***THE BF4 BENCHMARKS THREAD***

[LtMatt]

So you would agree that DX11 is better for anyone considering 4K.

Not necessarily. I'd need to try it on a proper 4k screen, as benchmarks have shown 290X runs 4k fine on Mantle on a proper 4k screen. The problem is resolution scaling uses a lot of vram, more than the 290X has so there you have the performance issue as its using page file to store the textures and why you have the stuttering. Mantle uses more vram than DX, no doubt about it, but how much extra is being used during the resolution scaling is the question.

 

[Kaapstad]

I'd need to try it on a proper 4k screen, as benchmarks have shown 290X runs 4k fine on Mantle on a proper 4k screen. The problem is resolution scaling uses a lot of vram, more than the 290X has so there you have the performance issue as its using page file to store the textures and why you have the stuttering.

It did not use a lot of memory for DX11 that 3.2gb figure I quoted for 4K was the maximum and that was rounded up a bit.

I also don't think there are many benchmarks for 4 x 290Xs using max settings and Mantle on BF4, reviewers normally have some of the settings turned down.

What I saw earlier was not even close, DX11 beat Mantle on mins, maxs, avg and VRAM usage by quite a margin.

 

[LtMatt]

It did not use a lot of memory for DX11 that 3.2gb figure I quoted for 4K was the maximum and that was rounded up a bit.

I also don't think there are many benchmarks for 4 x 290Xs using max settings and Mantle on BF4, reviewers normally have some of the settings turned down.

What I saw earlier was not even close, DX11 beat Mantle on mins, maxs, avg and VRAM usage by quite a margin.

HardOCP benched 4K on Mantle at x4AA and said there were no problems with running out of memory. They did however have some scaling issues i believe with quad crossfire under Mantle.

Mantle does uses more vram than DX though and resolution scaling definitely seems to exacerbate the situation. If its fine on a proper 4K screen though then that appears to back up what im saying.

Clearly there is still some work to be done on Quad Mantle at 4k. Its a very immature API though so i think you have to be patient. Its 5 months old vs DX11 which is like 6 years old. I'm just glad they have single card and dual gpu 290 crossfire working well, because on these much more common setups it ridicules DX in terms of overall performance.

Clearly DX is going to win in the situations you mentioned as Mantle has exceeded its vram limit. In all other situations though, DX gets its ass handed to it.

 

[Kaapstad]

HardOCP benched 4K on Mantle at x4AA and said there were no problems with running out of memory. They did however have some scaling issues i believe with quad crossfire under Mantle.

Mantle does uses more vram than DX though and resolution scaling definitely seems to exacerbate the situation. If its fine on a proper 4K screen though then that appears to back up what im saying.

Clearly there is still some work to be done on Quad Mantle at 4k. Its a very immature API though so i think you have to be patient. Its 5 months old vs DX11 which is like 6 years old. I'm just glad they have single card and dual gpu 290 crossfire working well, because on these much more common setups it ridicules DX in terms of overall performance.

Clearly DX is going to win in the situations you mentioned as Mantle has exceeded its vram limit. In all other situations though, DX gets its ass handed to it.

I would not take that HardOCP review seriously, the numbers they got for quadfire were considerably worse than what I get even when using Mantle.

 

[LtMatt]

I would not take that HardOCP review seriously, the numbers they got for quadfire were considerably worse than what I get even when using Mantle.

You can discard the fps figures, but you can't discard the vram usage figures. They will be accurate.

Resolution scaling aside, 290X's will have enough vram to run 4k maxed out on BF4. As long as you have enough of them to provide the grunt. If i can run 4k with resolution scaling and have just enough vram, they'll do it easier on a proper 4k screen without having to render the image twice, even with Mantles excessive vram usage.

No doubt they would've got better results using a X79 setup and a 6 core cpu though. So it definitely represents a worst case scenario for their fps figures.

