• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Gaming benchmarks: Core i7 6700k hyperthreading test (2018 update with GTX 1080 Ti)

Joined
Feb 22, 2009
Messages
698 (0.18/day)
System Name Artas1984
Processor Ryzen 5 1600
Video Card(s) Sapphire Nitro R9 Fury OC
Storage Plextor M9PeG 512 GB
Display(s) Dell S2415H
Case Fractal Design Define XL Titanium
Audio Device(s) Asus Xonar Essence ST
Power Supply Evga 750 GQ
Mouse Razer DeathAdder 2013
Keyboard Roccat Ryos MK Glow
Software Windows 10 Pro 1809
Benchmark Scores i've got a shitload of them over 10 years of work
Back in January 2016 i made a test about Intel Core i7 6700K virtual cores performance numbers in gaming. Short story back then was that hyperthreading offered no FPS improvement in any of my tested 21 games. Not only that, but about half of the tested games showed slight performance decrements with HT on.:

https://www.techpowerup.com/forums/threads/gaming-benchmarks-core-i7-6700k-hyperthreading-test.219417/

Things have changed a lot since then:

1. 17 new games have been added, way more modern in terms of performance demands, and only the most demanding 3 games back from the previous test will be retested per new.
2. GTX 1080 Ti will replace GTX 780 Ti to ensure no bottleneck on part of GPU will happen.
3. Seeing as Core i5 8400 beats Core i7 6700K in every modern game i have seen all over the internet benchmarks, i have to wonder is this not only because of the additional two cores of Core i5, but also due to the fact that virtual cores of Core i7 might actually hamper the performance in some games? Because we know that was the case in previous tests.

What i actually expect now is to see some performance gains from Core i7 hyperthreading just like from Core i3 hyperthreading (which was noted a millennia ago). It is now 10 years since Core i7 {4 cores/ 8 threads} processors have their name etched, but only recently compared to the whole 10 year perdiod we finally got our ''4+4'' HT supportive games like Ashes of Singularity, so let's see how further an overclocked GTX 1080 Ti can push CPU HT..

TEST SETUP

Intel Core i7 6700K 4.5 GHz
Asus Maximus 8 Ranger
Kingston Hyperx Fury 2X8 GB DDR4 2666 MHz C14
Reference GeForce GTX 1080 Ti OC {maximum boost clock 1974 MHz}

Windows 7 Pro 64 bit
NVIDIA 411.70


---------------------------------------------------------------------------------------------------------------------------

ASHES OF SINGULARITY



Starting with a fact that is world wide known by now, HT does increase performance. This is the only game in my list where CPU utilization is higher than GPU - 99 % vs. ''only'' 70 %... In non of the later games will CPU utilization reach over 50 %, while GPU will vary a lot with most times maxed out.

BATTLEFIELD 1



Reduce image quality, play in multi player, and you will likely see even more HT gains. Testing many times in a row, it is fair to say, that the evidence of HT performance gains is unquestionable.

CALL of DUTY WW2



Testing many times in row i must state that HT hurts performance very notably. WTF - in one of the latest and most demanding games of the franchise this is unacceptable!

COMPANY OF HEROES 2



Retesting a very demanding RTS confirms what was observed in my previous 2016 benchmark - HT hurts performance slightly. This is no margin of error, the benchmark results are very constant. With AA disabled the difference is even greater. CoH 2 built in benchmark is the most stable i have ever tested in any game - with variables lower than 1 % in each successive test.

CRYSIS 3



Retesting the benchmark god game again confirms the results of the 2016 test - constant performance drops with HT on. GTX 780 Ti was way too weak to handle Crysis 3 over 60 FPS minimum at max settings anyway. Having double checked the ''old'' games, let's move to more modern titles.

DEUS EX MANKIND DIVIDED



Core i7 4/4, having done many tests in a row, never reached the top results like Core i7 4/8 did, so i have to assume HT does increase performance here.

