Disable Hyperthreading on x570 TUF

[MHammett]

How do I disable hyperthreading on my x570 TUF motherboard?

[RedSector73]

MHammett wrote:
How do I disable hyperthreading on my x570 TUF motherboard?

You can't, the big problem is board does not support an Intel CPU or Intel's gift that does not stop giving 'hyperthreading'.
If you mean Simultaneous Multi Threading (SMT) ... well now you know what to look for.

Given your failure to know what to call it, I also doubt your logic behind disabling it, can you enlight us... Why would you want to disable this ?

[MHammett]

RedSector73 wrote:
You can't, the big problem is board does not support an Intel CPU or Intel's gift that does not stop giving 'hyperthreading'.
If you mean Simultaneous Multi Threading (SMT) ... well now you know what to look for.

Given your failure to know what to call it, I also doubt your logic behind disabling it, can you enlight us... Why would you want to disable this ?

Sorry for using the wrong term. It seems as that particular detail may not matter as "Intel branded this process as hyper-threading, but hyper-threading is the same thing as simultaneous multithreading." https://www.tomshardware.com/reviews/simultaneous-multithreading-definition,5762.html

I have a single-threaded process as a part of my workflow that will use 8% of my CPU (6 core, 12 threads) for long periods of time. I wanted to see if disabling hyperthreading (err SMT) would positively affect execution time of that task.

[MHammett]

I have found SMT in the manual, so when I get a chance, I'll conduct an operation in the current configuration, disable SMT, then conduct the operation again and see how the time differs.

[RedSector73]

MHammett wrote:
I have found SMT in the manual, so when I get a chance, I'll conduct an operation in the current configuration, disable SMT, then conduct the operation again and see how the time differs.

My gut feeling is it won't change significantly. I would be interested in the results either way.

[MHammett]

RedSector73 wrote:
My gut feeling is it won't change significantly. I would be interested in the results either way.

My first test had 5:08 with SMT on and 4:32 with SMT off. That is indeed faster, but not twice as fast as one would expect.

I'll have to run a few more of these with a bigger load (so that it magnifies the differences) and see what I get.

[RedSector73]

MHammett wrote:
My first test had 5:08 with SMT on and 4:32 with SMT off. That is indeed faster, but not twice as fast as one would expect.

I'll have to run a few more of these with a bigger load (so that it magnifies the differences) and see what I get.

Thats about 11.7% increase in speed which is not to be ignored either.

[MHammett]

RedSector73 wrote:
Thats about 11.7% increase in speed which is not to be ignored either.

Agreed. It's just hard to reconcile 11% when it should have been 100%. Something else must be afoot.

[RedSector73]

MHammett wrote:
Agreed. It's just hard to reconcile 11% when it should have been 100%. Something else must be afoot.

From what I remember with my Intel 980x, I struggle to accept there would be a 100% increase in intel chip with hyperthreading off.
Have you considered trying one CCX instead of two for testing purposes and SMT on/off with that configuration ?

[XIIIIX]

RedSector73 wrote:
You can't, the big problem is board does not support an Intel CPU or Intel's gift that does not stop giving 'hyperthreading'.
If you mean Simultaneous Multi Threading (SMT) ... well now you know what to look for.

Given your failure to know what to call it, I also doubt your logic behind disabling it, can you enlight us... Why would you want to disable this ?

If the goal is to prevent jumping processes, try Usmus powerplan.