My CPU benchmarks: GTA5 multicore testing

[scuzzycard]

So besides being active on forums, I have my own computer business, and I've been recommending that everyone and their mother get an i5 for games, since the i7 offers no advantage. Well, you know what - that's not completely true. GTA5 is a highly threaded game, and there ARE less frame drops when using an i7 vs i5.

Benchmark Setup:
i7 5820K - 4.6GHz Core / 4.4GHz Cache
Asus X-99 Motherboard
32GB DDR4 at 2400MHz CL11 / 500GB Samsung 950 Pro SSD m.2
Zotac GTX 980Ti 2X SLI overclocked to 1455/3748
3840X2160 resolution, FXAA, All settings on Very High

FRAPS 3.5.99 ------ MIN AVG
6 Cores, 12 Threads 96 163.8
6 Cores, 6 Threads 95 163.5
4 Cores, 8 Threads 96 162.9
4 Cores, 4 Threads 63 161.9
2 Cores, 4 Threads 38 157.7
3 Cores, 3 Threads 18 158.4
2 Cores, 2 Threads Failed to Load

I ran the 4-core/4-thread test 3 times just to make sure it wasn't a fluke. I'm going to test Crysis 3 the same way next, but I'm already going to start recommending the i7 to gamers who want every last of drop of performance.

 

[Gam3r01]

Youre talking about over a 100 dollar difference for 3 FPS gain in a very CPU heavy game. (Comparing 6600k (4C4T)to 5820k(6C12T), do note the 6600k does have better single threaded performance than the 5820k, so the difference is lost).

I do not consider that worthwhile.

I havent met anyone who says an i7 has zero gain over an i5 in games. However, these results reinforce the notion that there is zero reason to buy an i7 over an i5 for gaming.

 

[scuzzycard]

But look at the minimum frame rate - there's a big difference!

 

[Gam3r01]

In this case however the minimum framerates are still above acceptable levels, while I agree there is a difference, its not really a worthwhile difference.

 

[scuzzycard]

Well I agree with you there - for 99% of people, the 4C4T setup is enough, especially in this case. I think the difference would be more noticeable at slower clock speeds, and with older CPUs.

 

[scuzzycard]

I just want to add that I have no axe to grind either way - this is just informational.

 

[TerryLaze]

I just want to add that I have no axe to grind either way - this is just informational.

Well in the spirit of being informative...
[video="https://www.youtube.com/watch?v=J8DQl_VRiPw"][/video]

You get drops(and the loading takes forever) because the devs just can't be bothered to correctly setup a game.
Try this on all combinations of cores/threads and tell us if you still think that an i5 wouldn't be enough.

 

[scuzzycard]

The people I build systems for are not smart enough to use ProcessHacker - they just want something that will demolish their neighbor's PS4 or Xbox 1 and be just as easy to use. I'll still be selling lots of i5's, but I was surprised that the 4C8T setting was so much noticably faster than the 4C4T configuration (I also played the game with both, and I did notice the stuttering with HT disabled during game play, especially in very grassy areas and when things got particle-heavy. I will, however, be more likely to push i7's on clients with very large budgets, even if all they do is play games. That being said, I WILL bookmark your video for when/if I try to install GTA 5 again on my Pentium G3258. Thank you.

 

[scuzzycard]

OK, so I followed your video and re-ran the benchmark with 4 cores / 4 threads and, hey, you were right - it did a lot better after tweaking with ProcessHacker. MIN 92FPS, AVG 162.7FPS. Considering a 4C/4T Core2Quad is the minimum requirement, you would think they would've optimized it better for that configuration

 

[TerryLaze]

Yeah console cores are so weak that devs can't afford to put any synchronization into the code, they relay solely on the weak cores to cut off the speed of the threads. (doesn't work that well on the consoles either)

For windows a thread with a high priority means "finish this threads job first" but of course the main thread can't finish without the input of the helper threads.