An AMD Mantle users Testing BF4.

[Carfax83]

This is so typical for MS. It's really bad since MT support is supposed to be a DX11 feature, not a Win8.1 feature. This is like charging for a patch.

The increased performance under Windows 8.1 have nothing to do with MT as far as I know. It comes from better use of buffers or something. To quote Johan:

“We use DX11.1, there are some optimizations in it (constant buffer offsets, dynamic buffers as SRVs) that we got in to the the API that improves CPU performance in our rendering when one runs with DX11.1. This will be in BF4.”

It's probably more than that though. Windows 8.1 also has a smarter thread scheduler, and a faster kernel with lower latency than Windows 7.

 

[PPB]

More worst case scenario testing. I have in my possession now a gtx780 and an r9 290. Siege of Shanghai is a great map for showing CPU limits in BF4 under direct X. Server is 64 players conquest.

i5 2500k @ 4.5ghz
Windows 7
Same settings for all cards.
1920x1080
All settings to lowest values
Filtering 16x
Textures to Ultra
Mesh Quality to Ultra
HBAO/SSAO Off

The gtx780 is running at 1176mhz and mem at 6502mhz(highes stable overclock on stock bios), but is downclocking because the CPU isn't feeding it at this location.

R9 290 Tri-X OC @ stock clocks (1000mhz core) in direct x 11 I didn't have afterburner installed when I took these so I'm not sure if clocks were maxed in this screenshot or not.

R9 290 Tri-X OC in Mantle

I'm not sure how much the skyscraper being down effects performance. All I know is that the gtx780 and r9 290 are hitting the same CPU bottleneck in this location (lowest performing spot I could find on the map). LOD between the amd and nvidia card seems to be a wash in dx11. Some details missing that are in the other screen for both shots, and this may be due to texture crawling because of my lack of AA.

I'll will try later to get a shot in mantle of the same location with the skyscraper up. It should be noted that there are 2 tanks, and more players visible in the mantle shot so it should in theory be more taxing than the two dx11 shots.

Mantle nearly doubles my minimums in this location. It should also be noted that even when dx11 is running at 120fps mantle has a definite perceived smoothness advantage. Something about BF4 also makes anything below 90fps @ 120hz nearly unplayable. It introduces massive amounts of input lag and stuttering.

Tests these spots on Shangai:

- Looking Straight at the mall from the northest spawn location in D node. Either from the ground floor or from the top of the building (where a fairly amount of noobs try to snipe you from).
- Looking at the car park in A node standing just outside it a the south-west corner of the car park (in the crossing of both streets) and facing nouth-east (to where the skycrapper should be behind the big building of node A).

Those 2 are the major weak spots in regards to performance under my experience in that map. In those situations mantle is a 50 to 80% fps increase and I cant really believe that there is still people that claim that those kind of gains can happen from using W8.1 instead of W7. Considering the closest those people have been to the game was when they linked a youtube video of someone playing it, its fairly easy to assume they dont have a clue about how this game works as far as performance goes.

"AFAIKs" arent good enough here, people. This is a thread of BF4 testing, the nebulous fud regarding magical performance gains done in dubious tests wont be enough. Start to think that even the 90% of the tests done in MP wont really prove anything, as people usually fail at telling which server did the test on so you can go there too and tests under the same circunstances.

 

[OCGuy]

Also, remember that DX12 is primarily to improve performance on the XB1, because that is an AMD system it will automatically transfer to AMD GPU's

I'd be interested in some elaboration on this point, it sounds interesting.

 

[Gloomy]

I'd be interested in some elaboration on this point, it sounds interesting.

I think he means that Microsoft may be pushing for cross-platform development on the Xbox One, and since that uses an AMD GPU they will also be pushing best practices for AMD GPUs.

Which isn't totally out of the realm of possibility IMO but takes too many assumptions to consider seriously at this point.

 

[omeds]

More worst case scenario testing. I have in my possession now a gtx780 and an r9 290. Siege of Shanghai is a great map for showing CPU limits in BF4 under direct X. Server is 64 players conquest.

i5 2500k @ 4.5ghz
Windows 7
Same settings for all cards.
1920x1080
All settings to lowest values
Filtering 16x
Textures to Ultra
Mesh Quality to Ultra
HBAO/SSAO Off

The gtx780 is running at 1176mhz and mem at 6502mhz(highes stable overclock on stock bios), but is downclocking because the CPU isn't feeding it at this location.

R9 290 Tri-X OC @ stock clocks (1000mhz core) in direct x 11 I didn't have afterburner installed when I took these so I'm not sure if clocks were maxed in this screenshot or not.