DIRT 4



The results are within the margin of error.

DOOM



To confirm the abysmall performance of HT, i've done tests in several other levels, but with the same shocking outcome. HT is horrible in Doom.

F1 2017



HT increases performance for sure. The F1 game built in benchmark is quite stable and repetitive anyway.

FAR CRY 5



Because the max FPS are the same, i can not confirm that HT hurts performance here, but it definitely does not help. The built in benchark of Far Cry 5 seems to be very constant with results.

GHOST RECON WILDLANDS



Together with the built in benchmark i also tested the single player launch checkpoint as well as made tests on lowest preset possible to confirm that there is absolutely no performance gain with HT on.

MASS EFFECT ANDROMEDA



To think how GTX 1080 Ti performance might be hampered just due to the HT factor is sad. Second worse case after Doom.

PREY



Great minimal FPS gain with HT on, but you would not notice any difference unless you tweak the cfg, as the game runs at 144 FPS. I had to unlock 300 FPS to see the difference.

PROJECT CARS 2



No difference.

RISE OF TOMB RAIDER



The results are within the margin of error. Equal performance i'd say.

SNIPER ELITE 4



When you have got that many FPS, it does not matter, but HT does the job it was foretold to do in ages past.

TOTAL WAR WARHAMMER 2



Yet another very stable and constant build in benchmark shows clear HT performance gains. Enabling 4X AA would actually bottleneck GTX 1080 Ti, and show no difference of CPU performance! Brutal.

VANISHING OF ETHAN CARTER REDUX



Redux means that the game has been re-released in 2015 with Unreal4 engine instead of the original 2014 release with Unreal3 engine, and that shows - HT clearly adds to FPS and this is no margin error, because the results are very very constant within each run - great optimization.

WATCH DONGS 2



This must be a known case and i am glad to confirm that HT improves performance in Watch Dogs 2.

WITCHER 3



The final game is a revisit from the previous 2016 benchmark, and just as before we see that there is no performance difference whatsoever. This clearly had to be done, as GTX 780 Ti GHz Edition and GTX 980 G1 Gaming were averaging 59 FPS and 64 FPS, respectively, and clearly were holding the game compared to GTX 1080 Ti.
--------------------------------------------------------------------------------------------------------------------------------------------------------------

Unlike before, there is no general conclusion after this benchmark as what to think of the past generation Core i7 {4 cores/ 8 threads} processors hyperthreading these days. Many games show clear evidence of HT performance gains, but there are still so many games in which HT absolutely sucks. We have waited so long for Core i7 HT to gain ground on gaming, that eventually Core i5 {6 cores/ 6 threads } processors appeared and bested all existing previous generation Core i7 {4 cores/ 8 threads} processors in gaming! I guess what can i really say in the end is ''know your game'': do research before committing on long playing hours, recording and walkthroughs on your games with your Core i7.
 
Last edited:
Joined
Mar 18, 2015
Messages
2,168 (1.30/day)
Location
Long Island
Starting with a fact that is world wide known by now, HT does increase performance
Much like another axiom "th whole world knows 4 GB is not enough " (at 1080p) .... If i asked 10 people from within my tech circle, I'd expect at most 1 to concur with those two statements and at most 1 to say the opposite to both statements . So I guess I'm m not living in that world. I would expect 9 to say "It depends on several factors."

1. You mentioned "game runs at 144 hz" but is this being tested on your Dell S2415H 60 Hz Monitor or something else ? If so, other than AoS, I wouldn't consider the difference relevant.

2. But i think there's one factor that must be addressed. On my son's box for example, we have 3 BIOS profiles selectable via profile option option in in BIOS:

1. All Stock
2. Highest stable PC with HT ON (4.6 Ghz)
3. Highest stable PC with HT OFF (4.8 Ghz)

At the same OC (4.6 GHz), the cores ran 8 C hotter with HT ON. Some games will be problematic at 4.8 ... not many. But where HT leads i thinks this will eat into the advantage as the higher OC will negate some of that impact.
 
