....see attached zip file.

I have not yet tried the forced manual communication as the server seems to be down.

[Dresdenboy]

Originally Posted by Cruelty

I guess that any laptop (especially those equiped with P4)should be ready for high temperatures, and should not shutdown - besides, P4 utilize "thermal throttling" which should help lower the temperature if it becomes critical...

Dust or just some old components in a laptop (which actually is not designed for HPC) can change this situation quite a bit. We also had bad experiences with Dell laptops running Prime95. They seem to be just designed for the typical TDP mentioned by Intel, which doesn't include apps causing a higher power consumption. Throttling can only mask CPU cycles up to a certain level.

OTOH their are other components known to be the cause for failures.

[Cruelty]

Originally Posted by njcroquet1

This machine always runs on AC power.

I did find an interesting event log item under Administrative Tools....see attached zip file.

I also get such an event log error when Prime95 encounters a "possible hardware failure" problem, so in my opinion it says nothing more than that...