Gaming benchmarks: Core i7 3770 hyperthreading test (20 games tested)

[Artas1984]

Hyper-threading as we know now (nevermind Pentium 4 era) in games has been tested since the day Nehalem Lynnfield Core I5 arrived. In this benchmark we saw a direct comparison between equally clocked I5 750 and I7 920 processors.

Techpowerup actually made similar reviews in the past, like this

These reviews, however contain too few games, so i am bringing this Core i7 VS Core i5 debate once again.

I own I7 3770 and i5 3570K processors, but in this BENCHMARK (not a CPU review mind you) i will simply use I7 3770 without HT as my reference Core i5. This is because I5 3570K has all of it's 4 core clocks at 3.8 GHz in turbo mode, while I7 3770 has only one core at 3.9 GHz in turbo mode, while other cores are lower. Also replacing processors takes time, and i have a very nasty cooler, that is very difficult to dismount...

20 games have been tested for this review. I am using 1920X1080 resolution, since this is the most popular mainstream resolution for mid-range cards like GTX760/GTX960, R9 270X/R9 370X and gamers do not play at 1024x768 rez. anyway despite the fact, that games are more VGA bound at 1920x1080 rez. Game settings are set to highest playable, so if a game does not score at least 40 FPS at highest settings, i lower the settings down. No AA is used in any game except Alan Wake, since AA can not be disabled in that game.


Test setup
Intel Core I7 3770
2X4 GB DDR3 1600 MHz C9
GeForce GTX760 OC 2 GB
Windows 7 Pro 64 bit
Forceware 355.81

For those who prefer video presentation

Let's begin.

Alan Wake American Nightmare (highest settings)



Batman Arkham Origins (highest settings)



Battlefield 4 (highest settings)



Bioshock Infinite Burial at Sea (highest settings)



Company of Heroes 2 (highest settings)



Crysis 3 (high settings)



Far Cry 3 (highest settings)



F.E.A.R. 3 (highest settings)



Formula 1 2013 (highest settings)



Hard Reset (highest settings)



Hitman Absolution (highest settings)



Lost Planet 2 (highest settings)



Max Payne 3 (highest settings)



Metro Last Light Redux (high settings)



Resident Evil 6 (highest settings)



Serious Sam 3 (highest settings)



Starcraft 2 Wings of Liberty (highest settings)



Syndicate (highest settings)



Watch Dogs (highest settings)



Witcher 3 (high settings)



I've made the benchmarks several times in a row and they are as real as you can get.

CONCLUSIONS

Hyper threading (HT) does not improve any notable performance in gaming on smooth gaming settings, despite scoring barely higher maximum frame rates in some games. In fact, HT decreases minimal frame rates by a smal margin in these games: Crysis 3, Metro Last Light Redux, Hard Reset - these were not just random scores, i've made countless attempts into the same pattern.

Now, this does not mean HT is fundamentally non existent in gaming - in low quality resolution and settings HT might just pull a notable lead, but since FPS will be way above 100 FPS, it will not really matter. What matters is saving money for the correct gaming PC. I've bought a used Core I7 3770 to replace my I5 3570K because i need HT for video editing, not gaming.

 

[Ja.KooLit]

I believe that there is no significant gains with HT on. This is because alot of games is not optimized for multi core cpus. Thats where DX12 comes in. Hopefully, dx12 games will utilize all cores thereby alot of improvements with more cores or HT cpus.

 

[Dent1]

I believe that there is no significant gains with HT on. This is because alot of games is not optimized for multi core cpus. Thats where DX12 comes in. Hopefully, dx12 games will utilize all cores thereby alot of improvements with more cores or HT cpus.

Keep in mind HT doesn't constitute as multiple cores. It's a technique not a core.

A lot of the games above do support multi threading. I suspect most of these games are a few years old and the processors reviewed are at the very high end of performance. The games are probably reaching its frame rate limit regardless of whether HT is enabled or disabled. Basically these games are not stressing the CPU enough.

 

[Schmuckley]

Nice review!

 

[Ferrum Master]

well... so we have a strong argument... that lower-mid tier cards does not need i7, well yes... obviously... it would fun to see 970 and 380 as they are the mid tier really now.

 

[FordGT90Concept]

There's a few outliers but for the most part, it seems to not matter either way. The largest deviation (Metro Last Light Redux) in average FPS is 5.5%.

You should rerun Metro Last Light Redux with HTT on to double check those numbers aren't outliers. If the numbers aren't up where they should be, run it again setting the process affinity to core 0, 2, 4, 6. The problem might be that Metro Last Light Redux or Windows isn't load balancing well.

