Battlefield 1 CPU Performance (DX11 vs DX12)

[Sweepr]

Using the Fury X we see that DX12 allows for a few more frames per second on average with the Core i7-6700K. However, the minimum frame rate was oddly 19% lower. Moving to the Core i5-6600K we find that again DX12 boosted the average frame rate by a small margin, yet this time we see an even larger 29% decrease when looking at the minimum frame rate.

The situation worsens with the Core i3-6100T processor. DX12 looks amazing here if you focus solely on the average frame rate as it improved the Fury X's performance by 14%. Sadly, we saw a 53% reduction in minimum frame rate and the story is similar for AMD's FX processors.

It's the minimum frame rates that often take the biggest hit and even Pascal GPUs paired with an overclocked i7 are much slower when using DX12 versus DX11, and there isn't much to be gained by opting for DX12 on AMD GPUs either.

www.techspot.com/review/1267-battlefield-1-benchmarks

 

[NTMBK]

That minimum drop is weird. I wish they had more detailed frametime graphs- is that minimum drop due to a single big hit, from something like compiling shaders? Or is it due to persistent stuttering?

 

[itsmydamnation]

Really want to see multiplayer not single player and when DX12 is working better. its sad things aren't launching with better DX12 support because after the initial comparison people who aren't using it dont notice it because things are rareluy ever re-benchmarked.

In BF4 Mantle was significantly better then DX11 in terms of game play experience on my 290.

That minimum drop is weird. I wish they had more detailed frametime graphs- is that minimum drop due to a single big hit, from something like compiling shaders? Or is it due to persistent stuttering?

i just bought BF1 ( big mistake on my part i think....ekk) i'll get some frame timings on DX12 going once its finished Dl'ing. im not an excel guru will standard issue ~9 bin bell curve do?

 

[Mercennarius]

Interesting that average FPS was higher yet there was such a significant drop in minimum FPS. Looks like DX12 isn't as optimized currently here..

 

[Arachnotronic]

Interesting that average FPS was higher yet there was such a significant drop in minimum FPS. Looks like DX12 isn't as optimized currently here..

Minimum FPS is very important for smoothness. I don't know why anyone would subject themselves to the lower mins of the DX12 codepath.

 

[NTMBK]

i just bought BF1 ( big mistake on my part i think....ekk) i'll get some frame timings on DX12 going once its finished Dl'ing. im not an excel guru will standard issue ~9 bin bell curve do?

That would be really appreciated, thanks!

 

[LTC8K6]

Why a 6100T instead of a 6100?

 

[swilli89]

Really want to see multiplayer not single player and when DX12 is working better. its sad things aren't launching with better DX12 support because after the initial comparison people who aren't using it dont notice it because things are rareluy ever re-benchmarked.

In BF4 Mantle was significantly better then DX11 in terms of game play experience on my 290.

i just bought BF1 ( big mistake on my part i think....ekk) i'll get some frame timings on DX12 going once its finished Dl'ing. im not an excel guru will standard issue ~9 bin bell curve do?

That would work. DX12 has worked wonders in lots of games. Looks like the implementation here is lacking. However, as you said Mantle was superior to DX11 in BF4 so I think we will see this improve.

 

[LTC8K6]

And where did they get an FX-4320?

That is listed as an unreleased FX chip?

 

[AtenRa]

That minimum drop is weird. I wish they had more detailed frametime graphs- is that minimum drop due to a single big hit, from something like compiling shaders? Or is it due to persistent stuttering?

Single player still but fps timeline for the entire benchmark run.

http://www.hardocp.com/article/2016..._card_dx12_performance_preview/1#.WA-HECTvUgk

 

[coercitiv]

Minimum FPS is very important for smoothness. I don't know why anyone would subject themselves to the lower mins of the DX12 codepath.

Patiently waiting for subsequent patches to bump DX12 past DX11 even for minimums, then see how DX12 is still downplayed because some other reason. (developer bias will do before more refined approach emerges)

 

[LTC8K6]

What happened in the first second of the DX12 run that killed the 480? Without that, the minimum number is much better.

 

[pcslookout]

It is like everyone is scared to do multiplayer benchmark in a 64 slot conquest server.

 

[Erenhardt]

What happened in the first second of the DX12 run that killed the 480? Without that, the minimum number is much better.

In my testing DX12 seems to be choppy just after the load screen is over and the narrator is going about the map objectives. After that is is gone. Probably dx12 rendered and game engine fighting for resources as one is trying to load the rest of the assets and dx12 is trying to spit as many frames a possible, killing those CPU cores.

DX12 needs some work. But so do whole game. It feels rushed and have many minor bugs still.

OT: Did you guys see there were cheaters in the beta already? Servers are now filled with cheaters. Smoke grenades are worthless, as it only impairs your vision. Cheater will headshot you anyway

 

[AtenRa]

OT: Did you guys see there were cheaters in the beta already? Servers are now filled with cheaters. Smoke grenades are worthless, as it only impairs your vision. Cheater will headshot you anyway

https://forums.anandtech.com/thread...ng-october-18th.2462666/page-17#post-38454602

 

[Abwx]

And where did they get an FX-4320?

That is listed as an unreleased FX chip?

https://geizhals.de/amd-fx-4320-fd4320wmhkbox-a1387274.html?hloc=at&hloc=de

https://geizhals.de/?cat=cpuamdam3p

 

[LTC8K6]

https://geizhals.de/amd-fx-4320-fd4320wmhkbox-a1387274.html?hloc=at&hloc=de

https://geizhals.de/?cat=cpuamdam3p

Strange,and it still seems like an odd choice. A 4350 would match the rest of their choices.

http://www.cpu-world.com/CPUs/Bulldozer/AMD-FX-Series FX-4320.html

 

[Abwx]

Strange,and it still seems like an odd choice. A 4350 would match the rest of their choices.

http://www.cpu-world.com/CPUs/Bulldozer/AMD-FX-Series FX-4320.html

Scroll to the bottom of CPU Fantasy World page..

AMD apparently settled ultimately for the 4350, not only it has a full 8MB L3, wich is usefull in games, but it also consume the same as the 4320 despite its 5% higher frequency, it s a well binned chip that has 10% better perf/Watt than regular 43XXs.

AMD continued to gain momentum in virtual reality (VR) with the introduction of the most affordable VR-ready PC to-date, the limited edition CyberPowerPC, featuring the AMD FX 4350 processor and Radeon™ RX 470 graphics.

http://ir.amd.com/phoenix.zhtml?c=74093&p=irol-newsArticle&ID=2213570

In case someone was wondering about its 125W TDP rating it s only related to the cooling solution, otherwise Hardware.fr measured 60W for the CPU in a heavy integer loading that is more demanding than games.

 

[crashtech]

What happened in the first second of the DX12 run that killed the 480? Without that, the minimum number is much better.

Even without the first second, the frame rate is all over the place according to the graph. Puzzling.

 

[pcslookout]

Finally someone did a test of Multiplayer 64 slot Conquest online with a i5 vs a i7! The difference is huge!