3930k - Effects of Hyper Threading

[DerStig]

I would like to start this thread so that everyone is aware what kind of a little evil hyper threading is. Here is the hardware I am using:

3930k @ 5.0 Ghz
Asus RIVE
Corsair Dominator Platinum 2400 Mhz 16 GB (4 x 4 GB)
Samsung 840 Pro 512 GB
Evga GTX 680 4 GB
Corsair AX1200i

Case-labs
3 x 480 Black Ice SR1
EK-D5 Dual top
EK Supremacy
Heatkiller GTX 680
Monsoon fittings all around
Noiseblocker eloop fans

First things first, I don't have the best OC chip in the world. It's certainly not bad, and I'd say its above average, but not the best. I can do stock 3.8 Ghz with 1.095 vcore if that tells you anything. My main goal for building this rig was to have a 24/7 "quiet" PC that would be pushed to its limits. I wanted to hit 5.0 Ghz at any cost and keep it that way for as long as it goes.

As I put everything together and started OCing, I suddenly realized that to get to 5.0 Ghz, I would have to hit some crazy vcore numbers. No matter what I did, the heat would build up and the voltages I had to use were very high. Fast forward a few weeks, I suddenly realized that I never bothered to turn hyper threading off. It's something no games I play benefit from in fact, it hurts them (FSX and BF3).

To my surprise, turning hyper threading off made a massive difference. Not only my temps went down (13-14C difference!), my voltages went down as well. And strangely enough, my FPS in BF3 increased! The data below will speak for itself:

Hyper threading on (these numbers are Prime95 stable)

4.8 Ghz :

vcore : 1.495 (load 1.516)
vtt : 1.375
vccsa : 1.375
llc : high
power phase : 150%
dram : 1.67
bclk : 100
multi : 48
RAM : 2400
Timings : 10-12-11-31-2
Temperature Prime95 : 78C (Average)
Temperature Gaming (FSX) : 66C (Average)
Temperature Gaming (BF3) : 62C (Average)

5.0 Ghz :

vcore : 1.585 (load 1.607)
vtt : 1.4
vccsa : 1.4
llc : high
power phase : 180%
dram : 1.68
bclk : 100
multi : 50
RAM : 2400
Timings : 10-12-11-31-2
Temperature Prime95 : 86C (Average)
Temperature Gaming (FSX) : 72C (Average)
Temperature Gaming (BF3) : 66C (Average)

FPS in BF3 : Mostly in 50s, with sudden dips to 30s.

Hyper threading off (these numbers are Prime95 stable)

4.8 Ghz :

vcore : 1.450 (load 1.464)
vtt : 1.0
vccsa : 1.0
llc : high
power phase : 150%
dram : 1.65
bclk : 100
multi : 48
RAM : 2400
Timings : 9-11-11-31-2
Temperature Prime95 : 64C (Average)
Temperature Gaming (FSX) : 55C (Average)
Temperature Gaming (BF3) : 50C (Average)

5.0 Ghz :

vcore : 1.515 (load 1.536)
vtt : 1.2
vccsa : 1.2
llc : high
power phase : 150%
dram : 1.65
bclk : 100
multi : 50
RAM : 2400
Timings : 9-11-11-31-2
Temperature Prime95 :70C (Average)
Temperature Gaming (FSX) : 59C (Average)
Temperature Gaming (BF3) : 52C (Average)

FPS in BF3 : Mostly in 60s, sometimes in 50s, but the dip is not sudden and the numbers are smooth.

Conclusion:

* Voltages were dropped by a huge difference.
* Temperatures were down by 13-14C
* VTT/VCSSA and other motherboard settings are a lot lower (I could do 1.0 VTT in 4.8 Ghz!!!!)
* Ram timings were a lot tighter (I can do 9 tCAS at 5.0 Ghz! with 2400 Mhz ram)
* BF3 FPS increased

[Lvcoyote]

Nice post, and I believe what you are saying. Would be better with some screen shots though.....