 

[Aquinus]

The same tests with the 3770 set to only 2 cores enabled with and without HT would probably provide some more insightful numbers. It's not really new knowledge that games don't usually benefit from 8 logical threads.

 

[Artas1984]

well... so we have a strong argument... that lower-mid tier cards does not need i7, well yes... obviously... it would fun to see 970 and 380 as they are the mid tier really now.

I have to stop this right here, cause you are misleading other people. There are graphics cards meant for 1920X1080 resolution and there are graphics cards meant for 2500x1440 resolution. GTX970 is meant for higher resolution than 1920X1080, it has been quoted like that by many reviewers, not me, so it would not not make any difference in my setup. The graphics cards i can test atm - GTX760, GTX670, GTX960 are all by now mid-range products meant for no more than 1920X1080 HD gaming. Those are almost equal in performance, and there is no need for me to get a GTX970 for my 24" monitor, unless i want to play Crysis 3 and Witcher 3 on ultra settings. Get it now?

The same tests with the 3770 set to only 2 cores enabled with and without HT would probably provide some more insightful numbers. It's not really new knowledge that games don't usually benefit from 8 logical threads.

Nice idea. Perhaps i should do this.

There's a few outliers but for the most part, it seems to not matter either way. The largest deviation (Metro Last Light Redux) in average FPS is 5.5%.

There is only one outliner in the review - Hard Reset max frame rates peaking over 230 FPS on both processors, which does not really matter.

You should rerun Metro Last Light Redux with HTT on to double check those numbers aren't outliers. If the numbers aren't up where they should be, run it again setting the process affinity to core 0, 2, 4, 6. The problem might be that Metro Last Light Redux or Windows isn't load balancing well.

I already ten times checked it. When i've scored these results i was very surprised too, that is why i did 10 Metro Last Light Redux tests in a row to be absolutely sure that those numbers are not just coincidence. 10 out 10 times the minimal FPS would not go above 49 with HT on.

 

[R00kie]

You wanna see the difference? Get a higher end card, or an SLI setup, then you'll get your difference.

 

[Ferrum Master]

I have to stop this right here, cause you are misleading other people. There are graphics cards meant for 1920X1080 resolution and there are graphics cards meant for 2500x1440 resolution. GTX970 is meant for higher resolution than 1920X1080, it has been quoted like that by many reviewers, not me, so it would not not make any difference in my setup. The graphics cards i can test atm - GTX760, GTX670, GTX960 are all by now mid-range products meant for no more than 1920X1080 HD gaming. Those are almost equal in performance, and there is no need for me to get a GTX970 for my 24" monitor, unless i want to play Crysis 3 and Witcher 3 on ultra settings. Get it now?

No I don't. Because it is low mid tier... it is meant for 1080p... just as 970... as 1080p is the major used resolution by everyone and 970 delivers good 1080p performance at max, and will start to struggle at 1440p, you have to tone down settings. ANNO 2015 970 is mid tier card, that's it.

There is no use to bench such entry level cards for being CPU bond... The CPU single thread is so mighty that it maxes out the card, it is 99% fed with data.

 

[FordGT90Concept]

I already ten times checked it. When i've scored these results i was very surprised too, that is why i did 10 Metro Last Light Redux tests in a row to be absolutely sure that those numbers are not just coincidence. 10 out 10 times the minimal FPS would not go above 49 with HT on.

So try it using the CPU affinity as I said. That would show whether Windows' scheduler is screwing up or if Metro somehow breaks SMT.

 

[CounterZeus]

No I don't. Because it is low mid tier... it is meant for 1080p... just as 970... as 1080p is the major used resolution by everyone and 970 delivers good 1080p performance at max, and will start to struggle at 1440p, you have to tone down settings. ANNO 2015 970 is mid tier card, that's it.

There is no use to bench such entry level cards for being CPU bond... The CPU single thread is so mighty that it maxes out the card, it is 99% fed with data.

I use gtx970 for 1440p with pretty much maxed out games. I would agree if you said 4K. GTX970 is not a mid range card, not even in 2015, nor it delivers just 'good' 1080p performance. This card is complete overkill for 1080p.

 

[EarthDog]

So... where is the review? The thread starts on post #2??????

 

[Artas1984]

No I don't. Because it is low mid tier... it is meant for 1080p... just as 970... as 1080p is the major used resolution by everyone and 970 delivers good 1080p performance at max, and will start to struggle at 1440p, you have to tone down settings. ANNO 2015 970 is mid tier card, that's it.

There is no use to bench such entry level cards for being CPU bond... The CPU single thread is so mighty that it maxes out the card, it is 99% fed with data.

