Benchmark output Prime95

[lukerichards]

Hi,

Is there anything I can add to prime.txt to turn off benchmark output from the worker window or comm window of Prime95?

I'm running a lot of smallish (7-100hour) PRP tests and the multiple line output from the benchmarks is getting in the way a bit.

Have checked the undoc file, the only thing I can find looks like:

AffinityVerbosityBench=0

But I didn't know for sure if that's what I'm after.

[Prime95]

Originally Posted by lukerichards

Is there anything I can add to prime.txt to turn off benchmark output from the worker window or comm window of Prime95?

No. You can stop automatic benchmarking with AutoBench=0

[lukerichards]

Presumably no benchmarking -> lower performance?

[Prime95]

Originally Posted by lukerichards

Presumably no benchmarking -> lower performance?

Not necessarily. If there is an FFT implementation that is faster than the default FFT implementation -- then yes.

Two more items you can tweak:

	autobench_days_of_work = (double) IniGetInt (INI_FILE, "AutoBenchDaysOfWork", 7);
	autobench_num_benchmarks = IniGetInt (INI_FILE, "AutoBenchNumBenchmarks", 10);

The first controls how "deep" autobench looks into worktodo.txt.
The second controls how many benchmarks are accumulated for each FFT size. The default is 10 to eliminate the impact of other programs running during a benchmark and also the run-to-run variations that always occur when benchmarking. Set it to somewhere in the 2 to 5 range.

Also, you could look at your results in results.txt or results.json.txt for your results without benchmark output.

[lukerichards]

Originally Posted by Prime95

Also, you could look at your results in results.txt or results.json.txt for your results without benchmark output.

I do get the benchmark output saved to results.txt too - but I think I *can* turn that off?

[Prime95]

Originally Posted by lukerichards

I do get the benchmark output saved to results.txt too - but I think I *can* turn that off?

In 29.8?

[lukerichards]

Originally Posted by Prime95

In 29.8?

I could have sworn I'd updated, but your question prompted me to check and, no, it's 29.4.

[ixfd64]

Originally Posted by Prime95

Not necessarily. If there is an FFT implementation that is faster than the default FFT implementation -- then yes.

Two more items you can tweak:

	autobench_days_of_work = (double) IniGetInt (INI_FILE, "AutoBenchDaysOfWork", 7);
	autobench_num_benchmarks = IniGetInt (INI_FILE, "AutoBenchNumBenchmarks", 10);

The first controls how "deep" autobench looks into worktodo.txt.
The second controls how many benchmarks are accumulated for each FFT size. The default is 10 to eliminate the impact of other programs running during a benchmark and also the run-to-run variations that always occur when benchmarking. Set it to somewhere in the 2 to 5 range.

Also, you could look at your results in results.txt or results.json.txt for your results without benchmark output.

So there are undoc.txt features, and then there are undocumented features.