HyperThreading! Did anyone ever tested it?

[n3krO]

I'm tired of people claiming it's useless for most games except for bf3,4 and crysis 3.

 

So i wonder if someone who have an unlocked i7 ever tried to do the max stable overclock with HT and then without HT (Without HT it's supposed to achieve higher stable overcocks, right?) to test any of these 3 games to see if it's worth it or not.

 

All comparisons i see is with same clock.

[Volbet]

A test like this would be doable. If anyone is interested, I'll add it to my ´to-do-fast´ list. I just have to finish my Windows v. Linux gaming test first.

But I have a 3930K, so it might not be the best thing to test.

[qwertywarrior]

HT shows benefits on multithreaded software regardless of what the software is

 

if you want a prime example get a 2core intel cpu and turn HT on and off

[papeles]

HT has nothing to do with overclocking or even stability. I can't really understand what op means. You wanna know if HT gives any benifit to games? Yes it does, just a bit, for some games who require/take advantage of alot of cores. But it is minimal...

[vm'N]

Hyper-threading (Also known as SMT) is essentially running two threads (line of code) simultaneously on an single core.

 

SMT is viable because newer architectures have such an wide executing pipeline, it is very hard utilizing every resources by a single thread.

 

Both threads are having are dynamic relationship with the resources. Non are giving an static priority over the other.

 

SMT shouldn't decrease overclocking, however you might run into over-heating faster (you are essentially using the entire core, where without you maybe only utilized ~75).

 

 

Specially coded software for SMT does show increased performance over normal multithreaded software.

[harrynowl]

I'll do some testing on my 2600K at 4.2, 4.5, 4.7 and 4.9 tonight if enough people actually want to see the results.

 

I'll use BF4 for my testing (Siege 64player).

 

I can also test Wow/Medieval 2 total war, but they only use 1-2 cores so pretty much any result will be irrelevant.

[P e r s e p h o n e]

Based on the video below, though conjecture, I think the reason why the i3 4130 did so well up against the i5 4670k was due to is hyper-threading. However, moot point, in the end a 4^th generation i3 will barely pull gains ahead of an i5 or i7 in most video games. The only great disparage is in strategy games and Metro Last Light, even then its about 10-15% when pared with a high end GPU. In other words I'm saying it doesn't really matter if hyper-threading is utilized statistics show weak modern processors are sufficient for gaming.

 

 

Sorry this video isn't from Linus but Austin has exemplary benchmarks on this. I remember Linus mentioned the same thing in another video but I honestly just can't remember where it is.

 

[Prokart2000]

jayztwocents did a bottleneck vid

[deathjester]

I'm tired of people claiming it's useless for most games except for bf3,4 and crysis 3.

 

So i wonder if someone who have an unlocked i7 ever tried to do the max stable overclock with HT and then without HT (Without HT it's supposed to achieve higher stable overcocks, right?) to test any of these 3 games to see if it's worth it or not.

 

All comparisons i see is with same clock.

 

I have a 4770k 4.5 ghz, with verified screenshots of the temps/clock in the cinebench thread.  HT on/off. 1-3 FPS in BF4 on a GTX 770, and the 3 FPS might have been because more people were on screen and more explosions etc. If you turned HT on and off I literally could not tell which was which.

 

WoW, Guild Wars 2, Wild Star, HT off was actually slightly faster, which was pretty funny, but again that could have been more people in town. Other people noticed that HT was slower in Wild Star as well. Whether they just tested at different times, or there really is an issue? I don't know. Single player games like Tomb Raider I never bothered testing as a difference would be impossible since my CPU ran at freakin near idle in the game. The game is a GPU benchmark.

 

Crysis 3? I never bothered  buying. Multiplayer has always been trash in the games. Maybe someone else could test Crysis 3. Tom's Hardware has an I5 getting 10 fps higher on lows then a 8350 though, so these Crysis benchmarks appear that people don't know what the hell they are doing, and only a few "benchmarkers" are testing anything that actually gets CPU bound.

 

Short story. It makes no difference on my GTX 770. Recording the game with FRAPS? Yeah HT is awesome. I see a big swing there. Maybe someone else has a 780TI or 2 and sees a difference, but I sure don't on a GTX 770 in any game. Shooter, MMO, single player and about the only really big game I haven't played is Crysis 3. Dark Souls 2 I never bothered, but I highly doubt it runs better with HT . 

 

BTW WoW and Wildstar use all my threads. There is just zero difference in FPS lows. Like all games one thread is way higher then the other threads. Same with Planetside 2. That game would be really hard to benchmark but again, I could not tell the difference if you turned HT on and off and one thread is maxed out and the others are doing jack. Total CPU usage is like 25 percent on Planetside 2 with HT on.  