GTX970 might be whatever tier you like Ferrum Master (if you compare it to Titan X - then sure it is a mid range card), but the fact is that the most popular gaming computers where are live right now are made of Core I5 Haswell processors, 2X4 GB RAM and GTX960 2 GB video cards. Makes sense to compare hyper-threading in a similar performance type computer, so that people know whether should they upgrade their Core i5 processors to Core i7 for gaming purposes and i think i've just made their decision quite a lot clearer...

Earthdog, i know you for quite some time now and while i appreciate your advises generally (for instance like not to use AMD CPU for gaming benchmarks among others in the past), your sarcasm sometimes is out of place. I am going to correct the word "review" into a "benchmark", because this obviously not a CPU review. If i wanted to do real full time reviews, i would probably had to change my profession from a med. biologist/technician to a computer technical, PC service worker or a PC magazine reviewer. I do these hardware tests from time to time, why do you always pick on me and expect something "granduar"? Next time do not insult me publicly.

 

[BiggieShady]

This card is complete overkill for 1080p.

Depends if you like your minimal frame rate over 60

 

[Atomic77]

What about the Intel Core i5 4570??? I have played a quite a few games with my setup and only having a Intel HD4600 integrated graphics but of course my settings are not very high.

 

[Ferrum Master]

the fact is that the most popular gaming computers

The fact is based on what? STEAM, the 970 is the second MOST popular used GPU in it's system... ? How to it call it then? The most popular top end card? Please be reasonable. And yes it is a MID tier card, just as manufacturer positioned it price and performance wise versus the top performing product the 980ti. No personal feelings, no philosophy, plains facts. It is the most used dedicated silicon today.

Next time do not insult me publicly.

Nobody is insulting you, please calm down. If you can't take proper reasoning and arguments, then keep silent. You are in between experienced people also, who also have enough professional experience in this area.

Depends if you like your minimal frame rate over 60

Yes agree... stutter is the thing, if the game lags in the 30ties at minimum the card is not enough simple as that. No stable 60FPS for FPS? The card is not enough, despite being only 1080p. Yes there is a solid argument that the games are dated. Even UT3 based Bioshock has horrid min FPS, and no CPU can change it. It lacks GPU horsepower, I won't touch Witcher 3 and GTA5 where we can toss endless power inside. Yes even your tests have custom settings, that bring the situation in incomparable manner with other tests. You loose up the needed compute data needed for the GPU, giving it need for less overhead ie CPU job actually. It is like everyone blames AMD for not using anisotropy, as it causes lower bench results.

And yes the mainstream user will complain just only about that! The game stutters, as it breaks game play. The suggestion,order building a PC for someone is based on such needs. The patient people that don't mind having FPS stutter are a rare breed really. Everyone expects after spending money, maximum - perfect result. So let us keep our professionalism with daily common sense about average user behaviour.

So what should we think? Just gulp up the results and praise... yeah mate... pure biblical truth... or should we actually reason and bring out the real character of the graph, that actually the test won't change no matter what due to certain facts?

 

[RejZoR]

Wait for DirectX 12 where number of threads will matter far more than it does today. Today, even dual core is enough, because they stuff everything on first two threads anyway...

 

[Artas1984]

The fact is based on what?

Based on fact that the most popular gaming Core i5 computers in my country are being sold with a GTX960. And the price range. GTX960 computers are 600-700 EU in shops. GTX970 computers are above 1000 EU. Different market for different people. You think every second person in the world owns a GTX970? People in this planet are not that rich.. You have a wrong impression, because we who are lurking inside such forums like techpowerup have some knowledge about what is what, but believe me - i know tons of people who don't even speak English, want to play Tanks online and are looking for a below 500 Euro PC build - GTX970 is not an option for them.

Nobody is insulting you, please calm down. If you can't take proper reasoning and arguments, then keep silent. You are in between experienced people also, who also have enough professional experience in this area.

I have a history with Earthdog, this is not your businesses and there is no need to calm me as i am calm. And yes, there are far more experienced people in this reviewing business than me, like Earthdog, that's why there was no point of him mocking my benchmark. Even though i am not your regular "computer fan case" - i repair notebooks since 2007 and have sufficient exp in hardware. Just just leave it and continue with the benchmarks.

What about the Intel Core i5 4570??? I have played a quite a few games with my setup and only having a Intel HD4600 integrated graphics but of course my settings are not very high.

So what about it? Seems like a perfect CPU for gaming.

