i7 2600k Hyperthreading Y/N?

[Tanuki]

I am running an i7 2600k on Win7 64 with hyperthreading enabled. It turbos constantly at 3.8.

With Prime95 torture tests I am getting 61, 68, 63, and 61 as max on the four cores.

Was thinking about disabling hyperthreading and trying a higher multiplier like 42 or 44.

I care about games and emulators for games.

Any reason not to kill HT?

[Abram]

For games, most certainly yes, disable it. Having HT on gives an FPS loss in all games, and introduces stuttering in some.

Having it off raise the temps a bit, but also allows for higher overclocks. I have mine at 4.8 all the time, and I'm pretty sure i could have it at 5.0ghz but I'm too impatient for all that tweaking/testing/retweaking. Using a Noctua cooler with 2x 120mm fans. I have LLC at Level 7 and CPU voltage at 1.33, which is dynamically altered by my LCC setting. , I never even installed the stock cooler, though I would not expect more than 4.2ghz.


But HT is awesome for thing like encoding video and music. Cuts my encode time in half. It'll do a full 1080p BD-50 in about 80 minutes.

[Skid]

For games, most certainly yes, disable it. Having HT on gives an FPS loss in all games, and introduces stuttering in some.

Having it off raise the temps a bit, but also allows for higher overclocks. I have mine at 4.8 all the time, and I'm pretty sure i could have it at 5.0ghz but I'm too impatient for all that tweaking/testing/retweaking. Using a Noctua cooler with 2x 120mm fans. I have LLC at Level 7 and CPU voltage at 1.33, which is dynamically altered by my LCC setting. , I never even installed the stock cooler, though I would not expect more than 4.2ghz.


But HT is awesome for thing like encoding video and music. Cuts my encode time in half. It'll do a full 1080p BD-50 in about 80 minutes.

Just to correct one point, turning HT off lowers temps not raise them, because it reduces the load on the CPU, its the same reason why you can get higher overclocks with it turned off.

[Abram]

Haha yes I thought I had corrected that.

[dopefish]

I did a lot of benchmarking and tests with this when I first got my 2600k because I was curious.

After spending ridiculous amounts of time testing everything, I concluded that it was best to leave it on. I saw no performance loss in any game by having it enabled. I also didn't see much gain. There are a selection of games where it does increase performance.

On top of this I do a lot of video encoding, multi-tasking, and all kinds of stuff at the same time, and having HT enabled definitely increases performance, especially with video encoding.

I run mine with a turbo ratio of 4.8. I can run it at 5.0 but the temps make me feel uneasy so I backed it down to 4.8. It runs perfectly fine at 4.8 with HT enabled.

In fact, the only reason I can think of why HT might hurt gaming performance is due to heat. If you don't have a good cooler and you start generating too much heat, the processor will naturally cut back on performance to save the CPU. Get yourself a great cooler and you'll be fine.

I looked up others' benchmarks throughout the HardOCP forums and other various tech forums and the average seems to be that a slightly lower clock with HT enabled is faster than a slightly higher clock with no HT (in games that will utilize this many threads, of course—like Battlefield 3).

Anyway, I recommend you do a lot of benchmarking yourself. Raise your turbo multiplier (and voltage if needed), then try with and without HT. Monitor your temperatures throughout the entire thing. If you see that the temperatures are high then try without HT and then again at a slightly lower clock with HT.

[Tanuki]

Well HT off drops each core 5c.

I bumped to 4.5 and Prime95 was giving me 67, 73, 69, 66 for the cores max temp with no HT.

I then remembered to clean my dust traps and I now get 61, 69, 65, 62 for the max temps with Prime 95.

Think I will give it an hour more and then get some benches and then try the same clock with HT.

[ViciousXUSMC]

I would leave it on unless you run into a situation that demands having it off.

[Gilly]

Mine stays on all the time, and I don't have problems running games (stuttering), I am happy with the rendering performance in Sony Vegas, and the temps are fine. I have it at 4.2Ghz, haven't tried it higher, although with these temps I am sure I could push it to 4.6 or 8 no problems :)

[QuackingPlums]

I bought both an i5 and an i7 when I built my rig with the aim of testing overclockability in the same rig setup. Two other friends did the same thing to average out any batch issues. Ultimately the i7s didn't go as far with HT on, but gave similar results to the i5 with it off. With the games and benchmarks that we tested, HT gave no benefit to maximum or average FPS.
As my rig is games only and I was only bothered about this one metric, I left the i5 in there.