Joined
Feb 22, 2009
Messages
698 (0.18/day)
System Name Artas1984
Processor Ryzen 5 1600
Video Card(s) Sapphire Nitro R9 Fury OC
Storage Plextor M9PeG 512 GB
Display(s) Dell S2415H
Case Fractal Design Define XL Titanium
Audio Device(s) Asus Xonar Essence ST
Power Supply Evga 750 GQ
Mouse Razer DeathAdder 2013
Keyboard Roccat Ryos MK Glow
Software Windows 10 Pro 1809
Benchmark Scores i've got a shitload of them over 10 years of work
2. But i think there's one factor that must be addressed. On my son's box for example, we have 3 BIOS profiles selectable via profile option option in in BIOS:

1. All Stock
2. Highest stable PC with HT ON (4.6 Ghz)
3. Highest stable PC with HT OFF (4.8 Ghz)

At the same OC (4.6 GHz), the cores ran 8 C hotter with HT ON. Some games will be problematic at 4.8 ... not many. But where HT leads i thinks this will eat into the advantage as the higher OC will negate some of that impact.
It does run hotter with HT on, right... I've noticed this with many other Core i7 processors. Anything above 4.5 GHz in my tested friend's system layed too much heat on his Dark Rock heatsink, and also was not Prime95 ready. I had to settle with 4.5 GHz.

1. You mentioned "game runs at 144 hz" but is this being tested on your Dell S2415H 60 Hz Monitor or something else ? If so, other than AoS, I wouldn't consider the difference relevant.
Ups, minor mistake, i meant 144 FPS. Will correct that.
 
Joined
Jul 31, 2016
Messages
66 (0.06/day)
Processor i7 3770k @4.3GHZ
Motherboard Asus P8Z77-V
Cooling Hyper 212 Evo
Memory g.skill Sniper 1600 4GB x4
Video Card(s) Powercolor RX480 4GB
Storage Crucial MX100 x2
Display(s) LG 25UM56-P 25"
Case Antec 1100
Audio Device(s) Soundblaster Z
Power Supply Corsair HX750
Mouse Roccat Kone EMP
Keyboard CM Storm Trigger
Benchmark Scores https://www.3dmark.com/3dm/27435174
One game that I found is helped significantly by HT is Kingdom Come.
I upgraded from a 2500k to a 3770k and gained in the starter town:
9 minimum FPS (46 to 55)
7 Average (53 to 60)

In some areas it was even more but I didn't bench there.

Of course this was with an RX 480 and not a 1080 ti.


Also FYI you typo'd "watch dogs" and it's pretty funny :laugh:
 
Joined
Feb 22, 2009
Messages
698 (0.18/day)
System Name Artas1984
Processor Ryzen 5 1600
Video Card(s) Sapphire Nitro R9 Fury OC
Storage Plextor M9PeG 512 GB
Display(s) Dell S2415H
Case Fractal Design Define XL Titanium
Audio Device(s) Asus Xonar Essence ST
Power Supply Evga 750 GQ
Mouse Razer DeathAdder 2013
Keyboard Roccat Ryos MK Glow
Software Windows 10 Pro 1809
Benchmark Scores i've got a shitload of them over 10 years of work
The time is over for editing, so i will post the created video link here:

