CPU Hyperthreading and Gaming: Real-World Benchmarks

[grimpr]

My findings are as follows: all things being equal, in modern games, HT will either have no negative impact or will have a significant positive impact, especially as to minimums, which determine the smoothness of the gameplay. The benefit of HT is more pronounced on slower CPUs than on faster CPUs and on dual-cores than on quad-cores. When building a computer, I believe HT is a good investment when taking a long-term perspective on the usage of the system, and in fact, as between a fast dual-core with HT and a slower quad-core without HT, I might suggest the faster dual-core as the more cost-effective solution. Finally, a dual-core without HT simply isn't equipped to handle modern games with even a ~$100 GPU, let alone a higher-end GPU. I would not recommend, for instance, the current Pentiums for a gaming system.

Well said, thats the truth. I too found HT essential on an Intel dualcore for gaming and desktop usage, The Core i3 is truely worth the extra money compared to the Pentium and lower models, if only Intel would release a K model at 120$...

 

[2is]

again its a FACT that GTA 4 only uses four threads. I know from testing Deus Ex that going beyond 3 cores did zero and even 2 cores at 4.4 gave nearly the same performance as 4 cores at 4.4. and Toms even said Deus Ex does not go beyond four threads in their review. and I have spent hours testing Just cause 2 and know darn well that it does basically does NOT scale beyond 2 cores at 4.4 even at just 1024x768. your results are screwy.

One guy with extensive data says one thing, also welcomes the ther guy to post his data.

Other guy posts no data, just his own personal anecdotal evidence and claims the other guys reaults are screwy....

And yes, I know you posted Fraps numbers..... Later

Also, you can still benefit a good deal from say a quad core even on games that only may scale to 2 or 3 cores. There is an OS there as well as dozens of oter processes as well that need CPU cycles.

 

[toyota]

One guy with extensive data says one thing, also welcomes the ther guy to post his data.

Other guy posts no data, just his own personal anecdotal evidence and claims the other guys reaults are screwy....

And yes, I know you posted Fraps numbers..... Later

Also, you can still benefit a good deal from say a quad core even on games that only may scale to 2 or 3 cores. There is an OS there as well as dozens of oter processes as well that need CPU cycles.

AGAIN reviews show the same thing that I did and I did show FRAPS run to prove it. I also tested Just Cause 2 plenty of times so I know darn well that HT on quad is going to do nothing for that game as even my oced E8500 was not limiting it. I am not going to argue unless I have seen official reviews and/or I have already tested something myself so dont act like I am disagreeing for no reason.

and yes I know a quad can help out with overall system but the POINT was that it was not making any difference in Just Cause 2 or Dues Ex so how would HT on top of that help the game?

 

[Dark Shroud]

Well said, thats the truth. I too found HT essential on an Intel dualcore for gaming and desktop usage, The Core i3 is truely worth the extra money compared to the Pentium and lower models, if only Intel would release a K model at 120$...

Intel isn't going to do that because it would eat up sales of i5s.

This is what makes the Trinity A10 so great. It gives clear competition to the i3.

 

[Termie]

Hey all, I knew this would be controversial when I set out to put this test together, but I thought it would be because HT would have a big effect in BF3. In the end it was in that game that HT had the least impact of all, and the games I found a difference in I was totally surprised at.

Some have questioned my methods, some have questioned my results. At this point I'm just going to let it be. I spent many, many hours on this and have presented my approach as clearly as I could. Perhaps someone else will decide it's worth doing a follow-up test of hyperthreading on quad-core processors, just as another helpful user posted his results on a dual-core.

For now I'm going back to my gaming!

 

[Rvenger]

Termie, I respect all the time and effort you put into this thread and I think your numbers are accurate. I think I might actually run some tests and we can put them up against yours too. I just need to get some time to do so.

 

[Denithor]

Termie - good stuff in here!