It's worth noting that when I down-clocked to 4.3 the benchmarks reported next to no difference in the FPS scores either, so the concept of 'max FPS' is relative.

[Tanuki]

Ran some benchies. I know we are supposed to love benchies so here they are.

Used my Steam games with built-in benchmarks only. And Dolphin. NVIDIA driver settings were for max quality with AA and Vsync application controlled. In-game settings were maxed with no vsync at 1920x1080.

As an aside, if I didn't before, I now hate screen tearing.

Win7 64
i7 2600k 4.5
16gb DDR3 1600 ram
Intel 320 SSD 120GB
MSI Twin Frozer II 560ti 1GB Dual SLI
Xi-Fi Titanium PCIe



Mafia II no HT
51.8
48.7
50.8
Mafia II no HT Avg. 50.43

Mafia II HT
50.7
48.4
48.5
Mafia II HT Avg. 49.20



Just Cause 2 no HT
Concrete Jungle
49.66
49.92
49.68
Just Cause 2 no HT Avg. 49.75

Just Cause 2 HT
Concrete Jungle
49.17
49.36
49.28
Just Cause 2 HT Avg. 49.27



Lost Planet Extreme Condition DX10 no HT
Snow
23.3
24.2
23.6
Lost Planet Extreme Condition DX10 Snow no HT Avg. 23.7

Lost Planet Extreme Condition DX10 HT
Snow
23.5
23.3
23.4
Lost Planet Extreme Condition DX10 HT Snow Avg. 23.4



Lost Planet Extreme Condition DX10 no HT
Cave
39.5
38.8
39.1
Lost Planet Extreme Condition DX10 Cave no HT Avg. 39.13

Lost Planet Extreme Condition DX10 HT
Cave
38.7
39.2
38.9
Lost Planet Extreme Condition DX10 HT Cave Avg. 38.93



Batman: Arkham City DX11 no HT
39
39
38
Batman: Arkham City DX11 no HT Avg. 38.6

Batman: Arkham City DX11 HT
39
39
39
Batman: Arkham City DX11 HT Avg. 39



Half-Life 2 Lost Coast no HT
279.68
279.15
279.18
Half-Life 2 Lost Coast no HT Avg. 279.336

Half-Life 2 Lost Coast HT
281.60
282.46
281.35
Half-Life 2 Lost Coast HT Avg. 281.803



Shogun 2 DX11 Graphics High 1080p no HT
72.9625
72.75
72.9125
Shogun 2 DX11 Graphics High 1080p no HT Avg. 72.875

Shogun 2 DX11 Graphics High 1080p HT
72.875
72.8125
72.95
Shogun 2 DX11 Graphics High 1080p HT Avg. 72.879

Dolphin: Twilight Princess no HT
Dolphin: Twilight Princess no HT Inside Avg. 47
Dolphin: Twilight Princess no HT Outside Avg. 37

Dolphin: Twilight Princess HT
Dolphin: Twilight Princess HT Inside Avg. 56
Dolphin: Twilight Princess HT Outside Avg. 41

Lost Planet has the ability to increase concurent operations I left it at 4 for both. I could select 5-8 with HT enabled but 6-8 were flaky with stuttering in the Cave benchmark. Also, if anyone knows what the concurrent rendering option does in Lost Planet feel free to post.

Looking at the results, HT in PC games is a neglibile option. Perhaps newer games will start seeing more benefits but for pure gaming purposes it feels like the new X-Ram.

Dolphin however, had a noticiable improvement with HT enabled. Seeing as that was one of the reasons for overclocking in the first place I will be leaving it on. My max temps after 12 hours of Prime95 at 4.5 with HT were 68, 75, 70, and 68. They seemed to average a few Cs lower so I don't see a problem with leaving HT enabled or any need to downclock.

Out.