 
Joined
Aug 6, 2017
Messages
4,644 (5.79/day)
Location
Poland
Processor i7 5775c @4.3GHz/1.385v/EDRAM @2GHz
Motherboard Z97X Gaming 5
Cooling Noctua D15S
Memory Crucial Ballistix Tactical LP 1600 CL8 @2133 9-9-9-27 1T
Video Card(s) RTX 2070 Super Gaming X Trio
Storage SU900 128 (OS)/850 PRO 256+256+ 512,860 EVO 500,XPG SX950U 480,M9Pe(Y) 512 (games)/4TB HDDs (3+1)
Display(s) Acer XB241YU+Dell S2716DG dual monitor setup
Case Full tower
Audio Device(s) W830BT headphones
Power Supply Superflower Leadex Gold 850W
Mouse G903 lightspeed+powerplay,G403 wireless + Steelseries DeX + Roccat rest
Keyboard Razer Deathstalker
Software Windows 10
Benchmark Scores A LOT
"WATCH DONGS 2 "

 
Joined
Dec 6, 2005
Messages
10,716 (2.12/day)
Location
Manchester, NH
System Name Working on it ;)
Processor I7-4790K (Stock speeds right now)
Motherboard MSI Z97 U3 Plus
Cooling Be Quiet Pure Rock Air
Memory 16GB 4x4 G.Skill CAS9 2133 Sniper
Video Card(s) GIGABYTE Vega 64 (Non Reference)
Storage Samsung EVO 500GB / 8 Different WDs / QNAP TS-253 8GB NAS with 2x2Tb WD Black
Display(s) 34" LG 34CB88-P 21:9 Curved UltraWide QHD (3440*1440) *FREE_SYNC*
Case Rosewill Challenger
Audio Device(s) Onboard + HD HDMI
Power Supply Corsair HX750 (love it)
Mouse Logitech G5
Keyboard Corsair Strafe RGB & G610 Orion Red
Software Win 10 upgraded from Win 7 Pro
Nice work! Conclusion, HT definitely adds overhead for games that aren't thread dependent? No test with BF4? I recall that being one of the most thread dependent game out there.
 
Joined
Feb 22, 2009
Messages
698 (0.18/day)
System Name Artas1984
Processor Ryzen 5 1600
Video Card(s) Sapphire Nitro R9 Fury OC
Storage Plextor M9PeG 512 GB
Display(s) Dell S2415H
Case Fractal Design Define XL Titanium
Audio Device(s) Asus Xonar Essence ST
Power Supply Evga 750 GQ
Mouse Razer DeathAdder 2013
Keyboard Roccat Ryos MK Glow
Software Windows 10 Pro 1809
Benchmark Scores i've got a shitload of them over 10 years of work
Nice work! Conclusion, HT definitely adds overhead for games that aren't thread dependent? No test with BF4? I recall that being one of the most thread dependent game out there.
I did a test in Battlefield 4 single player, but my results might have been bottle-necked with GTX 760:

https://www.techpowerup.com/forums/threads/gaming-benchmarks-core-i7-3770-hyperthreading-test-20-games-tested.216466/

Back then i found no difference. That weak GTX 760 was the main point i wanted to redo this test with a GTX 1080 Ti...

"WATCH DONGS 2 "
Could be worse...
 
Joined
Dec 6, 2005
Messages
10,716 (2.12/day)
Location
Manchester, NH
System Name Working on it ;)
Processor I7-4790K (Stock speeds right now)
Motherboard MSI Z97 U3 Plus
Cooling Be Quiet Pure Rock Air
Memory 16GB 4x4 G.Skill CAS9 2133 Sniper
Video Card(s) GIGABYTE Vega 64 (Non Reference)
Storage Samsung EVO 500GB / 8 Different WDs / QNAP TS-253 8GB NAS with 2x2Tb WD Black
Display(s) 34" LG 34CB88-P 21:9 Curved UltraWide QHD (3440*1440) *FREE_SYNC*
Case Rosewill Challenger
Audio Device(s) Onboard + HD HDMI
Power Supply Corsair HX750 (love it)
Mouse Logitech G5
Keyboard Corsair Strafe RGB & G610 Orion Red
Software Win 10 upgraded from Win 7 Pro
I did a test in Battlefield 4 single player, but my results might have been bottle-necked with GTX 760:
I'm guessing that's the case then, interesting.
 