 

[foster grant]

OK maybe someone can help me out or explain why Yesterday afternoon after installing my new 290 Tri-X I was getting around 70-75 fps in BF4 my monitor resolution is 1440x900 ..yet when I was on BF4 last night I was lucky to get 50 fps and the game is sluggish at those fps

No idea why the difference

 

[LtMatt]

OK maybe someone can help me out or explain why Yesterday afternoon after installing my new 290 Tri-X I was getting around 70-75 fps in BF4 my monitor resolution is 1440x900 ..yet when I was on BF4 last night I was lucky to get 50 fps and the game is sluggish at those fps

No idea why the difference

Welcome to the online experience that is BF4. If the servers are overloaded, your gameplay experience may vary.

 

[foster grant]

Well quite

Think I'll send the card back and consider other options

 

[MyBrainz]

wow i just tried BF4 with DX and man wot a laggy mess. Even with all High settings and 4xMSAA i was dipping into the 30's. Same settings under Mantle is nowhere near. Lows of mid 50's at worst with Mantle.

I guess i gotta remind myself how good Mantle is now n again

 

[foster grant]

Mantle ?

 

[MyBrainz]

Well quite

Think I'll send the card back and consider other options

Put a user.cfg file in ur BF4 folder and put this in there..

RenderDevice.FramePacingMethod 2

or

RenderDevice.FramePacingMethod 0

Matt can help u out with the user file if u need it.

Mantle ?

Elaborate please?

 

[Kaapstad]

You can discard the fps figures, but you can't discard the vram usage figures. They will be accurate.

Resolution scaling aside, 290X's will have enough vram to run 4k maxed out on BF4. As long as you have enough of them to provide the grunt. If i can run 4k with resolution scaling and have just enough vram, they'll do it easier on a proper 4k screen without having to render the image twice, even with Mantles excessive vram usage.

No doubt they would've got better results using a X79 setup and a 6 core cpu though. So it definitely represents a worst case scenario for their fps figures.

Just fired the Titans up for good measure

DX11 @4K VRAM 3.1gb

DX11 @5120x3200 VRAM 4.9gb

 

[LtMatt]

Put a user.cfg file in ur BF4 folder and put this in there..

RenderDevice.FramePacingMethod 2

or

RenderDevice.FramePacingMethod 0

Matt can help u out with the user file if u need it.



Elaborate please?

According to Johan it does not work for single card. Nonetheless id use 1 or 0. 2 definitely lowers fps the most in crossfire. 1 has a small hit, 2-4 fps and 0 is full speed. When using Mantle i can't tell a difference between all three settings so i usually use 1 or 0.

 

[Kaapstad]

@LtMatt

I have just had a thought

Would a 290X need more VRAM than a 290P due the extra shaders ?

 

[LtMatt]

@LtMatt

I have just had a thought

Would a 290X need more VRAM than a 290P due the extra shaders ?

I doubt it, as shaders are not responsible for storing vram.

 

[Kaapstad]

I doubt it, as shaders are not responsible for storing vram.

No but more shaders equals higher output per cycle.

 

[thefogo]

Well quite

Think I'll send the card back and consider other options

You're sending the card back...because the game is buggy?

 

[MyBrainz]

According to Johan it does not work for single card. Nonetheless id use 1 or 0. 2 definitely lowers fps the most in crossfire. 1 has a small hit, 2-4 fps and 0 is full speed. When using Mantle i can't tell a difference between all three settings so i usually use 1 or 0.

1 stutters for me Matt.

 

[LtMatt]

No but more shaders equals higher output per cycle.

That's nothing to do with storing vram textures in memory though. Higher fps does not equal more vram usage.

1 stutters for me Matt.

Furry muff. Guess it does work for single cards then. I wasn't about to argue with him though, I was way out of my depth.

 

[Kaapstad]

That's nothing to do with storing vram textures in memory though. Higher fps does not equal more vram usage.

Nothing to do with higher fps

What I am saying is there are more shaders outputting at the same time on a 290X, that extra output has got to go somewhere.

It is not like overclocking a GPU where the core goes faster but the amount of info coming out per cycle remains the same so there is no need for extra VRAM providing it is fast enough.