Gaming benchmarks: Core i7 6700K hyperthreading test

[Kanan]

"Well if you really understand my point, you'd understand that it doesn't at all create a false impression,"

the false impression it creates (to the unwise) is that the cpu matters more in gaming situations than it really does.. at least that is how i see it and always have done..

Just ignore what we say, I don't think you are well informed - go and read some reviews and get some enlightenment if you don't believe us. You are just holding to your wrong informations, I guess because of your ego, or because you don't want to upgrade your CPU anytime soon - well basically the same thing (ego).

when you have to run (never used in the real world) extra low resolutions to even show up what difference there is should make my point.. mind you the reviewers do have to fill their pages up with something.. even nonsense..

The only nonsense I can see here, is from you. You simply don't understand it and all the websites doing it are wrong, and you are right. Sure. Ignorance / arrogance strong on your part.

they run some games to supposedly create a real world gaming situation.. but then bugger it all up by running a resolution that in the real world world never be used.. some make this clear but most dont..

You still don't understand it in the slightest. Well, alright. Last time I tried explaining it to you.

tis a bit like running 3dmark but using iceworld rather than firestike whilst trying to suggest the test means something useful in the real world to a gamer..

I give it up. ^^

so called real world gaming tests should use real world gaming resolutions.. this way they would produce real world results.. and we all know what those would be.. in simple terms the latest and greatest cpu wont make the slightest difference to the gaming experience.. keep your money in your pocket.. he he

trog

You don't see the difference between theory and practice. They / we do. That's it. And you still don't get it. Np. Ignorance is bliss.

 

[OneMoar]

there is so much wrong in this thread I can't even quantify the stupid ...

 

[Artas1984]

I have updated the thread with video presentation.

Trog, i want to hear no more comments from you, ok? You have made this thread into some kind of personal show of dissatisfaction, i am getting tired of opening my thread and seeing "Scott Pilgrim vs the World".

 

[trog100]

who the f-ck is scot pilgrim.. he he

but once a thread is open it is no longer your own private property Artas its open to member comment or at least it should be.. but i will respect your wishes and make no more comment in it..

being called ignorant by the ignorant aint much fun anyway..

consider me gone from your thread even though it was a good one..

this place is a little too tolerant of what could be called personal attacks and insults..

trog

 

[CAPSLOCKSTUCK]

i have run my Xeon X5670 @4.4 ghz
in the following comibinations

2 core ht off
2 core ht on
4 core ht off
4 core ht on
6 core ht off
6 core ht on

MOTOGP 15 runs at 95-100 fps regardless.


2 Core HT on

2 Core HT OFF

If i have time i will do the same for GTA V.

 

[the54thvoid]

I'm very lazy but for a quick check I ran a one minute run on BF4, MP, Dragon Valley (only me on server).

3930k @ 4.2Ghz
All setting maxed and at 150% DSR (4k effective)

HT on
Frames, Time (ms), Min, Max, Avg
4656, 60000, 63, 89, 77.600

HT off
Frames, Time (ms), Min, Max, Avg
4694, 60000, 57, 88, 78.233

Run consisted of flying Havoc heli between two caps, firing into water and through gap in trees for one minute each time.

I run a 980ti at 1500Mhz so the CPU does play 2nd fiddle. I think the discussion really needs to centre around low end CPU's. Obviously 12 threads is overkill, even in BF4 but when dropping core count to 2, perhaps that i3 discussion is more relevant.

However, I don't think the OP can state HT on hurts frame rates without looking at the evidence more scientifically. An explosion or any other physics derived effect will impact on frame rate by using CPU resources. It's impossible to negate these variables without set, predefined benchmarks. What is relevant is that HT does not help performance at a certain core count. What would be great to know from game devs is the core optimisations for games.

In years past, 'multi-core' started appearing on DVD packs. It would be so much better to know if it was optimised for 1, 2, 4 or more cores. Safe to say anyone that buys the next top end Broadwell E for gaming alone (at 10/20 cores/threads) needs to seriously think about giving their money to charity.

 

[Artas1984]

No. Same HTT. It's just that most games only need 4 threads/4 cores and a 4 core i7 has 8 threads out of 4 cores, it means your power gets divided, if the game starts to use the HTT thread instead of the real thread from the core itself.

That is the best (if a bit straightforward) answer that i could think of, that explains the loss of performance with HT turned on.

Trog, no hard feelings, it's just that your conversation with Kanan about the same stuff is getting pretty old, we need to change topics...

However, I don't think the OP can state HT on hurts frame rates without looking at the evidence more scientifically. An explosion or any other physics derived effect will impact on frame rate by using CPU resources. It's impossible to negate these variables without set, predefined benchmarks. What is relevant is that HT does not help performance at a certain core count. What would be great to know from game devs is the core optimisations for games.

Yes, exactly, you were right to correct me here.

 

[pascal]

Awesome work Artas1984, thanks for the analysis across all those titles.


The art of writing multi-threaded software is still being perfected. As you likely know an HT core shares the same cache and execution unit of the physical core. Depending on what is being threaded the overall performance can degrade if the overhead cost of core sharing is higher than the output of the threads. This can be seen if you thread some basic integer arithmetic on real cores vs real cores + HT cores.

