HTvsOC

[ScratchFive]

Alright peeps?

Couldn’t find one of these so decided to make one... hope its useful to someone.

I have read that most games don't use more than 4 cores, and that you can overclock higher with HT turned off, so turning HT off night be an advantage to many, right?

Im going to run benchmarks with HT on, on the highest BCLK value I can with a set of BIOS settings, then turn HT off, up the BCLK until it becomes unstable (using the same test as before) knock it back one (where it was last stable) and then run the same set of benchmarks, hopefully giving me a rule as to which is better, ok?

The main memory will change frequency as well as the CPU but thats the deal, right!? memtest will be run before both benchmarks to ensure stability.

My stability test will be 40 iterations of IBT on standard stress and with the appropriate amount of threads, ideally this would be longer but I don’t want this to take forever.

The benchmarks I will run will be 3DMark 11 and Heaven.

My setup

i7 860 (21x multi 1.264v LLC)
GTX 460 870MHz (tested stable)
Maximus 3
4gb GSkill RipJaws X (2:12 9)
CM GX650 PSU
Corsair A50 cooler

Results:

HT enabled:

BCLK 180

Heaven: FPS 47.9
Score 1206
min FPS 10.7
max FPS 114.5

3DMark: 3908


HT disabled:

BCLK 184

Heaven: FPS 48
Score 1209
min FPS 26.1
max FPS 111.9

3DMark: 4095

while these results are relatively mundane, the temperature difference was significant: the test with HT disabled was 9 degrees cooler under IBT and prime than the one with HT enabled (at the same voltage).

hottest core temps in degrees Celsius

HT – 72
OC – 63

this, I believe gives OC the clear advantage as there is lots of headroom for it to be improved.

hope this is a little interesting!

Have fun

 

[R.C.Anderson]

Personally I don't think this shows very much, as the sort of variance in your results is to be expected just from running the test a couple of times. Also, with the benchmarks are almost completely gfx card dependant, it doesn't really give a chance to demonstrate the differences.

Try running a few benchmarks like Wprime/SuperPi, and have a look at the differences, my guess would be that the difference in those would be significant, as they are CPU dependant, and as such, you can then decide if your extra couple of fps are worth the decline in CPU performance in other situations.

 

[paradigm]

And what, prey-tell, are the differences in CPU clock between HT on and HT off?

It makes less than 100MHz difference to my 920, so I leave it on.

 

[ScratchFive]

Personally I don't think this shows very much, as the sort of variance in your results is to be expected just from running the test a couple of times. Also, with the benchmarks are almost completely gfx card dependant, it doesn't really give a chance to demonstrate the differences.

I wanted this to be relevant for games, I was interested as to whether they can use 8 threads efficiently, when I get round to it I will run these benchmarks, but the reason I posted this was to show that there was very little or no difference for someone who was wondering the same as me. I was slightly surprised at the temperature difference though

And what, prey-tell, are the differences in CPU clock between HT on and HT off?

It makes less than 100MHz difference to my 920, so I leave it on.

its in there... (184-180)*21=84MHz but whose to say that doesn't make a difference?

P.S.
I have and will continue to run my overclock with HT ON...

 

[ScratchFive]

more tests

ok so... CPU benchmarks:

I ran 3 sets of BIOS settings

Test one:

HT: on
BCLK: 180
SuperPi 4M calc (2 run avg): 55.1615s
wPrime 32 (5 run avg): 7.6192s
max temp under IBT: 69


Test two:

HT: off
BCLK: 184
SuperPi 4M calc (2 run avg): 53.9995s
wPrime 32 (5 run avg): 9.4752s
max temp under IBT: 60

Test three:

HT: off
BCLK: 190
SuperPi 4M calc (2 run avg): 52.4235s
wPrime 32 (5 run avg): 9.1696s
max temp under IBT: 68
(note that there was a voltage increase to get this to just under the temp of test one)


as you can notice SuperPi has a linear response to the overclocking regardless of HT probably because it has <4 threads...
HT owns in wPrime. no more need be said