AMD GPUs for BF4 with Ultra settings at 3k Resolution

[Wild Thing]

How about 1 x 295X2 ?

 

[RaulF]

Start with 2 290's and you can add another if is not enough.

I did run BF4 @ 1080 with 200% scale with everything else maxed out in a couple of 290's (not X) and game never went below 50. Very smooth, by the way, not like SLI D:

 

[3DVagabond]

Check out the 290 Vapor-X. Obsoletes most of the 290X's IMO. Source KitGuru

63° gaming temp (forget Furmark ) is fantastic for Hawaii on air.

VRM temps in the 60's! Likely has a lot to do with the O/C headroom.

The Sapphire R9 290 Vapor-X OC is the star of the show today. We feel the R9 290 makes more sense than the more expensive R9 290X. With the amount of overclocking headroom available on this Sapphire card, it is straightforward enough to outperform the fastest *[factory] overclocked GTX780 Ti solutions, nevermind a reference R9 290X.

2 of these, with a space between them preferably, should do the trick, IMO.




*My edit. It's obviously not faster than the fastest manually O/C'd 780 ti.

 

[Silverforce11]

That Gigabyte Windforce 780ti OC has a out of box boost around 1.2ghz, so it isn't shabby at all.

Its hard to justify the cost of R290X over R290, especially when you can get them with awsome coolers such as the Tri-X, Vapor-X and PCS+

 

[wand3r3r]

As much as I dislike the anemic reference card I'm not sure if there's an alternative with 3 cards. 2 reference and 1 custom should work. I'll be curious to see how the end result is. I'd want 3 custom cards but I really don't know if the could handle the heat from the cards below.

Personally I'd go open bench with 3 custom cards, I believe that would work fine but open bench is usually not an option.

Have you had reference cards before?

A final note, I would have considered 290 for price reasons and the off chance that the very slightly cut down card would be further from throttling.

 

[Skurge]

My top card barely stays under 90c. I doubt 3 customs card would not throttle.

 

[UaVaj]

That Gigabyte Windforce 780ti OC has a out of box boost around 1.2ghz, so it isn't shabby at all.

Its hard to justify the cost of R290X over R290, especially when you can get them with awsome coolers such as the Tri-X, Vapor-X and PCS+

reason for choosing the 290x over the 290 is for that extra TUM.

as for the reasoning. the answer would cause a major sidebar to this thread. best for a new thread.

 

[UaVaj]

My top card barely stays under 90c. I doubt 3 customs card would not throttle.

top card is at 90C with only 2. what is your spacing between card? 1 slot? or 2 slot? how is your case airflow?

initial planning for 3 non reference was to have
gpu1 in slot 1/2
gpu 2 in slot 5/6
gpu 3 in slot 9/10
allowing 2 slot between gpu.
with lots and lots and lots of positive air flow.

 

[3DVagabond]

top card is at 90C with only 2. what is your spacing between card? 1 slot? or 2 slot? how is your case airflow?

initial planning for 3 non reference was to have
gpu1 in slot 1/2
gpu 2 in slot 5/6
gpu 3 in slot 9/10
allowing 2 slot between gpu.
with lots and lots and lots of positive air flow.

The hard part is finding a mobo that allows 3x triple slot (1, 4, 7) spacing. The Asrock X79 Champion does, but good luck finding it.

 

[Attic]

Going 1, 5, 9 on the slots of the PC case right? I know you had mentioned some of this to me in a PM. Couldn't ya go 1, 5, on mobo and then us an extender to get slot 7 from MB repositioned to slot 9 of the PC case expansion slots?


I think with 2 slots space for air between each card you're going to be fine if your driving about ~150-200cfm over the 3 cards and do the other airflow right.

 

[UaVaj]

2 Asus 290x Reference + 1 Gigabyte 290x Reference + Startech DP2DVI is on route.

will find out real soon how many 290x is necessary to get the job done.

if 2 is all that is needed - as many of u have suggested. will under clocked all three for reduce heat (throttle prevention) and for reduce noise (skipping underwater).