I wonder what would happen if you ran the GTX 670 on the i7 860 with/without HT enabled? Would the differences be about the same or higher? I have a feeling the later, as with the older CPU making more efficient use of threads matters more than on today's SB/IB chips.

 

[wand3r3r]

I don't have enough games to test but I do have a 6 core 1366 which I could test if there are demos or other repeatable scenarios to compare 1 to 6 cores w/ w/out hyperthreading. I do have BF3 but apparently that is a moot point as you found it to have no affect using 4 cores vs. 4 cores w/ HT?

Anyways, thanks for the effort. It's food for thought at the least, informative at any level.

I too am quite skeptical of "review" sites as they seem to follow the directions of the sponsor and my results tend to be a lot more negative then reviews proclaim.

Borderlands 2 is a perfect example of that, as well as BF3. Both vary dramatically depending on the number of players and FPS drop tremendously even on cards like the gtx 690, but a review shows the 690 purring along at high FPS. Yeah single player may work fine but it doesn't tell the story.

I would agree that hyperthreading is another one of those obscure areas and is apparently difficult to measure the effect accurately.

 

[guskline]

As mentioned above, Great job Termie. You need a chuckle after all that work. How about I run the tests on a 8150 Bulldozer since I don't have hyperthreading but rather "more coars" or is that more Coors?

 

[Termie]

Termie, I respect all the time and effort you put into this thread and I think your numbers are accurate. I think I might actually run some tests and we can put them up against yours too. I just need to get some time to do so.

Given that you have basically the same system, it would be interesting to see if you come up with similar numbers.

Termie - good stuff in here!

I wonder what would happen if you ran the GTX 670 on the i7 860 with/without HT enabled? Would the differences be about the same or higher? I have a feeling the later, as with the older CPU making more efficient use of threads matters more than on today's SB/IB chips.

I definitely think there's a pattern of HT helping more with slower CPUs, and that would be amplified by a stronger GPU. Unfortunately I cannot fit the 670 in my HTPC case, so I can't test it directly.

I don't have enough games to test but I do have a 6 core 1366 which I could test if there are demos or other repeatable scenarios to compare 1 to 6 cores w/ w/out hyperthreading. I do have BF3 but apparently that is a moot point as you found it to have no affect using 4 cores vs. 4 cores w/ HT?

Anyways, thanks for the effort. It's food for thought at the least, informative at any level.

I too am quite skeptical of "review" sites as they seem to follow the directions of the sponsor and my results tend to be a lot more negative then reviews proclaim.

Borderlands 2 is a perfect example of that, as well as BF3. Both vary dramatically depending on the number of players and FPS drop tremendously even on cards like the gtx 690, but a review shows the 690 purring along at high FPS. Yeah single player may work fine but it doesn't tell the story.

I would agree that hyperthreading is another one of those obscure areas and is apparently difficult to measure the effect accurately.

Thanks for that BL2 testing you did - it was exactly the kind of thing enthusiasts can be adding to the body of knowledge, and which was notably absent in professional reviews, at least at the beginning.

Also, to any of you who are interested in conducting similar testing (guskline included!), let me offer four tips:
(1) don't bother with BF3 multiplayer benchmarking unless you have lots of time and patience! wand3r3r - you have a 690, and I'd really like to see more testing with that much GPU power, similar to what was done by user Moonbogg on a system almost identical to yours (except he didn't actually play, he just benched as he looked out over the map, and now I know why!). If you have the time, go for it.
(2) for single-player games, choose a game and level you know very well, so you don't have to find a path that won't be interrupted by unpredictable events in game (or cutscenes), or actually risk getting lost. You should be able to do it in your sleep or it's just too hard to repeat.
(3) try to do 3 identical runs of 3-4 minutes each. I chose four different runs of 1 minute each and then attempted to repeat them at the other setting, which was a lot of work and probably caused too much variation. The difficulty, as noted above, will be finding a 3-4 minute run that is not interrupted in different ways each time.
(4) watch out for in-engine fps caps. I put about 30 minutes into testing and reviewing data on Batman:AC only to realize that it has a 22-62fps smoothing cap built in, even without vsync. I had to scrap the B:AC data altogether, because I didn't want to mess with config files or re-running my benches.

