TortureHyperthreading, does it help for stability testing?

sp00n · 2022-09-10, 11:45

Is there any meaningful difference for a torture test regarding stability testing if you activate TortureHyperthreading in prime.txt or not?

I.e. for single core stability testing with two threads, and manually setting the affinity to one specific core, is there a difference if you let Prime95 run with

Code:

prime.txt
----------
TortureHyperthreading=1


local.txt
---------
NumThreads=1
WorkerThreads=1

vs.

Code:

prime.txt
----------
TortureHyperthreading=0


local.txt
---------
NumThreads=2
WorkerThreads=2

In both cases one core is fully utilized with two threads. The second option is the "old" behavior until up to version 30.6, in 30.7 the new TortureHyperthreading setting was introduced. So I wonder, does it somehow affect the outcome of the stability test (better? harder?)?

As far as I can see, when the setting is activated Prime95 uses two distinct threads for anything up to an FFT size of 240k, and for 256k and up it uses one thread spread across two cores.

Unfortunately this introduces inconsistencies in the log file, where for the first part, it writes two log entries for each FFT size, and for the second part only one line, which makes parsing the log file rather cumbersome:

Code:

[Thu Sep  8 16:51:56 2022]
Self-test 224K (thread 2 of 2) passed!
Self-test 224K (thread 1 of 2) passed!
Self-test 240K (thread 2 of 2) passed!
Self-test 240K (thread 1 of 2) passed!
Self-test 256K passed!
[Thu Sep  8 16:53:11 2022]
Self-test 288K passed!
Self-test 320K passed!
Self-test 336K passed!
Self-test 384K passed!

Which is why I reverted back to the "old" mechanism for now for CoreCycler, but if TortureHyperthreading does offer a valuable improvement over the previous implementation, I would adjust my log parser for it.

Prime95 · 2022-09-10, 15:14

If you are using OS tools to restrict prime95 to run on one specific core, then your two examples are equivalent.

sp00n · 2022-09-10, 15:25

Great, thank you.
Yes, the tool automatically assigns the affinity of Prime95 to a single or both virtual cores of a physical core to test the automatic overclocking functionality of modern CPUs like Ryzen (PBO / Curve Optimizer).

2022-09-10, 11:45	#1
sp00n Dec 2020 2×5 Posts	TortureHyperthreading, does it help for stability testing? Is there any meaningful difference for a torture test regarding stability testing if you activate TortureHyperthreading in prime.txt or not? I.e. for single core stability testing with two threads, and manually setting the affinity to one specific core, is there a difference if you let Prime95 run with Code: prime.txt ---------- TortureHyperthreading=1 local.txt --------- NumThreads=1 WorkerThreads=1 vs. Code: prime.txt ---------- TortureHyperthreading=0 local.txt --------- NumThreads=2 WorkerThreads=2 In both cases one core is fully utilized with two threads. The second option is the "old" behavior until up to version 30.6, in 30.7 the new TortureHyperthreading setting was introduced. So I wonder, does it somehow affect the outcome of the stability test (better? harder?)? As far as I can see, when the setting is activated Prime95 uses two distinct threads for anything up to an FFT size of 240k, and for 256k and up it uses one thread spread across two cores. Unfortunately this introduces inconsistencies in the log file, where for the first part, it writes two log entries for each FFT size, and for the second part only one line, which makes parsing the log file rather cumbersome: Code: [Thu Sep 8 16:51:56 2022] Self-test 224K (thread 2 of 2) passed! Self-test 224K (thread 1 of 2) passed! Self-test 240K (thread 2 of 2) passed! Self-test 240K (thread 1 of 2) passed! Self-test 256K passed! [Thu Sep 8 16:53:11 2022] Self-test 288K passed! Self-test 320K passed! Self-test 336K passed! Self-test 384K passed! Which is why I reverted back to the "old" mechanism for now for CoreCycler, but if TortureHyperthreading does offer a valuable improvement over the previous implementation, I would adjust my log parser for it.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Best way to test CPU and RAM stability?	Caribou007	Software	3	2013-07-12 05:50
Test GPU stability	TimSorbet	GPU Computing	25	2013-03-15 02:37
AMD Core unlock, stability?	nuggetprime	Hardware	0	2011-03-05 19:07
Remote Stability Test?	spaz	Software	3	2009-12-10 15:16
stability of system?	aspicot	Hardware	3	2003-03-18 16:12

2022-09-10, 15:14	#2
Prime95 P90 years forever! Aug 2002 Yeehaw, FL 17·487 Posts	If you are using OS tools to restrict prime95 to run on one specific core, then your two examples are equivalent.

2022-09-10, 15:25	#3
sp00n Dec 2020 A₁₆ Posts	Great, thank you. Yes, the tool automatically assigns the affinity of Prime95 to a single or both virtual cores of a physical core to test the automatic overclocking functionality of modern CPUs like Ryzen (PBO / Curve Optimizer).