• 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.

***THE BF4 BENCHMARKS THREAD***

Maybe how the frames are renders from one resolution to another then compressed for display? Must be how dice have implemented it then as native driver down sampling is pretty much spot on :)
 
Results are in. I only tested Mantle once as it uses more vram than DX, but i will add the 4k mantle test later.

24/7 clocks 1000/1250 (stock 975/1250)
So running a very small overclock 2.5%



1440P Ultra Preset +AAx4

200% res scaling - DX11.1
40FPS - unplayable

VFOpz5g.jpg



200% res scaling - Mantle


45FPS - unplayable - better fps than dx, but more spiking

JLWSWZa.jpg



150% res scaling 4K res - DX11.1

89FPS - Playable for me - Still exceed the vram limit though. Every 10-15 seconds there is a slight hitch which shows itself as a small cpu spike. However this did not impact on the gameplay much and i would consider this playable. Nonetheless if it was me i would knock AA down to x2 to reduce vram usage and increase fps further. Whoever said when you exceed the vram limit you move down to single digit fps lied.

rWWCw2S.jpg

screenshot - 4K res + Ultra Preset + 4x AA

4k.png



175% scaling

I didn't grab gpu-z info for this as i know it will be over the limit. Nonetheless this was also fairly playable. Though the cpu spike was slightly bigger this time.

174.png


I'm actually pretty impressed with the performance of 290 Pro crossfire at 4k. I may decide to run 150% SS 24/7 and back off on the AA settings insread. If i could get it running in Mantle with similar vram usage i would get a nice fps boost as well as smoother performance and better render times. Anyway, thanks PG Tips for getting me to try it. :)
 
Last edited:
Just realised i ran the 4k test in Mantle, according to the screenshot. That explains why i exceeded the vram limit. Let me run again with DX11.1 at 4k.
 
150% and lower msaa would be the way to go for any of us, at 1920x1200 i was pleasantly surprised at how it ran with x4 msaa. Dropping aa would help even more. It still felt quite smooth though even on the max settings, albeit it was on the test range. No point me even trying 200% on a single 780.
 
My mistake. That Mantle screenshot i showed with 89 FPS was 150% res scaling, Ultra preset +2xAA, hence the high fps. Running it again with x4AA netted me lower fps.


150% scaling - Mantle
Ultra Preset +x4AA

8qAzpXM.jpg

ScreenshotWin32-0001.png


150% scaling - DX11.1
Ultra Preset +x4AA

Cr7JShB.jpg


What i find quite surprising is on Mantle the game appears to be quite smooth and playable even when i exceed the vram limit.
 
Smooth^^^

200% 1440p:

ScreenshotWin32-0001.jpg

1440 150% (4k sample)

ScreenshotWin32-0002.jpg

Both with no AA.

Nice. I'll put up a no AA comparison soon.

Lets make a new rule from now though. Post the starting location. Don't touch the mouse or keyboard, just post the very start. Should be easier to compare. :)
 
NP, I just tried it with 32x CSAA too (forced via nvcp), didn't appear to change much of anything, from frame rates to IQ.

Also is there a quick way to render screenshot or do I have to manually type it in everytime?
 
I thought I'd go 110% as it were. Running Backup in the background also

Floodzone 4K / Res scale 110% (4224x2376) (3032MB VRAM) MSAA 2X

Seems to run ok.


5wlo.png
 
Last edited:
Its ok I'll manage away with typing it as it gives a lossless png file :)

What point were you spawning at Matt? Couldn't seem to spawn (at least looking in the same direction as you) where you are. So spawned at Point A and here are some IQ comparisons at 200%...some IQ may have been lost moving it from an 8mb PNG to a
~1.5mb jpeg

All are 1080 200% (4K sample)

No AA:

4K_0xmsaa.jpg

2 x AA

4K_2xmsaa.jpg

4 x AA

4K_4xmsaa.jpg

And what happens when you run out of Vram...

VRAMLIMITHIT.jpg

Looks stunning that frosty, how do I get the sexy green text at the top? Always thought it was a Mantle thing...
 
Last edited:
Its ok I'll manage away with typing it as it gives a lossless png file :)


Looks stunning that frosty, how do I get the sexy green text at the top? Always thought it was a Mantle thing...

I resized it for you, was like 15mb so people were bound to kick off :D

as above render.drawscreeninfo 1 :)

full size here : http://imageshack.com/a/img845/4614/5wlo.png

100 fps with stock clocks with a physical horizontal pixel count over 4K isn't bad :D

Siege of Shanghai 4K (100% scaling) Ultra 4X MSAA

Runs absoloutely fine but the frame times are no where near as flat as Mantle. Note I am not using the BF4 resolution scaling, so I'm not sure how comparable it is in truth. But looks as though I'm roughly getting the same performance in most instances as Mantle over two cards (maybe ever so slightly better :p). Good times for AMD crowd

mbzy.jpg
 
Last edited:
24/7 clocks 1000/1250 (stock 975/1250)
So running a very small overclock 2.5%

Really small point, but remember that a 290 stock is 947/1250, the clocks you've given as stock are your stock factory-OC clocks. So compared to 'absolute' stock it's a 5.6% OC.

I know someone said it had been proven, but to get an idea how the scaling affects VRAM usage, etc. could someone at 1440p run at 720p and 200% scaling?
 
Back
Top Bottom