EDIT - I know I should have got the 2GB cards. No need to tell me.

[Tanuki]

Quick question world.

At idle, my vcore is around 1.376-1.384. When is gets more load it goes down. Vdroop IIRC.

I am at 1.312-1.328 at full load 8 threads in Prime95.

Is that too high of an idle vcore?

[dopefish]

Idle should be super low.

Mine idles at around 0.8-0.9v when it's running at 1600MHz.

When turbo kicks in and it goes up to 4800, it jumps up to 1.4v (which is what I have set).


What tool are you using to read those results? I'd verify with CPU-Z first. It's more accurate at measuring the voltage than a lot of other software.

[Tanuki]

I was giving the CPU-z results. ASUS AI Suite is very close to CPU-z as well. I am using a Sabertooth P67 mobo.

It is always showing as 4500 mhz even with just CPU-z running and the system at idle.

I also have all the power saving thingies enabled except EPU.

I guess I should take some pics of my bios.

[Tanuki]

I am guessing I borked something here.

[dopefish]

Sounds like you have SpeedStep disabled or something.

Here's what's different from my setup:
I use By Per Core instead of By All Cores and I set each one to 48, or 45 in your case.
Internal PLL Overvoltage set to Auto.
Load-line Calibration set to Ultra High.
VRM Frequency set to Manual.
VRM Fixed Frequency Mode set to 350.
Duty control set to Extreme.

I have my turbo set higher, so some of my options apply here and not yours.

I don't see anything in your screens related to SpeedStep (I don't think mine has a setting for that either, it's just always on.)

Edit:
Actually, it may be your By All Cores setting. This is left up to the OS to decide what ratio to use. You might need to install some Intel driver or something to manage the ratios within Windows. When set to By Per Core, the BIOS manages it. Try it out and let me know.

[Tanuki]

Changing tp per core did not make any perceptable difference. I think all cores allows a bios utility to change in Windows.

Internal PLL Overvoltage had no difference.

LLC was very, very bad. No change at idle but load went to 1.46+ vcore and one core was over 80C in seconds.

I was too scared to try the VRM frequency settings.

NO change with Duty Control.

However, on a lark I changed all the C states from Auto to Enabled and it worked I think.

Cores now idle at 1600 or 4500. They switch between. Vcores seem to be 1.0~~ to 1.2~~ when idling and 1.312-1.328 at full load with Prime95.

So, win, I think. Thanks.

[ugzz]

crap crap crap!
can't find the maximum PC article, it might have only been in the actual magazine and not online..

Basically they said that the performance was negligable in most tests, the only real time HT made a differance was apps that are multithreaded to work with more than 4 cores. and in those apps they saw as high as a 20% boost. and made the statement. Adding 4 more virtual cores can give you enough boost that its similar to adding 1 more physical core. but only in a very small percentage of apps.

if i can find somethign more solid than " i read this" ill link to it :)

[dopefish]

Hyper-threading isn't the same as a "virtual core" nor is it the same as a physical core.

The only reason you see a gain in a small percentage of apps is because not every app requires more than four threads. Applications that use more than four threads will always see a gain in performance. This is why audio and video applications often see large gains in performance while games do not. Old games are almost always run on a single thread. Even when two physical processors and dual-core processors became more common games were still designed primarily around running on a single thread. It wasn't until the Xbox 360 and PS3 that games really started to take advantage of multiple threads. This is due to the nature of the Xbox 360 having three cores capable of executing two threads each (six threads total), and the PS3 having one primary core with 6 effective smaller cores. To get the maximum performance out of games developers had to start multi-threading their games. For quite awhile the Xbox 360 and PS3 were both ahead of PC performance because at the time most PC users only had dual-core systems.

Anyway, because of this, as more console games get ported over to the PC, and new games come out in general, they will be taking advantage of more threads and cores and it will be beneficial to have HT.

[Tanuki]

So I was getting some display flickering after cold boots that had no power going to the components for a while.

Figured I screwed up and forgot everything I learned about troubleshooting. Messed with the BIOS settings, software, and connections before wising up and attempting to record it in FRAPS. No problems show. I guess it is a problem with the LCD. I think I am under warranty.