2 Asus 290x Reference and Startech DP2DVI is here. There was some shipping issue with the GigaByte 290x so that order was canceled.

2 more Asus 290x Reference on order. Should be here be the end of this week for Quad 290x.

 

[n0x1ous]

2 Asus 290x Reference and Startech DP2DVI is here. There was some shipping issue with the GigaByte 290x so that order was canceled.

2 more Asus 290x Reference on order. Should be here be the end of this week for Quad 290x.

you read Karlitos thread on quad 290?

 

[UaVaj]

you read Karlitos thread on quad 290?

Karlitos culprit is 2560x1440 on the third monitor which output only via native display port. the DP to DVI-DL adapter is broken.

for now. I am running 1920x1080 for 1080p eyefinity on the third monitor so I do not share his problem. DP to DVI-SL is working fine.

in the near future. I am considering upgrading to 1440p eyefinity (30" Dell U3014 x3). Only ""IF"", again ""IF"" 290x x4 can deliver BF4 ultra setting at 70+fps and 55fps absolutely minimum. based on my research - this is NOT achievable with ALL the eye candy.

will not duplicate Karlitos issue. U3014 supports native display port.

 

[UaVaj]

1) RESULTS w/ 290x x2.

Resolution 6000x1080 (3k bezel compensated) - Ultra Setting - 100% Resolution Scale.

290x x2 simple NOT cutting it. on open map like silk road and close map like locker. getting average low 70fps with dipping into the low 50fps. on texture filled map like zavod, parcel storm, dragon pass, guilin peaks, and especially siege of shanghi (after the sky scraper falls) and caspian border. averaging mid 50fps and dipping to low 40fps.



2) STUTTER.

CrossfireX on the 290x is smooth as butter. If you think NVidia SLI (680 in this case) was smooth. CrossfireX is smoother. (this is confirmation for 2 gpu).



3) DX11 vs Mantle.

a) Mantle gets a 5fps gain.
b) Viewing distance or whatever you want to call it is definitely washed out somewhat. Feels exactly like that claritin clear commercial.
c) Slightly longer to load game. not a deal breaker.
d) Stutter was more pronounce (0.53ms). FLA calcaluator frame variance confirms dx11 is smoother (0.35ms).

Given the added fog to view distance and only 5fps gain and more stutter.
Staying with DX11. just add more GPU.

 

[n0x1ous]

Karlitos culprit is 2560x1440 on the third monitor which output only via native display port. the DP to DVI-DL adapter is broken.

for now. I am running 1920x1080 for 1080p eyefinity on the third monitor so I do not share his problem. DP to DVI-SL is working fine.

in the near future. I am considering upgrading to 1440p eyefinity (30" Dell U3014 x3). Only ""IF"", again ""IF"" 290x x4 can deliver BF4 ultra setting at 70+fps and 55fps absolutely minimum. based on my research - this is NOT achievable with ALL the eye candy.

will not duplicate Karlitos issue. U3014 supports native display port.

:thumbsup: Hope it all works out for you as a fellow hardware nut and BF4 lover.

 

[Gloomy]

1) RESULTS w/ 290x x2.

Resolution 6000x1080 (3k bezel compensated) - Ultra Setting - 100% Resolution Scale.

290x x2 simple NOT cutting it. on open map like silk road and close map like locker. getting average low 70fps with dipping into the low 50fps. on texture filled map like zavod, parcel storm, dragon pass, guilin peaks, and especially siege of shanghi (after the sky scraper falls) and caspian border. averaging mid 50fps and dipping to low 40fps.



2) STUTTER.

CrossfireX on the 290x is smooth as butter. If you think NVidia SLI (680 in this case) was smooth. CrossfireX is smoother. (this is confirmation for 2 gpu).



3) DX11 vs Mantle.

a) Mantle gets a 5fps gain.
b) Viewing distance or whatever you want to call it is definitely washed out somewhat. Feels exactly like that claritin clear commercial.
c) Slightly longer to load game. not a deal breaker.
d) Stutter was more pronounce (0.53ms). FLA calcaluator frame variance confirms dx11 is smoother (0.35ms).

