Possible hardware errors...

[blahpy]

Originally Posted by SverreMunthe

It's on Worker 6 (thread 11 & 12)

Does your CPU actually have 12 cores, or are you trying to use both threads on 6 hyperthreaded cores? You may like to know that Prime95 will actually run slower if you do that (and use a lot more power, too).

[SverreMunthe]

I'm just running it out of the box, it has 6 cores, all threaded. So 6 workers and each worker is a core with two threads. It's an Intel 3960X.

[blahpy]

Originally Posted by SverreMunthe

I'm just running it out of the box, it has 6 cores, all threaded. So 6 workers and each worker is a core with two threads. It's an Intel 3960X.

You're better off setting each worker to one thread, then. It will actually work slightly faster AND generate less heat. One thread is enough to maximise usage of a core since the program is built very optimally. What I mean is the bottleneck is the CPU - not the threads.

[TheMawn]

I can confirm the hyperthreading generating more heat and operating slower.

My first gen i3 is about 10C warmer with two workers per core. Someone I know had an i7-920 running eight workers at ~44 ms per iteration, doing some double checks. Took off 5C and reduced to ~21 ms per iteration by running only the four workers instead.

[kracker]

My 3rd gen i3 has no effect at all with hyperthreading with P95, only 10C hotter... (fft times are same, evened out)

[tului]

Originally Posted by Prime95

The message you are seeing states that one hardware error was detected since the start of the test. You may not have an ongoing problem, but rather a rare problem, or even a one-time-only-stray-cosmic-ray problem.

See undoc.txt for options to control how often that message is output (or wait for the LL test to end).

Those darned antineutrinos!