• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

What does hyperthreading stress when enabled

got a nice little fancy overclock on a 920 to 4.21Ghz HT off,, stable.. I just thought i'd see if it's stable with HT on and it crashed LinX after 3 iterations.. what is HT stressing to cause it to crash.. got a BSOD

i'm guessing NB voltage too low? 1.23v

just curious

EDIT: currently on 29 iterations LinX HT off no problems
 
I think it is about 10% more silicon used to manage it, and of course the extra using of parts that would otherwise be allowed to idle now and then instead of doing actual work.

beyond that, I've no idea exactly.
 
HT uses more power, so your voltage is probably to low. Evidence of that is noticeably higher (2-4C) temps with HT turned on.
 
I think it is about 10% more silicon used to manage it, and of course the extra using of parts that would otherwise be allowed to idle now and then instead of doing actual work.

beyond that, I've no idea exactly.

you're right
a single, possibly 2 notch voltage bump should be enough to overcome the latter (extra using of parts that would otherwise be allowed to idle). If not, then maybe the HT silicon is not as well optimized and is a critical path.
 
You could lower the OC 100Mhz as well to make it stable. The bad part is I don't think much of anything benefits from it. I would rather Intel have used those resources for something else.
 
You could lower the OC 100Mhz as well to make it stable. The bad part is I don't think much of anything benefits from it. I would rather Intel have used those resources for something else.

anything running code that is heavily threaded that is not exclusively integer computation benefits from it.
 
Interesting discussion. When I set up my i7-2600K/Z68 system, I chose to do the over-clocking with HT "Enabled." I should be able to turn it off and get a higher over-clock with the same thermal results, and if not the same voltage, then not much more.
 
anything running code that is heavily threaded that is not exclusively integer computation benefits from it.

Actually its the opposite. FP code gets less gains then on Integer.

Nehalem

SpecInt06: 13% gain
SpecFP06: 7% gain

There are very few instances that won't gain in multi-threaded scenarios. If it never goes further than L1 cache, or code has average IPC close to the width of the CPU(in this case, 4).

Here it shows that in games, Core 2 has CPI of 1.2 to 1.7, meaning it doesn't even reach 1 instruction per clock(since IPC=CPI/1). Problem is that in games, it usually only works well to 3 or 4 threads, making extra 4 gained by Hyperthreading in quad core CPUs worthless.

However in dual core: http://hardware-review24.com/load/c...pentium_g840_and_pentium_g620_review/1-1-0-60

You can see that 5 out of 6 games have gains that are better than pure clock speed differences indicate(plus clock speeds do not scale linear), and both G850 and Core i3 2100 have same amount of cores and cache.
 
Last edited:
Actually its the opposite. FP code gets less gains then on Integer.

Nehalem

SpecInt06: 13% gain
SpecFP06: 7% gain

There are very few instances that won't gain in multi-threaded scenarios. If it never goes further than L1 cache, or code has average IPC close to the width of the CPU(in this case, 4).

Here it shows that in games, Core 2 has CPI of 1.2 to 1.7, meaning it doesn't even reach 1 instruction per clock(since IPC=CPI/1). Problem is that in games, it usually only works well to 3 or 4 threads, making extra 4 gained by Hyperthreading in quad core CPUs worthless.

However in dual core: http://hardware-review24.com/load/c...pentium_g840_and_pentium_g620_review/1-1-0-60

You can see that 5 out of 6 games have gains that are better than pure clock speed differences indicate(plus clock speeds do not scale linear), and both G850 and Core i3 2100 have same amount of cores and cache.

apologies, fp code counts as "integer" to me. What I meant by integer was "maths".
IE, compiling code, running code that's more than just maths, can greatly benefit from hyperthreading thanks to stalls. Maths rarely lead to stalls.

games-- that may change if they get better at threading-- IE starcraft2 should have been able to scale to infinity threads given all those units but, wasn't....
 
On my 930 I can run prime with HT at 4000mhz stable 1.31-1.33
At 4200mhz I can only run 7 cores in prime.
It hard be 1 more core can cause a crash but it does.
My choices were use 1.4+v or shut off ht.
I shut off HT and left it at 4200mhz .
 
Back
Top