Joined
Nov 13, 2007
Messages
7,537 (1.73/day)
Location
Austin Texas
System Name _
Processor 8700K @ 5.1 Ghz
Motherboard MSI Z370-A PRO
Cooling 120mm Custom Liquid
Memory 32 GB 3600 Mhz DDR4 16-16-16-36-380 trfc - 2T
Video Card(s) Gigabyte GTX 2080 Ti Windforce (Undervolted OC 1905MHz)
Storage 3x1TB SSDs
Display(s) Alienware 34" 3440x1440 120hz, G-Sync
Case Jonsbo U4
Audio Device(s) Bose Solo
Power Supply Corsair SF750
Mouse silent click gaming mouse
Keyboard tenkeyless
Software Windows 10 64 Bit
its kind of amazing how little difference it makes. DOOM being an exception;

I leave the HT off on the 7820x as the extra 300Mhz OC room (4.7ghz va 4.95ghz) definitely feels overall faster than the extra 8 threads.
 
Joined
Feb 22, 2009
Messages
698 (0.18/day)
System Name Artas1984
Processor Ryzen 5 1600
Video Card(s) Sapphire Nitro R9 Fury OC
Storage Plextor M9PeG 512 GB
Display(s) Dell S2415H
Case Fractal Design Define XL Titanium
Audio Device(s) Asus Xonar Essence ST
Power Supply Evga 750 GQ
Mouse Razer DeathAdder 2013
Keyboard Roccat Ryos MK Glow
Software Windows 10 Pro 1809
Benchmark Scores i've got a shitload of them over 10 years of work
A final bumb to close 2018.

I've been making these HT tests continuously, and the further they go, the better HT performance is noticed all around. Next year i will compare HT performance in games with Ryzen 7 vs. Intel Core i7. I will keep my GTX 1080 Ti for a lil longer too.
 
Joined
Aug 3, 2011
Messages
912 (0.30/day)
Location
At the Pub
System Name Sli Lake
Processor i5 7600K w.c
Motherboard Z170 Pro Gaming
Cooling Hardline Acrylic w.c
Memory 2x8GB Corsair 3000
Video Card(s) GTX 1070 Sli w.c
Storage M.2 850 Evo 512
Display(s) 4K 40" BDM4065UC
Case Corsair 750D
Audio Device(s) Asus Zonar
Power Supply Corsair AX1200
Mouse Logitech G502
Keyboard Logitech G910
Software Win 10 x64 atm
Benchmark Scores It Chops
Are all these game benchmarks "Single Player"?

It would be interesting what HT does to Multiplayer.
 
Joined
Feb 22, 2009
Messages
698 (0.18/day)
System Name Artas1984
Processor Ryzen 5 1600
Video Card(s) Sapphire Nitro R9 Fury OC
Storage Plextor M9PeG 512 GB
Display(s) Dell S2415H
Case Fractal Design Define XL Titanium
Audio Device(s) Asus Xonar Essence ST
Power Supply Evga 750 GQ
Mouse Razer DeathAdder 2013
Keyboard Roccat Ryos MK Glow
Software Windows 10 Pro 1809
Benchmark Scores i've got a shitload of them over 10 years of work
Joined
Feb 22, 2009
Messages
698 (0.18/day)
System Name Artas1984
Processor Ryzen 5 1600
Video Card(s) Sapphire Nitro R9 Fury OC
Storage Plextor M9PeG 512 GB
Display(s) Dell S2415H
Case Fractal Design Define XL Titanium
Audio Device(s) Asus Xonar Essence ST
Power Supply Evga 750 GQ
Mouse Razer DeathAdder 2013
Keyboard Roccat Ryos MK Glow
Software Windows 10 Pro 1809
Benchmark Scores i've got a shitload of them over 10 years of work
Seems lately there has been a lot of fuzz about the new Intel vulnerability issues with hyper threading. Since this benchmark is a fairly recent one, i am bringing this up just to remind that there is no single answer if HT should be kept for games or not, because every game behaves manually - games like Doom and Mass Effect Andromeda hate hyperthreading, while Total War 2 Warhammer and Prey love it.

