• 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.

BF4 Retail CPU scaling measured

i7 4770k at stock, gpu's at stock.

a4sQJfz.png
 
Played a couple of hours now and really smooth without a single in game crash ... had to kill the exe a couple of times on leaving the game though! Could not find busy servers an a couple of occasions as well. Only technical issue I have with in game is that the sound is weird and kind of choppy. FPS is well over 100 mostly with some scenes dipping into the 90s .. tbh overclocking is not required on my setup! Graphs below of average GPU usage per core and average CPU averaged over 12 threads ..... the third graph combining both sections looks at CPU only showing Max loaded thread, Min loaded thread and average thread load ..

nga0.jpg


tbse.jpg


mk6w.jpg
 
Last edited:
Played a couple of hours now and really smooth without a single in game crash ... had to kill the exe a couple of times on leaving the game though! Could not find busy servers an a couple of occasions as well. Only technical issue I have with in game is that the sound is weird and kind of choppy. FPS is well over 100 mostly with some scenes dipping into the 90s .. tbh overclocking is not required on my setup! Graphs below of average GPU usage per core and average CPU averaged over 12 threads ..... the third graph combining both sections looks at CPU only showing Max loaded thread, Min loaded thread and average thread load ..

nga0.jpg


tbse.jpg


mk6w.jpg

Very nice, I'm guessing there's little to no stuttering on that setup judging by the graphs.
 
Nearly linear scaling with clock speed, clear sign of being CPU bound.

Not sure I'd agree, not seeing anywhere near 100 % CPU utilisation on any core even at 4.6 GHz. The gains are small with higher CPU clocks on the 8320. Will test my 3930k system too at various clocks and I would bet overclocking this would also result in some minor gains (and I doubt the 3930k would cause any sort of bottleneck at 4.5 GHz+)


What program are you using to create those slides?

Yes Excel once the raw data is produced by FRAPS.

Look like Excel ones to me?

The GPU not maxing out could be drivers, can't say it's a bottleneck unless we see other results.

Yes will try to see if similar behaviour occurs on the 3930k. Was also thinking of trying the FX8320 system at 1440p to see if I can max GPU utilisation. Heard that Crossfire at 1440p is not the best though. You may be right, drivers may be responsible.

Andy you're right about Win8/8.1. I am sure I would see gains in BF4. Is the reduced performance in Win7 anything to do with core parking as I have disabled this.
 
You don't need 100% CPU usage to get a bottleneck.
It could be maxing the parts it can max (Ergo, the bottleneck is down to the shared resource, the FPU, which is maxed out)
 
Last edited:
You don't need 100% CPU usage to get a bottleneck.
It could be maxing the parts it can max (Ergo, the bottleneck is down to the shared resource, the FPU, which is maxed out)

Granted but in a game with an apparently better AMD optimised instruction set, I would have thought one would get closer to 100% utilisation prior to reaching an FPU max out. Remains to be seen either way. Will do more testing.
 
It has 1 FPU to 2 cores.
It's pretty logical that you can max out the FPU performance it can give without maxing out all cores.

But again, it could be down to drivers.
 
It has 1 FPU to 2 cores.
It's pretty logical that you can max out the FPU performance it can give without maxing out all cores.

But again, it could be down to drivers.

Not disputing this at all. What I don't know is at what point this happens yet and I am not so sure what I've seen so far suggests there is a bottleneck. At 4.6 GHz if this was the case, I'd expect to see different (lower) GPU utilisation behaviour to 4.8 and 5 GHz. That didn't happen.
 
Very nice, I'm guessing there's little to no stuttering on that setup judging by the graphs.

No stuttering .. Beta was terrible on ultra on these GPUs on my 920@ 4.2ghz. The Shift to 4930k, windows 8.1 and probably some game engine improvements have saved the day, lol!
 
Last edited:
No stuttering .. Beta was terrible on ultra on these GPUs on my 920@ 4.2ghz. The Shift to 4930k, windows 8.1 and probably some game engine improvements have saved the day, lol!

lol, I bet you're glad it is performing noticeably better than your old CPU, as it's pretty expensive upgrade. The CPU alone cost's more than a CPU and motherboard usually does. Would love one but think I'll treat myself to one of those Ezio 240Hz monitors next. ;)
 
lol, I bet you're glad it is performing noticeably better than your old CPU, as it's pretty expensive upgrade. The CPU alone cost's more than a CPU and motherboard usually does. Would love one but think I'll treat myself to one of those Ezio 240Hz monitors next. ;)

Ha ha .... Had that 920 for 4 years - cost me £500 odd with board and ram back then and never felt OTT ... have to say that this does kind of feel OTT lol, except in Crysis 3 though :p Carried over my TRUE cooler and RAM to keep the wallet hit down a bit as well ..

My daughter gets the 920 on stock cooler now ... but before that I'll benchmark her athlon 640/GTX460 rig on BF4 :)
 
Look at the numbers - CPU speed goes up 8%, so do the FPS, ergo, bottleneck.

You may be right. Going to be doing more testing to get a bigger sample size. I hear win8/8.1 also makes a large difference (larger than 8% touted). Teppic has an excellent point, bus speed increase could account for minor fps bumps too.
 
well i have to say im very happy with my new i3 ivybridge :)
i3 with 4 threads enabled and a 6850 on medium settings its letting me get average frames of 72, upgraded from a phenom x4 where i could barely play at low.
 
It seems the Sweclockers results were for 64 player maps and used an average of three 180 second runthroughs. There were at least 50 players during each runthrough.
 
Back
Top Bottom