[n3krO]

 

I have a 4770k 4.5 ghz, with verified screenshots of the temps/clock in the cinebench thread.  HT on/off. 1-3 FPS in BF4 on a GTX 770, and the 3 FPS might have been because more people were on screen and more explosions etc. If you turned HT on and off I literally could not tell which was which.

 

WoW, Guild Wars 2, Wild Star, HT off was actually slightly faster, which was pretty funny, but again that could have been more people in town. Other people noticed that HT was slower in Wild Star as well. Whether they just tested at different times, or there really is an issue? I don't know. Single player games like Tomb Raider I never bothered testing as a difference would be impossible since my CPU ran at freakin near idle in the game. The game is a GPU benchmark.

 

Crysis 3? I never bothered  buying. Multiplayer has always been trash in the games. Maybe someone else could test Crysis 3. Tom's Hardware has an I5 getting 10 fps higher on lows then a 8350 though, so these Crysis benchmarks appear that people don't know what the hell they are doing, and only a few "benchmarkers" are testing anything that actually gets CPU bound.

 

Short story. It makes no difference on my GTX 770. Recording the game with FRAPS? Yeah HT is awesome. I see a big swing there. Maybe someone else has a 780TI or 2 and sees a difference, but I sure don't on a GTX 770 in any game. Shooter, MMO, single player and about the only really big game I haven't played is Crysis 3. Dark Souls 2 I never bothered, but I highly doubt it runs better with HT . 

 

BTW WoW and Wildstar use all my threads. There is just zero difference in FPS lows. Like all games one thread is way higher then the other threads. Same with Planetside 2. That game would be really hard to benchmark but again, I could not tell the difference if you turned HT on and off and one thread is maxed out and the others are doing jack. Total CPU usage is like 25 percent on Planetside 2 with HT on.  

 

 

But you are doing clock to clock comparisons, which is, in my opinion, wrong.

 

Btw, i also think this  test would be more apropiate with a better GPU just to make sure the GTX 770 isn't bottlenecking the CPU.

 

 

Hyper-threading (Also known as SMT) is essentially running two threads (line of code) simultaneously on an single core.

 

SMT is viable because newer architectures have such an wide executing pipeline, it is very hard utilizing every resources by a single thread.

 

Both threads are having are dynamic relationship with the resources. Non are giving an static priority over the other.

 

SMT shouldn't decrease overclocking, however you might run into over-heating faster (you are essentially using the entire core, where without you maybe only utilized ~75).

 

 

Specially coded software for SMT does show increased performance over normal multithreaded software.

 

So HT doesn't change the maximum stable overclock speed?

[deathjester]

But you are doing clock to clock comparisons, which is, in my opinion, wrong.

 

Btw, i also think this  test would be more apropiate with a better GPU just to make sure the GTX 770 isn't bottlenecking the CPU.

 

 

 

So HT doesn't change the maximum stable overclock speed?

 

What are you looking for like stock vs stock? That is going to be skewed since the 4770k has a higher stock turbo on 1-2 cores or 4. 1-2 cores a stock 4770k is like 3.9ghz so it will beat a stock I5 in something like Dolphin Emulator by quite a bit, or a OOOOOLD game that actually can't use multiple threads.

 

4 cores a 4770k is at 3.7. A I5 is 3.6. That is what makes these benchmarks when they show a 4770k 1 fps higher kind of non legit. One is clocked higher. The 4770k also has better cache but it doesn't appear to matter at all gaming. 

 

Here are all the turbo speeds of current Haswell.

http://www.tomshardware.com/reviews/core-i7-4770k-haswell-performance,3461.html

 

As far as overclocking. HT runs like 10 degrees hotter in a render. You are limited by heat as far as overclocking. So for like my 30 dollar air cooler? If I took HT off I could go like 4.6 and be ok on a render or like when gaming and streaming. My I7, 1.25v would simply be too hot. Then again when rendering and or streaming and playing a game the I7 is going to beat the I5 by quite a bit, so it is kind of a silly argument. Add to that you never know what you are gonna get on a OC on haswell so buying one thinking it will be higher is silly. You can be at 4.1-4.7 below 1.3v on water. It is all a crapshoot. The 4770k's are supposedly binned higher but you can lose just as bad on a current I7 on the silicon lottery. Going by the Devil's Canyon turbo numbers the new I7 really does appear to be much higher binned then the new I5, but we will see.

 

As far as a 780TI bottlenecking over a 770? I would say that ain't happening going by my CPU usage. Two? Possibly a I7 would show more of a difference.  Everything I have seen shows crossfire/sli on good cards being much better on an I5 then a 8350 though, so it might not be a big difference at all. Probably someone on this forum with dual 780's or 2 290's that could test it. 