Also, for multithreaded programs - the more cores your processor has, the less gain from mirror virtual cores will it gain. Here is an example:

HANDBRAKE 4K conversion test.

Xeon E5-2683 V4 2.3 GHz 16 C/32 T: 20:45 min.
Xeon E5-2683 V4 2.3 GHz 16 C/16 T: 22:45 min.

Gain from HT = 9 %

Xeon E3-1245 V5 3.6 GHz 4 C/8 T: 38:17 min.
Xeon E3-1245 V5 3.6 GHz 4 C/4 T: over 70 min.

Gain from HT = over 90 %.
 
Last edited:
Joined
Dec 31, 2009
Messages
15,407 (4.31/day)
Also, for multithreaded programs - the more cores your processor has, the less gain from mirror virtual cores will it gain. Here is an example:
Id bet that is application specific. What about cinebench? POV-Ray? Etc?

EDIT: It may be that application doesn't scale as well or the bottleneck is elsewhere (ram bandwidth, for an example) for that test. I would be interested to see more thorough testing of this theory across various applications to formulate a conclusion. I think you may be right in some cases, but it will vary by app.

I ran Cinebench R20 for example on my 7960x at 16c/16t and 16c/32t and saw more typical ~30% scaling.

Seeing 90% scaling for HT isn't a typical result in applications, note. That will vary wildly with application and load with how HT works.


EDIT2: I found something comparable, 7Zip, and ran a quick test..... 6900K vs 7960x (8c/16t vs 16c/32t).

In my quick testing, the 7960x hit 103070 MIPS at 16c/16t. Enabling HT yielded a score of 126336. an increase of 22%. In this test, a 6900K (8c) scored 30830 without and 38015 with HT enabled, a difference of ~23%. Seems similar to me and we doubled core and thread count. And that is just one example. ;)

Take a look at the Handbrake page as well... you'll see little scaling regardless in that testing compared to others. Look at the 7600K to 7700K, that is what, 10%? How did you get 90% scaling? May want to double check that result? And their words about Handbrake -
Since these are two lightly threaded workloads there isn’t much (or any) benefit when SMT is enabled. Handbrake is slightly better in this respect but again, other system components get in the way of delivering a CPU-focused result.
So, as I said, its the application (and potentially that odd result) which caused this curious blanket statement/conclusion.
 

Attachments

Last edited:
Joined
Feb 22, 2009
Messages
698 (0.18/day)
System Name Artas1984
Processor Ryzen 5 1600
Video Card(s) Sapphire Nitro R9 Fury OC
Storage Plextor M9PeG 512 GB
Display(s) Dell S2415H
Case Fractal Design Define XL Titanium
Audio Device(s) Asus Xonar Essence ST
Power Supply Evga 750 GQ
Mouse Razer DeathAdder 2013
Keyboard Roccat Ryos MK Glow
Software Windows 10 Pro 1809
Benchmark Scores i've got a shitload of them over 10 years of work
Very nice insights Earthdog. Great job providing some additional tests with those killer processors. The biggest surprise was the new Cinebench R20 - i was wondering if there was ever to be a new version, because R15 was getting very old, especially in GPU department where GTX 770 managed to outscore GTX 1070 and Vega 56 - strong case of software lethargy crying for help..

Now when it comes to handbrake, the 80 % difference (corrected after recalculating time) between 4 cores/4 threads and 4 cores/8 threads was due to the fact, that i ordered the CPU to process a 1154 Mbps 4K raw video file to 64 Mbps H.265 4K MKV output. When i add some common job to make some 1080p MP4, the HT only adds up like 30 % at best.
 
Last edited:
Top