The developer can choose which cores they want their software to thread on (1 core, all cores, real cores only, or real core 1 and 2 and only HT core 4).

One could theorize that the games that don't reveal a performance delta when HT cores are toggled are threading only on real cores or are experiencing a bottleneck elsewhere.

[edit] remove article caused by backspacing + rewriting sentence fragment.

 

[trog100]

"Trog, no hard feelings, it's just that your conversation with Kanan about the same stuff is getting pretty old, we need to change topics.."

that old i have complexity forgotten about.. who is kanan.. he he

in truth who gives a f-ck its all time passing waffle.. you take part in it or you dont..

trog

 

[CAPSLOCKSTUCK]

in truth who gives a f-ck its all time passing waffle.

https://www.facebook.com/

 

[Dethroy]

This debate is quite pointless.
Why would software that isn't optimized for the usage of more than 4 cores benefit from HT (exception being i3 obviously)?

for as long as i can remember good PC gaming as always been about graphics card power..

You must be rather young then.

 

[ASOT]

For games 6600K is better choice,also cheaper..

I had 6700K and was not so happy with HT i switch to 6600K and OC

That 2 mb cache difference seems to less for 8 threads

 

[Mussels]

bumping a slightly older thread, as i've been doing testing myself with my new CPU with HT on/off in gaming (takes 10C off load temps, for no gaming loss) and i googled it leading me to here...


My question is: is FRAPS the problem here. It's basically recording the 3D and playing it back, and the CPU load may be entirely different to rendering it initially. Have you tested with in-game benchmarks where available to see differences there?

Not sure if i'm wording my concern correctly, anyone else able to expand on the idea?

 

[Schmuckley]

Well...Time to turn HT off

 

[CAPSLOCKSTUCK]

I paid the princely sum of 5 quid each for my threads........ so i am leaving mine on !!!

 

[Shou Miko]

HT does a little in gaming if the game can figure out what to do with the extra threads.

Many buys the i5 bcs of the value to performance compared to the i7 and if u disable HT on the i7 and let both do 4ghz for a fine base clock with boot to 4.2ghz than it should be a question of the 6 vs 8mb cache catfight.

 

[Mussels]

HT does a little in gaming if the game can figure out what to do with the extra threads.

Many buys the i5 bcs of the value to performance compared to the i7 and if u disable HT on the i7 and let both do 4ghz for a fine base clock with boot to 4.2ghz than it should be a question of the 6 vs 8mb cache catfight.

i have a 100Mhz difference between my 3770k and 2600k, so that catfight is a daily occurence here

 

[Shou Miko]

i have a 100Mhz difference between my 3770k and 2600k, so that catfight is a daily occurence here

but u r now also fighting with a 2. gen vs 3. gen cpu so there should be a difference even with that 100mhz difference

 

[Mussels]

but u r now also fighting with a 2. gen vs 3. gen cpu so there should be a difference even with that 100mhz difference

PCI-E 3.0 and clocking the ram higher is about the extent of it, to stay on topic HT on and off has made zero noticeable difference.

 

[RejZoR]

I've always had HT enabled and never looked back. Didn't buy a hexa core to save power. Besides, HT does come into action a lot for 7zip that I do use to compress rather big data regularly.

 

[Mussels]

I've always had HT enabled and never looked back. Didn't buy a hexa core to save power. Besides, HT does come into action a lot for 7zip that I do use to compress rather big data regularly.

could you do a test with and without?

HT off
HT on, but 7 zip set to 4 cores (this alloes the HT threads to help with background tasks in the OS etc, much like you'd get with gaming)
HT on, 7 zip all cores

 

[RejZoR]

Tested on Core i7 5820K (6 physical cores, 12 max threads with HT) @ 4.5 GHz with 32GB 2400MHz DDR4 RAM with HT enabled and disabled for different scenarios.

6 Threads / 6 cores (HT OFF)

6 Threads / 6 Cores (HT ON)

12 Threads / 6 Cores (HT OFF)

12 Threads / 6 Cores (HT ON)

MADE TWO MORE FOR TEH LOLZ BEYOND AVAILABLE USABLE THREADS:

16 Threads / 6 Cores (HT ON)

18 Threads / 6 Cores (HT ON) (This is as high as RAM would allow me)

 

[Mussels]

PURDY NUMBERS!

thanks for those, i'll wrap my head around the results in the morning

 

[Shou Miko]

18 Threads / 6 Cores (HT ON) (This is as high as RAM would allow me)
View attachment 75514

How can a Hexa-core have 18 threads?

what haxx do u use?

 

[RejZoR]

That's threads utilized by the software. You can cram several threads onto one physical core. It is beneficial if cores aren't utilized 100% by one software thread. You don't gain anything if one software thread already occupies entire core. I've used this just to see how CPU with HT performs under absurd conditions when all cores are fully utilized and software tries to feed it even more work.

If you open up Task Manager, Performance tab you can see number of threads used by Windows and apps on it. 1160 at the moment on my system. But since basically all of them are just idling, they aren't causing any performance issues.