[DerStig]

Nice post, and I believe what you are saying. Would be better with some screen shots though.....

I'm sorry, there is no way I am going to turn this thing on again. I don't know what screenshots you need though, I'm sure anyone with this chip can prove me right (or wrong).

I guess it is safe to say that as you go higher in clocks and volts, the impact of hyper threading is more obvious in terms of heat and the amount of extra voltage that the chip requires.

The difference in BF3 performance is amazing though and those FPS figures might sound low even with HT off, but I use 2560 x 1600 monitor and I set everything to maximum.

[eskimochaos]

I wonder what other software benefits by turning virtual cores off.

[Lvcoyote]

I'm sorry, there is no way I am going to turn this thing on again. I don't know what screenshots you need though, I'm sure anyone with this chip can prove me right (or wrong).

That's my point, with proper documentation of your claims, there would be no need for anyone to prove you right or wrong. It's just something for you to consider next time you make a similar post, not a big deal.

[peugot]

from what i have read, this is old problems that did appear to sandy bridge only, but i dont know personally..? does this happen to all cpu with HT or just those 6-cores? i have a quad core 3820 with hyperthreading on, and i havent ran into any issues with it(not in bf3 either). smooth even @4.675 with vcore 1.390 (original 3.60Ghz@ v1.200, no problem in bf3 then either). but will i get higher fps if i turn off HT today, do you know this?

[freeagent]

a 5ghz 24/7 hexa chip seems a tad unreasonable to me

Ive played fsx before, and it ran great even with a lower oc than yours. A bad oc can make things run a bit sluggish too, wich could be what you are experienceing, considering youre other thread asking what a stable oc is

[DerStig]

The problems with FSX is not as simple as it is with BF3. With BF3, I saw immediate FPS increase at my overall final FPS and I also stopped seeing the constant and absurd FPS drops that would happen here and there.

With FSX, as I used a 111111111111 affinity mask, having virtual cores load textures would cause stutters. This is well proven in any intel chip with HT, its not something I am making up. FPS wise with or without HT, my FSX runs pretty smooth.

[DerStig]

That's my point, with proper documentation of your claims, there would be no need for anyone to prove you right or wrong. It's just something for you to consider next time you make a similar post, not a big deal.

Fair enough, but I'm not really the type of guy who makes these types of posts really (look at my post count). I wasn't prepared for something like this I didn't even think I would have to go through this to get to my target OC. I just wanted to share my experiences that's all.

[Bobnova]

Do be aware that on properly threaded stuff HT gains you ~30% CPU processing power. Some things it's not good at, a few things it's actively bad at, others it is very very good at.
Play with Cinebench a little bit for instance

[DerStig]

With what I do, I need raw core power. So to compare (temperature wise) :

(Even 4.8 Ghz with HT ON is hotter than 5.0 Ghz with HT OFF, so I used 4.6 Ghz as I remember 4.5 Ghz with HT ON was colder than 5.0 Ghz HT OFF)

HT ON : 4.6 Ghz/2400 Mhz/10-12-11-31-2
HT OFF : 5.0 Ghz/2400 Mhz/9-11-11-31-2

With HT on, I get 12 cores, but they run slower, with slower ram timings. With HT off, I get 6 cores, but they run faster and with faster timings. No game that I play (in fact let's make it even more general, most popular games) will perform better in the first scenario. 6 cores is plenty (in fact its more than enough) for any game out there.

I also get to run my RAM at the advertised speed and timings at much lower VTT/VCSSA voltages which adds to the performance and also will extend the CPU life time.

[EarthDog]

Thanks for documenting slthese things...there are a rare few games that do show worse fps with ht on, not many though. Its generally know about the voltages and ht, so more evidence of that is cool too.

Also, at least with my limited experience with a 3930k I could hit the same memory and timings regardless of clockspeed on ambient temps so long as I wasn't running HOT.

That said, screenshot are a huge +.