[n3krO]

What are you looking for like stock vs stock? That is going to be skewed since the 4770k has a higher stock turbo on 1-2 cores or 4. 1-2 cores a stock 4770k is like 3.9ghz so it will beat a stock I5 in something like Dolphin Emulator by quite a bit, or a OOOOOLD game that actually can't use multiple threads.

 

4 cores a 4770k is at 3.7. A I5 is 3.6. That is what makes these benchmarks when they show a 4770k 1 fps higher kind of non legit. One is clocked higher. The 4770k also has better cache but it doesn't appear to matter at all gaming. 

 

Here are all the turbo speeds of current Haswell.

http://www.tomshardware.com/reviews/core-i7-4770k-haswell-performance,3461.html

 

As far as overclocking. HT runs like 10 degrees hotter in a render. You are limited by heat as far as overclocking. So for like my 30 dollar air cooler? If I took HT off I could go like 4.6 and be ok on a render or like when gaming and streaming. My I7, 1.25v would simply be too hot. Then again when rendering and or streaming and playing a game the I7 is going to beat the I5 by quite a bit, so it is kind of a silly argument. Add to that you never know what you are gonna get on a OC on haswell so buying one thinking it will be higher is silly. You can be at 4.1-4.7 below 1.3v on water. It is all a crapshoot. The 4770k's are supposedly binned higher but you can lose just as bad on a current I7 on the silicon lottery. Going by the Devil's Canyon turbo numbers the new I7 really does appear to be much higher binned then the new I5, but we will see.

 

As far as a 780TI bottlenecking over a 770? I would say that ain't happening going by my CPU usage. Two? Possibly a I7 would show more of a difference.  Everything I have seen shows crossfire/sli on good cards being much better on an I5 then a 8350 though, so it might not be a big difference at all. Probably someone on this forum with dual 780's or 2 290's that could test it. 

 

I'm looking for the best OC you can get on HT and the best OC you can get without HT. And compare how both perfomance on battlefield 4 for example. That's the only way you can really know if HT is worth it or not on that game imo. Not comparing clock to clock....

[Jumper118]

jayztwocents did a bottleneck vid

that was in valley which is a benchmark

[deathjester]

I'm looking for the best OC you can get on HT and the best OC you can get without HT. And compare how both perfomance on battlefield 4 for example. That's the only way you can really know if HT is worth it or not on that game imo. Not comparing clock to clock....

 

Just tested for ya with tons of crashes lol . I need the exact same voltage for HT and non HT. 4.6 was 1.25vish. Just like my HT. Temps are lower so I guess I could get 100 mhz more on a evo 212 with push/pull with HT off. Was rendering at 71C as opposed to mid 70's with HT on. Rendering score of course it gets smoked by HT. 

 

For BF4? I wouldn't do it. They would prob perform exactly the same and you lose the ability to stream. Dolphin emulator. Yeah might be worth it on some games, especially if you got a crappy overclocking Haswell at like 4.1-4.2. MMO's? I might use the profile in Wild Star. Maybe another 2 FPS with HT off? I would test it but beta is up. Then again if I am streaming or recording the game, HT on wins in a landslide.

 

I would never buy a 4770k to turn HT off though. That is a colossal waste of money.

 

Dolphin emulator and MMO's are the only reason I could see to do it though. All about heat/cooling and it might get ya 100 mhz more, but that ain't much. I guess a better test would be with skype or vent on. I will try that with Wild Star when I get the full release version . It might be worth it for an MMO and might not. 

[Vrbas]

HT benefits i3s more then i7s currently for games. Though some games will use i7 hyperthreading to lower the overall CPU usage.

[papeles]

 

So HT doesn't change the maximum stable overclock speed?

No. HT has nothing to do with clocks. You are probably confusing it with turbo boost or SpeedStep?

[deathjester]

No. HT has nothing to do with clocks. You are probably confusing it with turbo boost or SpeedStep?

 

He meant if you turn HT off could you clock higher at the same voltage or achieve higher overclocks. I tried. Answer was no for me, though you can push more voltage and keep temps lower but 100 mhz is about all I can get out of it and for BF4 that is useless. Prob a push, while losing streaming/recording performance. 

[n3krO]

Just tested for ya with tons of crashes lol . I need the exact same voltage for HT and non HT. 4.6 was 1.25vish. Just like my HT. Temps are lower so I guess I could get 100 mhz more on a evo 212 with push/pull with HT off. Was rendering at 71C as opposed to mid 70's with HT on. Rendering score of course it gets smoked by HT. 

