6 core vs 4 core gaming.

[Master&Puppet]

The Idea

I was dreaming about getting myself a 3930 after watching TTLs review on it last night and it got me thinking - what exactly is the effect of having 2 extra cores?

We know that games are rarely 'coded' to make use of more than 4 cores and many aren't even coded for that. However we also know that Windows 7 has a sneaky habit of spreading the load across all available cores.

So with that in mind I played a few games of BF3 and recorded the fps.

Test rig

Crosshair formula 4

4 x 4gb Corsair XMS3 1600

2 x 5850 @ 788Mhz/1125mem

And one 1100t:

for the 6 core test all 6 cores were used and the whole thing clocked to 3978 (17*234) @ 1.475 vcore.

for the 4 core test the worst 2 cores were turned off manually and the 4 remaining were clocked to 4216 (17*248) @ 1.475 vcore.

I then sat down and played the same level of the bf3 sp campaign and recorded the min/av/max fps using fraps.

Results and Conclusion

You can see the results in sexy table and graph format here

Clearly the 6 core is making a significant difference even with me giving the 4 core setup all the help I could by overclocking it to 4.2. So in principle a more powerful cpu is more power to the game but in reality it depends upon whether the cpu is bottle necking the gpu(s) - which mine clearly is.

It would intrigue me to know whether hyper threading on the Intel chips would do a similar thing (ie 2500k vs 2600k at the same frequency).

Anyway, that was just a Saturday morning mini experiment.

M&P

 

[SnW]

Interesting mate

will read carefully later

Greetings Ray........

 

[SPS]

That's pretty interesting.

As for the 2500k vs 2600k I would probably say that it would produce similar results.

 

[Nathlm]

Was this at all highest settings?

 

[MaXiMiZe]

Interesting...

 

[marsey99]

nice 1 dude

i am sure i was reading that the fake cores of ht on the 26k didnt have a big impact on gaming as they aint real cores but i am sure one of the guys with a 26k and bf3 will run a few tests to show how they stack up

ps, get a 25k ;D

 

[d6bmg]

Pretty good comparison done. For your hard work, a +1 for your post

 

[Master&Puppet]

Was this at all highest settings?

Nearly but my 1gb gpus can't handle it - everything is on ultra except AA deferring (OFF) and Anisotrophic filtering (1x).

nice 1 dude

i am sure i was reading that the fake cores of ht on the 26k didnt have a big impact on gaming as they aint real cores but i am sure one of the guys with a 26k and bf3 will run a few tests to show how they stack up

ps, get a 25k ;D

Yea I've seen a couple of reviews so far and it's made me wonder about cpu/gpu balance - in order to test it the cpu must be bottlenecking the gpu(s) by quite a lot:

1. If the gpu is heavily limited by the cpu then all the gains can be attributed to the HT/extra cores.

2. If the gpu has slighty better relative performance over the cpu then people assume that HT/extra cores do very little.

3. If the cpu is equal to or has great relative performance then people assume HT/cores do nothing.

Therefore these fps results have a ceiling which is created at the point where the relative performance of the cpu equals or beats the gpu (i.e. cpu bottlenecking no longer exists).

Problem is - if you test this on a 2600k is there any chance that the gpu is going to be bottlenecked?

To be honest what I should have done is UnderClocked my cpu and checked the difference in performance then. You could expect even great performance scaling then and SPS might be right.

Might do that quickly!

M&P

 

[Master&Puppet]

Ok test has been run again using the same method as above however the 4 core and 6 core tests were run at the stock 3.3ghz.

Second Results

Here's the new results in sexy table and graph format

As expected the 4 core runs performed better with the OC but interestingly neither as well as the 6 core runs.

The overclocked 4 core run performed less well than the stock 6 core run. The pc is definitely making full use of all 6 cores and is not 4 thread biased.

Second Conclusion

I'm in no doubt that the pc uses all the cores and the performance difference is quite shocking. I would say that it would be even greater if my gpus were more powerful which proves what I was saying in the post above. So if anyone does try to see if HT has an effect make sure you underclock your cpu until you see the performance dip before making the comparison!

I'm going to install Shogun 2 overnight and do the same thing using the benchmark just to check that this isn't some BF3 anomally but so far it looks like the whole 'games only use up to 4 threads' is looking pretty rubbish. If you have cores which aren't doing much then the cpu has probably got more life in it than you might have expected. Similarly if HT has an affect the 2600k might offer better future-proofing than the 2500k...

And lastly, on a personal level - there was no change to performance between the 6 core results. I had assumed my cpu was bottlenecking the gpus but I was wrong, all that OC is actually being wasted!

M&P

 

[SPS]

How many times did you run the test? Because it's not a huge difference and I think you would need an average of at least three to make your results accurate

 

[KING_OF_SAND]

For all the Intel fans, I hate to burst your bubble but because games don't use hyperthreading a 6 core 1090t will out perform a 2700k pretty handily in a 6 core game.

 

[SPS]

For all the Intel fans, I hate to burst your bubble but because games don't use hyperthreading a 6 core 1090t will out perform a 2700k pretty handily in a 6 core game.

Which games are coded directly to use 6 cores?

 

[hmmblah]

For all the Intel fans, I hate to burst your bubble but because games don't use hyperthreading a 6 core 1090t will out perform a 2700k pretty handily in a 6 core game.

Games don't get programmed to use hyperthreading because they don't need to be. It's handled by the OS. A thread is a thread to a game whether it is hyperthreaded or a true core. So no, a 6 core 1090t will not out perform a 2700k.

 

[KING_OF_SAND]

Which games are coded directly to use 6 cores?

The new dirt game can run eight.

 

[SPS]

The new dirt game can run eight.