If you have graphs or data, I'm happy to post it in the OP.

 

[Madrox]

AGAIN reviews show the same thing that I did and I did show FRAPS run to prove it. I also tested Just Cause 2 plenty of times so I know darn well that HT on quad is going to do nothing for that game as even my oced E8500 was not limiting it. I am not going to argue unless I have seen official reviews and/or I have already tested something myself so dont act like I am disagreeing for no reason.

and yes I know a quad can help out with overall system but the POINT was that it was not making any difference in Just Cause 2 or Dues Ex so how would HT on top of that help the game?

You guys made me curious about the possible effect of HT in older games, also while we discuss this in our own forum (different language).

So i downloade DEHR an Fraps and made a serie of 60sec runs, just like the TS.

My config: 3.2Ghz Core i7 920, 6GB DDR 3, Asus X58-E motherboard, HD7850, 80GB intel SSd and a six-disk Raid-5. OS= Win7 64bit. Drivers: latest...


First scores are with Vsync ON, default graphics settings.

HT-ON
2013-06-12 04:52:30 - dxhr
Frames: 3596 - Time: 60000ms - Avg: 59.933 - Min: 58 - Max: 61
2013-06-12 04:53:54 - dxhr
Frames: 3598 - Time: 60000ms - Avg: 59.967 - Min: 58 - Max: 62
2013-06-12 04:55:01 - dxhr
Frames: 3598 - Time: 60000ms - Avg: 59.967 - Min: 59 - Max: 61
2013-06-12 05:01:46 - dxhr

HT-OFF
Frames: 3583 - Time: 60000ms - Avg: 59.717 - Min: 58 - Max: 62
2013-06-12 05:02:56 - dxhr
Frames: 3597 - Time: 60000ms - Avg: 59.950 - Min: 58 - Max: 62
2013-06-12 05:04:03 - dxhr
Frames: 3598 - Time: 60000ms - Avg: 59.967 - Min: 58 - Max: 62

Nice....

Oke, Vsync OFF and game settings all on maximum with FXAA.


HT-OFF (while i was in non-HT mode i run these test first this time).

2013-06-12 05:30:28 - dxhr
Frames: 6256 - Time: 60000ms - Avg: 104.267 - Min: 85 - Max: 115
2013-06-12 05:31:37 - dxhr
Frames: 6121 - Time: 60000ms - Avg: 102.017 - Min: 82 - Max: 117

Oke, the HT test i still need to do but wat strikes me is this:

How can my infeririour system score a min FPS of atleast 80, while your newer and stronger setup drops to 40/50 min-FPS!

Myth Busted ???

 

[toyota]

what?

I guess you are referring to the OP and not me.

 

[Termie]

...

HT-OFF (while i was in non-HT mode i run these test first this time).

2013-06-12 05:30:28 - dxhr
Frames: 6256 - Time: 60000ms - Avg: 104.267 - Min: 85 - Max: 115
2013-06-12 05:31:37 - dxhr
Frames: 6121 - Time: 60000ms - Avg: 102.017 - Min: 82 - Max: 117

Oke, the HT test i still need to do but wat strikes me is this:

How can my infeririour system score a min FPS of atleast 80, while your newer and stronger setup drops to 40/50 min-FPS!

Myth Busted ???

This is a necro-thread, so I'm not going to spend a lot of time responding, but in short, the i7-3770K/GTX670 combo crushed your system, while my i7-860@3.2/GTX460 lost pretty badly, but of course that's because you're using an HD7850.

Also, the level you test in Deus Ex makes a huge difference. You can't just compare any scene in the game to another.

