How much is hyperthreading hurting my chess performance...

[Terry McCracken]

Ok, thanks very much. So likely it is a good idea for quality analysis to turn of HT. But Turbo Boost? That sucks...

There are two parameters in question: turbo boost and the other is SpeedStep. SpeedStep allows the OS to adjust the speed up and down based on needs (mostly a green thing - you don't need hi GHz when nothing is happening).

I turn both off and OC my computer beyond what Turbo would do. If my computer is on, it is doing something.

It is up to you how accurate you want to be on your testing.

Yeah, I get it but don't overclock but would like to keep the boost when pushing it.

[CRoberson]

Ok, thanks very much. So likely it is a good idea for quality analysis to turn of HT. But Turbo Boost? That sucks...

There are two parameters in question: turbo boost and the other is SpeedStep. SpeedStep allows the OS to adjust the speed up and down based on needs (mostly a green thing - you don't need hi GHz when nothing is happening).

I turn both off and OC my computer beyond what Turbo would do. If my computer is on, it is doing something.

It is up to you how accurate you want to be on your testing.

Yeah, I get it but don't overclock but would like to keep the boost when pushing it.

The turbo boost is just a milder OC. Sure keep it on.

[Terry McCracken]

Ok, thanks very much. So likely it is a good idea for quality analysis to turn of HT. But Turbo Boost? That sucks...

There are two parameters in question: turbo boost and the other is SpeedStep. SpeedStep allows the OS to adjust the speed up and down based on needs (mostly a green thing - you don't need hi GHz when nothing is happening).

I turn both off and OC my computer beyond what Turbo would do. If my computer is on, it is doing something.

It is up to you how accurate you want to be on your testing.

Yeah, I get it but don't overclock but would like to keep the boost when pushing it.

The turbo boost is just a milder OC. Sure keep it on.

Ok, and thanks again!

[Mincho Georgiev]

Terry, I'll try to explain my experience with SB, it could be useful for you, because I'm trying to figure out what's working best during my tests also.
Now, first I'm using 80% of the time XP x64 as my primary OS.
What I've noticed is that the CPU advanced features turned on /HT especially/ are generating extremely noisy results and huge amount of randomness which I've never had before with any CPU (Intel or AMD).
I can't tell for sure how is it under linux, since I've only devoted it 20% of the time, but under Windows, and I suppose that this goes out to WIN7 as well, the thread scheduling with HT gets badly messed up. I never run with more than 30% utilization and still, the active core executes other threads. This is only a suspicion of mine and I can't give 100% guarantee that is correct, but at least IMO it is just like that.
What I'm doing right now is running all the tests without ANY of the advanced features turned on. I even switch off virtual technology and isochronious support and everything goes smooth.
There is no speed difference for me, but I don't use 8 threads as I assume you would want to.
Anyway, I don't think that other CPU features would have an impact on
the results correctness, but under XP64 hyper-threading and intel EIST - definitely.
However, I assume that this is a bit far from the subject, you wanted to know if there is a speed difference and I'm talking about correctness, but just wanted to share my experience.

[daws]

but under Windows, and I suppose that this goes out to WIN7 as well, the thread scheduling with HT gets badly messed up.

Well actually there is a difference in hyperthreading between XP 64 and Windows 7 64. Basically, XP can't tell the difference between a logical and a physical core whereas Windows 7 can. Theoretically, Windows 7 should always schedule available physical cores before logical cores, but this is not the case for XP.

For myself, I switched hyperthreading off, on my I7 860, almost two years ago and have never felt the need to switch it back on again. Although I remember noticing that tasks such as graphics rendering do benefit from hyperthreading. So I guess it's horses for courses.

[Mincho Georgiev]

but under Windows, and I suppose that this goes out to WIN7 as well, the thread scheduling with HT gets badly messed up.

Well actually there is a difference in hyperthreading between XP 64 and Windows 7 64. Basically, XP can't tell the difference between a logical and a physical core whereas Windows 7 can. Theoretically, Windows 7 should always schedule available physical cores before logical cores, but this is not the case for XP.

Possible. I'm unaware if that's the case.