• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

I did some benchmarking today...

Soldato
Joined
12 May 2011
Posts
6,272
Location
Southampton
I've been impressed at how well my little 1.6GHz Windows tablet handles some modern games (namely GRID Autosport) and I thought I'd see how low I could go with another well optimised game- GTA V. I wanted to see what effect massively reducing the clock speed and thread count had on frame rate and also compare it to Just Cause 3 (a 'poorly optimised' game).

My PC:
i7 4770k @ 4.2GHz
8GB 2133MHz DDR3 RAM
390x
Samsung Evo 850 SSD

I therefore set the games to run at medium-y settings, at 720p windowed, and manually started limiting the clock speed using windows power settings. I'd then set the affinity to work on 8, 4, 3 and 2 cores. Like this 4c, 1.6Ghz example:

4c%201.6GHz.png

I did a 30 second benchmark running forwards in a busy part of Los Santos. Anyway basically the results came out like this. It took a little while to realised that I needed to allocate alternate threads (e.g. thread 0, 2, 4 for the 3 thread run) unless I was using all threads. The random jumps in the frame rate is when I got run over and the camera points to the ground (high frame rate) or sky (low frame rate)!

GTA%20V%202%20Threads.jpg


GTA%20V%203%20Threads.jpg


GTA%20V%204%20Threads.jpg


GTA%20V%208%20Threads.jpg


Average frame rate for each run.
GTA%20V%20Averages.jpg


This really surprised me - I could play it fluidly at 1.6GHz 3 cores, 1.2GHz on 4 cores, and I even scraped a sort of playable 23fps at 800MHz using 8 cores. The frame rate was consistent too, not all over the place (except for aforementioned hit and runs!)

I did then cranked the settings back up to my normal Very High/Ultra level @1080p, using 3 cores @1.6GHz and the frame rate was the same. I drove around fast and did stuff and it was just as playable as the 'benchmark' run.

ANd then I tried Just Cause 3 (only 8 threads tested):
JC3%20Averages.jpg


The dip which occurs at around the same time is when I turn a corner onto a street scene. Again, Considering how this game "seems" to run terribly even at 4.2GHz, it runs very nicely at 1.6GHz. Again, I whacked the settings back up @ 1080p and played it for 15 mins or so at 1.6GHz and it stayed pretty playable.

I'm not sure what I was hoping to prove with this, maybe just point what we already knew, that games seem to be going towards many cores and fewer GHZ (e.g consoles) ?
 
Last edited:
Good stuff.

I'm not sure what I was hoping to prove with this, maybe just point what we already knew, that games seem to be going towards many cores and fewer GHZ (e.g consoles) ?

But that's not what the results show. That second to last chart is the best one. 2 to 3 threads gives a big jump, 3 to 4/8 is much of a muchness at most clock speeds. All show a strong CPU speed scaling though - a factor of about 4 between 0.8 GHz and 4.2 GHz.

In terms of "feels" though you said

I could play it fluidly at 1.6GHz 3 cores, 1.2GHz on 4 cores

Those look to be averaging high and low 30s repectively. If we draw a line at 30 FPS (say) the playable specs look to me like

2 threads: 2.5 GHz+
3 threads: 1.6 GHz+
4 threads: 1.2 GHz+

That would suggest that even the cheapest Skylake CPU (for example) on ocuk (the £54, 3.3 GHz dual core G4400) would do a perfectly playable job in GTA V. I'd expect the minimum frame times and variance to be bad, but if you say it feels fine then maybe it's not a problem.

Also those who like 60-144 Hz will probably disagree with you!
 
Agree with joeyjojo.

What it basically shows is Intel's approach of having a good balance between IPC and core count is the key, being at either end of the extreme (ie. having too few big cores or too many slow cores) is counterproductive.
 
I can see exactly where Almoststew is coming from given that his graphs show 4 threads at 2.5GHz are faster than 2 threads at 4.2GHz.

You also have to remember that while there is little or no difference between an i5 and an i7 'in this' games are becoming more dependant on more threads.

i5's are still good CPU's but i7's are coming into their own, I certainly wouldn't recommend an i3 for modern games at all now, dual cores are 2008 middle ground and should be concined to that history, 6 core CPU's as high premium products in this age is sick joke.

Hopefully Zen will offer a more sensible product structure.
 
Last edited:
Tbh not that long ago it was cheaper to buy a hex core i7 5820k than it was to get a 4 core i7 6700k.

Yeah, it was £300. a mistake, or test no doubt.

6 cores shouldn't be anymore than 300, 4 cores shouldn't be more than £200.

The i5 should be the new i3 but they keep the thing going as if dual core CPU's are still a premium product.
 
The thing is the i3's, i5's and i7's have been with us in the same form as they are now for the best part of a decade.

You replace a 2500K or a 2700K with a 6600K or 6700K and in pretty much anything you do with it you will only see a token difference.

The only thing that has changed in a significant way about Intel's CPU's since then is the price.

I hope AMD's bottom run Zen is a 4 core 8 thread Haswell IPC for £119.99!

oh... and a 10 core 20 thread monster for £500.

#### Intel.
 
Last edited:
Back
Top Bottom