On my singlecore netbook with Intel Atom N450 (N455) CPU, engines benefit greatly from a second thread although it's just hyperthreading. I was surprised. The old dogma of HT. being not useful for chess engine calculations, seems outdated.
The cpu type is acknowledged by Cpuz.
The tests were with Windows 7 32-bit, 256 MB hash, comparing time to depth x in three different positions each. The new Stockfish gains more than 2:1.
(This whole system is of course rather slow in comparison to an i7 etc. still.)
To get real values, disable hyperthreading in the BIOS and benchmark the engine. Then enable hyperthreading and test it again, a) with one core then b) with two cores. Then come back with your results and I explain them to you.
Why I do not explain them right now ? Maybe my theory is wrong therefor I wait for your results...
georgerifkin wrote:this is interesting I've experience the same
does anyone have an explanation for why hyperthreading works well with atoms?
This was observed and explained here before. The reason is that the Atom is not an out-of-order machine, but a derivative of the good old Pentium I. It thus spends most of its time waiting for the completion of other instructions or memory fetches. That provides excellent opportunity for the other hyperthread to execute its instructions without hindering each other.
The reason is that the Atom is not an out-of-order machine
I new the Atom was shite, but i never realised this fact.
That places the atom up there with the crappy pentium 4 as a shit (and vastly successful) product shoved on the average home user by intel's marketing machine.
What is the actual benchmark difference with/without hyperthreading ?
I don't know about Atoms, but I can tell you about my experience with i7 Sandy bridge. I'll be grateful if someone else share his experience too with that chip (Albert Silver for example) for the same regard. Using Hyper-threading, my results was highly noisy and controversial with almost every testing method I've tried. Although of course, I'm not using the full capacity, only the half physical, which is 25% with hyper-threading. Switching it off however eliminated that almost completely and results get back to normal and much close to the other CPU's results I've got.
Mincho Georgiev wrote:I don't know about Atoms, but I can tell you about my experience with i7 Sandy bridge. I'll be grateful if someone else share his experience too with that chip (Albert Silver for example) for the same regard. Using Hyper-threading, my results was highly noisy and controversial with almost every testing method I've tried. Although of course, I'm not using the full capacity, only the half physical, which is 25% with hyper-threading. Switching it off however eliminated that almost completely and results get back to normal and much close to the other CPU's results I've got.
Zdravei Mincho,
Hyper-Threading e kato morskoto svinche....nito e morsko,nito e svinche
Pozdravi za tsiala Bulgaria i po-specialno za vsichki burgazlii ,
Dr.D
_No one can hit as hard as life.But it ain’t about how hard you can hit.It’s about how hard you can get hit and keep moving forward.How much you can take and keep moving forward….
Mincho Georgiev wrote:I don't know about Atoms, but I can tell you about my experience with i7 Sandy bridge. I'll be grateful if someone else share his experience too with that chip (Albert Silver for example) for the same regard. Using Hyper-threading, my results was highly noisy and controversial with almost every testing method I've tried. Although of course, I'm not using the full capacity, only the half physical, which is 25% with hyper-threading. Switching it off however eliminated that almost completely and results get back to normal and much close to the other CPU's results I've got.
I have no doubt about your question. I use two i7 computers, one with 4 real cores + 4 HT and the other with 6 real cores + 6 HT. After intensive testing my conclusions are clear (for i7 environment):
1.- No engine works worse with HT than whithout it.
2.- The engine that takes the maximum advantatge from HT is DR4.
3.- Houdini 1.5 and Houdini 2.0 take only a very marginal advantadge from the use of HT.
I suggest you to try it yourself. Perhaps my computers work in a different way than all other in the world.
Mincho Georgiev wrote:I don't know about Atoms, but I can tell you about my experience with i7 Sandy bridge. I'll be grateful if someone else share his experience too with that chip (Albert Silver for example) for the same regard. Using Hyper-threading, my results was highly noisy and controversial with almost every testing method I've tried. Although of course, I'm not using the full capacity, only the half physical, which is 25% with hyper-threading. Switching it off however eliminated that almost completely and results get back to normal and much close to the other CPU's results I've got.
I have no doubt about your question. I use two i7 computers, one with 4 real cores + 4 HT and the other with 6 real cores + 6 HT. After intensive testing my conclusions are clear (for i7 environment):
1.- No engine works worse with HT than whithout it.
2.- The engine that takes the maximum advantatge from HT is DR4.
3.- Houdini 1.5 and Houdini 2.0 take only a very marginal advantadge from the use of HT.
I suggest you to try it yourself. Perhaps my computers work in a different way than all other in the world.
Regards from Barcelona.
Tomcass
I didn't meant running SMP. What I'm pretty sure about is that running single core test is giving me very noisy results. Whether parallel program
runs better or not with HT is another subject. By I7 you mean Sandy Bridge, or not?
,