 

For BF4? I wouldn't do it. They would prob perform exactly the same and you lose the ability to stream. Dolphin emulator. Yeah might be worth it on some games, especially if you got a crappy overclocking Haswell at like 4.1-4.2. MMO's? I might use the profile in Wild Star. Maybe another 2 FPS with HT off? I would test it but beta is up. Then again if I am streaming or recording the game, HT on wins in a landslide.

 

I would never buy a 4770k to turn HT off though. That is a colossal waste of money.

 

Dolphin emulator and MMO's are the only reason I could see to do it though. All about heat/cooling and it might get ya 100 mhz more, but that ain't much. I guess a better test would be with skype or vent on. I will try that with Wild Star when I get the full release version . It might be worth it for an MMO and might not. 

 

I'm looking foward for  i7 4790k which is rumored to run at 4 Ghz (4.4Ghz boost) and to have better materials for better heat dissipation (as far as i understand it's not only the thermal paste).

 

So disabling HT would only give about 100 mhz extra clock speed? Thanks for your time. If i7 4790k is released at it's rumored price (same as i7 4770k atm) i think i'll get it instead of i5 3690k, 80 eur for extra 500mhz and HT seems quite nice.

[TheProfosist]

You should be able get a higher oc without HT because of less heat.

[deathjester]

I'm looking foward for  i7 4790k which is rumored to run at 4 Ghz (4.4Ghz boost) and to have better materials for better heat dissipation (as far as i understand it's not only the thermal paste).

 

So disabling HT would only give about 100 mhz extra clock speed? Thanks for your time. If i7 4790k is released at it's rumored price (same as i7 4770k atm) i think i'll get it instead of i5 3690k, 80 eur for extra 500mhz and HT seems quite nice.

 

Yeah Devil's Canyon's should be pretty good clockers based on their turbo's and stocks. Current Haswell? I wouldn't really bother water cooling or going all out. 

 

I have seen 4.2 on water on a Asus board. Motherboard doesn't matter. Water can get you like 100-200 more but you just hit a huge wall with heat around 1.3v. The new chips are supposed to run cooler and they are binned higher. If looking for top clocks definately wait.

 

Higher binning plus running cooler they might go like 1.35v safely and that would be pretty nice with a high bin on water.

[Faa]

I'm looking for the best OC you can get on HT and the best OC you can get without HT. And compare how both perfomance on battlefield 4 for example. That's the only way you can really know if HT is worth it or not on that game imo. Not comparing clock to clock....

HT might require a bit more vcore to get stable, thats all. Don't expect that you need a huge vcore bump for it. 99% of the time you won't need a vcore bump

[papeles]

holy balls this thread is giving me cancer. Can some1 explain me whats the relation between hyperthreading and overclock or voltages? I still didn't get what's all this discussion about the stability of whatever when using HT

 

ht: http://www.intel.com/content/www/us/en/architecture-and-technology/hyper-threading/hyper-threading-technology.html

 

It doesn't change anything related to overclock or voltages unless you have a haswell and auto oc settings on bios, wich has nothing to do with HT.

[TheProfosist]

holy balls this thread is giving me cancer. Can some1 explain me whats the relation between hyperthreading and overclock or voltages? I still didn't get what's all this discussion about the stability of whatever when using HT

 

ht: http://www.intel.com/content/www/us/en/architecture-and-technology/hyper-threading/hyper-threading-technology.html

 

It doesn't change anything related to overclock or voltages unless you have a haswell and auto oc settings on bios, wich has nothing to do with HT.

HT crates more heat, thats about it. It was quite ommon to get a i7-920 or a i7-2600K and turn on HT to get a better/bigger OC. This was because when core count was even less important and HT did increase temps quite a bit.

[vm'N]

Hyper-threading, intels implementation of SMT allow the normally unused resources of the core to be utilized. This means that programs have an easier way to utilize the entire core.

 

So with hyper-threading you are utilizing more, which means the core will get hotter and will use more electricity. This is generally in the small amount.

 

So in theory hyper-threading can reduce the overclock of an chip slightly (Almost unnoticeable).

 

However you will get more performance per clock, and in the end you will mostlikely have more performance having hyper-threading enabled than having an core running 100MHz faster.

[Faa]

holy balls this thread is giving me cancer. Can some1 explain me whats the relation between hyperthreading and overclock or voltages? I still didn't get what's all this discussion about the stability of whatever when using HT

 

ht: http://www.intel.com/content/www/us/en/architecture-and-technology/hyper-threading/hyper-threading-technology.html

 

It doesn't change anything related to overclock or voltages unless you have a haswell and auto oc settings on bios, wich has nothing to do with HT.

TL;DR nothing.

If you crash it means you want stability and you need more vcore to maintain stability again. Hyperthreading doesnt require 99,999% of the time more vcore, its a rare case. Anyone who claims it, take it with a grain of salt.