You should rerun Metro Last Light Redux with HTT on to double check those numbers aren't outliers. If the numbers aren't up where they should be, run it again setting the process affinity to core 0, 2, 4, 6. The problem might be that Metro Last Light Redux or Windows isn't load balancing well.

There is no point in that. Obviously what is happening is that when HT is turned on, the main cores do not reach their full clock speed defined by turbo boost. When HT is turned off, the cores work to their maximum clock speed.

 

[FordGT90Concept]

There is no point in that. Obviously what is happening is that when HT is turned on, the main cores do not reach their full clock speed defined by turbo boost. When HT is turned off, the cores work to their maximum clock speed.

If you're so convinced of that, you can test it as well by turning off turbo boost.

SMT has little to do with turbo boost; in fact, they have an inverse relationship: the more load on the processor, the more likely SMT is to improve performance and the less likely turbo boost will be enabled.

 

[Ferrum Master]

Based on fact that the most popular gaming Core i5 computers in my country are being sold with a GTX960

Just chill mate. Don't take it personal. Get a beer. I gave you steam statistics and those are hard evidence. Those allmost 4% of 970 users allmost ~350 000 users. And that is a lot of people. And... If you sum up the 7970 front, that actually is more powerful than 960, despite being 3 years old going with castels... You will get a half a million of active gamers that actually mold the PC gaming front and demand.

If you cry about poor countries, please top mine. We may be poor overall, we have no spare money for defence budget, but we are hard skinned and have endured very hard times also. And those who do their job, just live fine and have the same desires for gaming and art as everyone else here. And I know the ambitions. I also am a hardware repair technician since 2003 and not only in PC hardware, so who cares. So leave the ambitions and epeen level out of here, I beg of you. Your personal clash with other members? Leave the grudge in online deathmatch, instagib.

I agree... 1K system is 970. But now it is a Skylake build? Not worth? The most of 970 owners are upgrade, or older ivy/beenwell owners. Their system cost especially on anniversary pentiums can be so cheap, so they use this delta to buy a better GPU. And I cannot blame them. The CPU dependancy has become weak. Only exception is GTA5. Other that matters are highly clocked two cores really. Just as many said.

 

[Frag_Maniac]

I have to stop this right here, cause you are misleading other people. There are graphics cards meant for 1920X1080 resolution and there are graphics cards meant for 2500x1440 resolution. GTX970 is meant for higher resolution than 1920X1080, it has been quoted like that by many reviewers, not me, so it would not not make any difference in my setup. The graphics cards i can test atm - GTX760, GTX670, GTX960 are all by now mid-range products meant for no more than 1920X1080 HD gaming. Those are almost equal in performance, and there is no need for me to get a GTX970 for my 24" monitor, unless i want to play Crysis 3 and Witcher 3 on ultra settings. Get it now?

That's a bit harsh and inaccurate really. You're implying running anything more powerful than a midrange card at 1080p is pointless. Well, having an AMD card that's more powerful than any of those you mentioned, I can most definitely tell you it is in fact not. There's lots of games I don't run at max settings at 1080p, and with the cards you mentioned, it would be even more the case.

You're also basing your assessment on your size of display to justify your point of view. There are LOTS of people using much bigger than 24" displays now. I play on 32" at 1080p, but that is dwarfed by what a lot of people play on. Even in high end monitors, 27" is quickly becoming the new norm.

Tests like this really need be done with a range of hardware to get an accurate idea, but HT is always been one of those things that is mostly irrelevant to even bother testing because so few games actually make use of it. The last person that argued with me that HT can make a big difference only talked of Crysis 3, and he dropped res from 1080p to 720p to do so, making it a pretty silly statement really.

 

[CounterZeus]

Depends if you like your minimal frame rate over 60

Well, I'm going to play happy on 1440p and let you play at a mere 1080p (a resolution we've been stuck with for way too long) with a 350 euro 'mid range' card.

There will always be new future proof games or terribly coded ones that will kick you in the nuts at pretty much every resolution. And yes the longer you wait, the more valid your statement becomes. For example, I was playing every game at 1080p with a 9600GT all those years ago. Now that card is obsolete. In another year, I might have to drop some eye candy to get stable fps on 1440p with a gtx970.

 

[GreiverBlade]

Very nice review, that comfort me in my 6600K choice over a 6700K.

Thanks!

 

[RCoon]

I don't understand why you turned off AA in every game for your benchmarks.

Also, OCAholic did it with separate processors as opposed to disabling HT: http://www.ocaholic.ch/modules/smartsection/item.php?itemid=1061&page=13

Similar results to yours, but thats to be expected. Almost no game uses more than four cores. The only people who buy i7's instead of i5's are usually those that need them for different reasons other than gaming.