Good because I can't wait for it

 

[Master&Puppet]

Right, so here's some more results from a few more games.

Same method as before, although this time I ran benchmarks provided by the games and averaged several runs together.

The main purpose was to compare 4 core and 6 core results running at stock 3.3 ghz but I also included a 6 core at 4ghz to test for a gpu performance ceiling.

Shogun 2

Just Cause 2

I also ran benchmarks for Mafia II but the results are spurious. I was getting fps deviation of 20-30 for min and max frame rates even when running the benches back-to-back! Results are included for what it's worth. Only conclusion I can draw is that minimum frame rates seem to be lower than maximum frame rates and the average sits somewhere in between

I would say that there was a general trend towards all the frame rates preferring more cores but the utilisation of the 5th and 6th varies. BF3 makes significant use of them but these other games benefit only a little. My gut instinct is that these last games only make use of 4 cores (or perhaps even fewer than 4 cores) but the OS is able to pass off the background threads to the spare cores which allow the games to run slightly faster. BF3 meanwhile sucks up all the spare capacity of the 5th and 6th cores too. I suppose I could test this by tracking cpu usage during the benches...I'll have a think about that.

Games don't get programmed to use hyperthreading because they don't need to be. It's handled by the OS. A thread is a thread to a game whether it is hyperthreaded or a true core. So no, a 6 core 1090t will not out perform a 2700k.

This is my thinking at the moment but I haven't got the chip to test it with! I would say that things have started heading upwards in thread usage, BF3 using 6 threads now and if the new dirt is looking at 8 as KoS said then who knows where this will end up.

Of all the things you have to consider the Hyper-threading is the most vital thing. At just short of £250 the price gap between the i7-2600K and the i5-2500K is large, especially for the normally small gap in performance. So if your tasks are largely based upon needing to make an enormous amounts of calculations in the tiniest timescale, then the extra £80 is a worthy expenditure. Otherwise, unless you must have the best at any price, we'd advise the Core i5-2500K as the best of the new breed.

It will certainly make cpu choice more interesting if hyperthreading is utilised. We could really do with some testing on HT from someone because it could well make a significant difference to chip choice for IvyB - the equivalent 2500k being the 3570K (non-HyperT 4 thread) and the 2700k equivalent being the 3770k (HT 8 thread) is about £70 difference I think.

And as for the 2011 socket its usefulness remains to be seen. Firstly, 6-8 threads might well become the standard over the next year or so but I don't expect to see 12 during the life of IvyB. Secondly, even if threads do rise above 8 then I still can't see that meaning an IvyB 1155 is going to bottleneck the 7xxx or 6xx series gpus. It's certainly not worth getting the SB-e just to use cores 5 and 6 to run the idle/background processes!

To summarise: BF3 could be the game which forces us to take the leap into HT. The 2500k was the perfect chip before-hand but now maybe the 2600k has become the boss. What do people think?

M&P

 

[Master&Puppet]

My gut instinct is that these last games only make use of 4 cores (or perhaps even fewer than 4 cores) but the OS is able to pass off the background threads to the spare cores which allow the games to run slightly faster. BF3 meanwhile sucks up all the spare capacity of the 5th and 6th cores too. I suppose I could test this by tracking cpu usage during the benches...I'll have a think about that.

Just a quick test of this - I thought that by testing again using the 4 core @ 4.2ghz we'd be able to see whether there was any difference.

Shogun 2

Just Cause 2

I reckon the Shogun 2 scores are probably topping out because of the gpus bottlenecking the fps and the Just Cause scores are too close to call.

 

[Nculver]

The Idea

I was dreaming about getting myself a 3930 after watching TTLs review on it last night and it got me thinking - what exactly is the effect of having 2 extra cores?

We know that games are rarely 'coded' to make use of more than 4 cores and many aren't even coded for that. However we also know that Windows 7 has a sneaky habit of spreading the load across all available cores.

So with that in mind I played a few games of BF3 and recorded the fps.

Test rig

Crosshair formula 4

4 x 4gb Corsair XMS3 1600

2 x 5850 @ 788Mhz/1125mem

And one 1100t:

for the 6 core test all 6 cores were used and the whole thing clocked to 3978 (17*234) @ 1.475 vcore.

for the 4 core test the worst 2 cores were turned off manually and the 4 remaining were clocked to 4216 (17*248) @ 1.475 vcore.

I then sat down and played the same level of the bf3 sp campaign and recorded the min/av/max fps using fraps.

Results and Conclusion

You can see the results in sexy table and graph format here

Clearly the 6 core is making a significant difference even with me giving the 4 core setup all the help I could by overclocking it to 4.2. So in principle a more powerful cpu is more power to the game but in reality it depends upon whether the cpu is bottle necking the gpu(s) - which mine clearly is.

It would intrigue me to know whether hyper threading on the Intel chips would do a similar thing (ie 2500k vs 2600k at the same frequency).

Anyway, that was just a Saturday morning mini experiment.

M&P

Nice work man, thanks for putting all of that together!

 

[jonnyfgm]

Interesting posts, I have a feeling however that BF3 is a really well coded game, i saw near perfect 2x scaling when i went SLI on my 450s so perhaps might not be the best game to try this particular test on

I will probably give the HT on/off a go when i get my 7970 later this week

 

[yassarikhan786]

Interesting posts, I have a feeling however that BF3 is a really well coded game, i saw near perfect 2x scaling when i went SLI on my 450s so perhaps might not be the best game to try this particular test on

I will probably give the HT on/off a go when i get my 7970 later this week

For me, HT on/off didn't provide any noticeable difference. I neglected to check the fps with HT on and off, but visually I didn't notice any difference.