Given the added fog to view distance and only 5fps gain and more stutter.
Staying with DX11. just add more GPU.

Change Mantle's frame pacing.

In your user.cfg put:

RenderDevice.FramePacingMethod 0

if you don't want frame pacing

RenderDevice.FramePacingMethod 1

if you want type 1

RenderDevice.FramePacingMethod 2

if you want type 2 (this is default)

Also, there should be NO difference between Mantle and DX11. Something is broken on your config if this is the case.

 

[UaVaj]

Gloomy

thanks for the frame pacing method info. will give it a try.

as for mantle or dx11. that washed out distance is deal breaker for me.
whatever it is the cause. dx11 is simply more clear.
for now sticking with dx11

also with mantle. there is the occasional where the gpu goes haywire and spike/dips all over the place requiring BF4 restart.

 

[Gloomy]

There is no difference. We've already gone over this. That was patched a long time ago. I've even posted screenshots.

So if you're seeing a difference, then the problem is on your end, not Mantle's.

 

[n0x1ous]

There was a difference right after Mantle was released, but I no longer can tell the difference either. There is a VRAM memory leak in Mantle though which will require reboot after VRAM fills up completely. For me though, the Mantle codepath is so much smoother that I am playing 3 rounds and then closing restarting rather than switch back to DX11

 

[wand3r3r]

1) RESULTS w/ 290x x2.

Resolution 6000x1080 (3k bezel compensated) - Ultra Setting - 100% Resolution Scale.

290x x2 simple NOT cutting it. on open map like silk road and close map like locker. getting average low 70fps with dipping into the low 50fps. on texture filled map like zavod, parcel storm, dragon pass, guilin peaks, and especially siege of shanghi (after the sky scraper falls) and caspian border. averaging mid 50fps and dipping to low 40fps.



2) STUTTER.

CrossfireX on the 290x is smooth as butter. If you think NVidia SLI (680 in this case) was smooth. CrossfireX is smoother. (this is confirmation for 2 gpu).



3) DX11 vs Mantle.

a) Mantle gets a 5fps gain.
b) Viewing distance or whatever you want to call it is definitely washed out somewhat. Feels exactly like that claritin clear commercial.
c) Slightly longer to load game. not a deal breaker.
d) Stutter was more pronounce (0.53ms). FLA calcaluator frame variance confirms dx11 is smoother (0.35ms).

Given the added fog to view distance and only 5fps gain and more stutter.
Staying with DX11. just add more GPU.

It is awesome, I agree. I really don't know about the exact gains but I expect more. I haven't benched mantle/dx11 since my FPS are absurdly high at only 1200p and 290x crossfire. I'm very impressed with the bf4 playability in crossfire though.

:thumbsup: Hope it all works out for you as a fellow hardware nut and BF4 lover.

:awe:
Same here on both points.

 

[Gloomy]

Yeah, Mantle is much better for me also. I'm on a single card though :awe:

I RMA'd my 290 today, and I'm using an old 6950. I can't hold 60 with DX, even on low settings and my 4.5GHz 4770K... it's extremely annoying to see yellow numbers now. :/

 

[UaVaj]

Gloomy

one thing for sure. mantle is inferior to dx11 when it comes to visual quality displayed on the screen. real question is - can you tell the difference - and does it bother you?

i do accept your conclusion. i could be much much more sensitive to visual quality. after all - i will only play with all the eye candy on or will not play at all.

so for that measly 5fps gain and the lost of visual. no mantle for me.



n0x1ous

that explains why sometimes mantle goes haywire.

 

[Gloomy]

There is no difference dude. We've been over this already, sorry you missed the memo.

http://forums.anandtech.com/showpost.php?p=36171722&postcount=82
http://forums.anandtech.com/showpost.php?p=36171879&postcount=86

If there were a difference, I would notice it.

 

[UaVaj]

we can agree to disagree.

 

[Leadbox]

we can agree to disagree.

Or you could post evidence to back up your point of view?
Those images Gloomy posted have me leaning towards mantle for superior iq