If you're interested in this topic, you should view this much newer thread: http://forums.anandtech.com/showthread.php?t=2322595

 

[Maximilian]

Pretty good thread to necro tbh.

Still relevant :thumbsup:

 

[moonbogg]

Well, the thread got resurrected, but the only thing that can bring my interest in this kind of testing back from the dead is BF4. Soon enough fellas. Soon enough.

 

[Borealis7]

Crysis 3 loves them threads let me tell ya.

 

[grimpr]

Well, the thread got resurrected, but the only thing that can bring my interest in this kind of testing back from the dead is BF4. Soon enough fellas. Soon enough.

Seems AMD is pushing hard for heavy multithreading, good news for all multicore users!

"Today we're proud to announce that Battlefield 4 will be exclusively optimized for AMD Radeon GPUs, AMD A-Series APUs and AMD FX CPUs!"

 

[BrightCandle]

The gamegpu.ru data suggests the following games benefit the most from Hyperthreading:

Far Cry 3
Starcraft 2
Battlefield 3
Crysis 3
Medal of Honor Warfighter
Planetside 2
Guild wars 2

Their data also shows the 3930k beating out the 2500k by more than 10% in the current games (and basically the same list against the 2600k):

Metro last light
Marvel heroes
Far Cry 3
Battlefield 3
Crysis 3
Hitman Absolution
Need for speed most wanted
Medal of Honor Warfighter
Project cars
Planetside 2


The biggest gainers in terms of threads of this lot are:

Metro last light
Crysis 3
Hitman absolution
project cars

These would be the games that at least with SLI show the highest gain in performance and hence likely are the best games to test in more detail.

What I find interesting from the course data that gamegpu.ru gives me is that the HT benefitting games are not the same as the games that benefit from 6 cores. There is an overlap of course but just because a game benefits from HT does not mean it benefits from 6 cores (Starcraft 2), and those that benefit from 6 cores do not necessarily benefit from HT (Project cars).

The biggest loser incidentally is warframe, where a 3930k v 2500k looses 10% of its frame rate.

 

[Madrox]

This is a necro-thread, so I'm not going to spend a lot of time responding, but in short, the i7-3770K/GTX670 combo crushed your system, while my i7-860@3.2/GTX460 lost pretty badly, but of course that's because you're using an HD7850.

Like i said, your system is superiour.

ps: what is necro?

Also, the level you test in Deus Ex makes a huge difference. You can't just compare any scene in the game to another.

Oke, what scene did you test, can you tell where i can find it and do the test again?

If you're interested in this topic, you should view this much newer thread: http://forums.anandtech.com/showthread.php?t=2322595

[/quote]
Thank you.

 

[Termie]

Like i said, your system is superiour.

ps: what is necro?


Oke, what scene did you test, can you tell where i can find it and do the test again?

Thank you.[/QUOTE]

The level I tested was the Hengsha landing pad, and I ran around the lower levels for 60 seconds. Passing an area with a large flame coming from a kitchen interestingly sent every CPU I've tested into a nose-dive on minimums, and on a dual-core without HT, it almost ground to a halt.

Having tested 8 different configurations in Deux Ex, I can say with complete certainty that HT always performs better in this game than no HT.

"Necro" means that this thread was posted a long time ago, in Oct. 2012 in fact.

 

[guskline]

I'm also glad this thread is being "resurrected". The last time I posted on this thread I had a 2500k rig @4.5Ghz and a 8150 rig @4.6Ghz. Since then I've swapped out the 8150 for a 8350 at the same clockspeed and swapped out the 2500k for a 3770k running just a little slower at 4.4Ghz. In gaming, even with the same video card used (GTX680) the 3770k @ 4.4Ghz always has higher fps than the FX8350 @ 4.6Ghz. The difference was clearly detectable in the charts but not as obvious in the "feel" of game play.

Perhaps later this summer we will get to read a review of AMD's 9570 chip (PileDriver 8 core @4.7Ghz turbo to 5ghz). I don't know if that will make much difference.

For those owning the "big boys" - Socket 2011 3930k/3960k/3970k I sure wouldn't be worried. You have the horsepower to run any game well 6 "true cores" and 12 threads.

 

[toyota]

The gamegpu.ru data suggests the following games benefit the most from Hyperthreading:

Far Cry 3
Starcraft 2
Battlefield 3
Crysis 3
Medal of Honor Warfighter
Planetside 2
Guild wars 2

Their data also shows the 3930k beating out the 2500k by more than 10% in the current games (and basically the same list against the 2600k):

Metro last light
Marvel heroes
Far Cry 3
Battlefield 3
Crysis 3
Hitman Absolution
Need for speed most wanted
Medal of Honor Warfighter
Project cars
Planetside 2


The biggest gainers in terms of threads of this lot are:

Metro last light
Crysis 3
Hitman absolution
project cars

These would be the games that at least with SLI show the highest gain in performance and hence likely are the best games to test in more detail.

What I find interesting from the course data that gamegpu.ru gives me is that the HT benefitting games are not the same as the games that benefit from 6 cores. There is an overlap of course but just because a game benefits from HT does not mean it benefits from 6 cores (Starcraft 2), and those that benefit from 6 cores do not necessarily benefit from HT (Project cars).

The biggest loser incidentally is warframe, where a 3930k v 2500k looses 10% of its frame rate.

for Hitman Absolution using only 3 cores of my 2500k gave only 2 less fps as 4 cores. I see no way that HT can help on top of a quad.

 

[Essence_of_War]

The gamegpu.ru data suggests the following games benefit the most from Hyperthreading:

Far Cry 3
Starcraft 2

That seems confusing to me...I thought that SC2 basically used only 2 cores. Is this from HT on an i3, or HT on an i7?

 

[ShintaiDK]

The gamegpu.ru data suggests the following games benefit the most from Hyperthreading:

Far Cry 3
Starcraft 2
Battlefield 3
Crysis 3
Medal of Honor Warfighter
Planetside 2
Guild wars 2

Their data also shows the 3930k beating out the 2500k by more than 10% in the current games (and basically the same list against the 2600k):

Metro last light
Marvel heroes
Far Cry 3
Battlefield 3
Crysis 3
Hitman Absolution
Need for speed most wanted
Medal of Honor Warfighter
Project cars
Planetside 2


The biggest gainers in terms of threads of this lot are:

Metro last light
Crysis 3
Hitman absolution
project cars

These would be the games that at least with SLI show the highest gain in performance and hence likely are the best games to test in more detail.

What I find interesting from the course data that gamegpu.ru gives me is that the HT benefitting games are not the same as the games that benefit from 6 cores. There is an overlap of course but just because a game benefits from HT does not mean it benefits from 6 cores (Starcraft 2), and those that benefit from 6 cores do not necessarily benefit from HT (Project cars).

The biggest loser incidentally is warframe, where a 3930k v 2500k looses 10% of its frame rate.

Are they use dualcores vs dualcore with HT as baseline? Else something is terrible wrong in their measurement. If they think games like GW2 and SC2 works better with 4+ threads.

I have a feeling its their setup, rather than the games that utilize more threads.

 

[aigomorla]

i have told u guys this for many many years now.

HT is NOT a real core.
It translates to roughly 1/3 or 1/2 the equvilant of a real core.

If work is done on a HT core, it is slower then work being done on a real core.

That is why ur seeing less FPS on a HT enabled CPU which has fetching off.
The work being fetch shouldnt be done on a HT thread unless the real core was already loaded.

Typically this doesnt happen on anything higher then a quadcore, as very very few games today even can use more then a quadcore.

Fetching issues should of been fixed and resolved, which is why u dont see as big as a difference as u used to see on a i7 920 with HT ON.