R9 290 Tri-X OC in Mantle

I'm not sure how much the skyscraper being down effects performance. All I know is that the gtx780 and r9 290 are hitting the same CPU bottleneck in this location (lowest performing spot I could find on the map). LOD between the amd and nvidia card seems to be a wash in dx11. Some details missing that are in the other screen for both shots, and this may be due to texture crawling because of my lack of AA.

I'll will try later to get a shot in mantle of the same location with the skyscraper up. It should be noted that there are 2 tanks, and more players visible in the mantle shot so it should in theory be more taxing than the two dx11 shots.

Mantle nearly doubles my minimums in this location. It should also be noted that even when dx11 is running at 120fps mantle has a definite perceived smoothness advantage. Something about BF4 also makes anything below 90fps @ 120hz nearly unplayable. It introduces massive amounts of input lag and stuttering.

Sorry I don't have a 2500k to test, but here's a 4770k under win 8.1. Same settings, single 780 at 1176mhz

No action

Action

As for BF4 stuttering/smoothness, it likes to be at a solid frame cap. That will make the lines in the graph smooth.

 

[HWfreak]

More worst case scenario testing. I have in my possession now a gtx780 and an r9 290. Siege of Shanghai is a great map for showing CPU limits in BF4 under direct X. Server is 64 players conquest.

i5 2500k @ 4.5ghz
Windows 7
Same settings for all cards.
1920x1080
All settings to lowest values
Filtering 16x
Textures to Ultra
Mesh Quality to Ultra
HBAO/SSAO Off

The gtx780 is running at 1176mhz and mem at 6502mhz(highes stable overclock on stock bios), but is downclocking because the CPU isn't feeding it at this location.

R9 290 Tri-X OC @ stock clocks (1000mhz core) in direct x 11 I didn't have afterburner installed when I took these so I'm not sure if clocks were maxed in this screenshot or not.

R9 290 Tri-X OC in Mantle

I'm not sure how much the skyscraper being down effects performance. All I know is that the gtx780 and r9 290 are hitting the same CPU bottleneck in this location (lowest performing spot I could find on the map). LOD between the amd and nvidia card seems to be a wash in dx11. Some details missing that are in the other screen for both shots, and this may be due to texture crawling because of my lack of AA.

I'll will try later to get a shot in mantle of the same location with the skyscraper up. It should be noted that there are 2 tanks, and more players visible in the mantle shot so it should in theory be more taxing than the two dx11 shots.

Mantle nearly doubles my minimums in this location. It should also be noted that even when dx11 is running at 120fps mantle has a definite perceived smoothness advantage. Something about BF4 also makes anything below 90fps @ 120hz nearly unplayable. It introduces massive amounts of input lag and stuttering.

Thats a healthy boost in Mantle

 

[mindbomb]

I'm getting a 7870 pretty soon. What should I expect from BF4's mantle? I understand it's optimized for gcn 1.1, effectively only hawaii based cards, cause the others aren't really suitable for gaming.

 

[HWfreak]

I'm getting a 7870 pretty soon. What should I expect from BF4's mantle? I understand it's optimized for gcn 1.1, effectively only hawaii based cards, cause the others aren't really suitable for gaming.

Well you can see my post the only thing is Mantle tends to use high vRam, so Ultra setting is not possible. i will put up a screen shot tomorow of the settings i use, its a mixture of High and Ultra, with my FPS looked to 80 thats what i get pretty much 100% of the time.

 

[VulgarDisplay]

I need a faster CPU and MOAR COARS!!!

 

[Erenhardt]

I'm getting a 7870 pretty soon. What should I expect from BF4's mantle? I understand it's optimized for gcn 1.1, effectively only hawaii based cards, cause the others aren't really suitable for gaming.

Expect to go out of vram. Mantle uses more graphics memory and you will have stuttering if you fill it all. They optimized 3+GB cards only, 2GB will be patched later.

I run my 7870 at stock
1920x1080
Ultra, high terrain, no AA,
or high preset.
GPU running on 95-100%, slight CPU bottleneck with phenomII 965 3,5GHz.
I could OC my CPU to 4GHz to see if there are any gains to be had.

 

[HWfreak]

I'm getting a 7870 pretty soon. What should I expect from BF4's mantle? I understand it's optimized for gcn 1.1, effectively only hawaii based cards, cause the others aren't really suitable for gaming.

2GB is not great for Mantle because of the way it currently manages Texture.

With 2GB of vRam the best settings you can get to be stutter free and smooth as silk are the settings in picture #1

To give the vRam some wiggle room i use the settings in pic #2

With these setting you pushing the limits of nice high FPS performance on a 7870 anyway.

PS: i have my FPS lock to 80 in user.cfg, thats why you only see 80 FPS in the screen shots.

#1




#2

 

[VulgarDisplay]

As I've been showing in my screenshots. The most benefit from mantle will come from a mid to low range CPU with a lot of GPU horsepower. If you push the bottleneck entirely to the CPU with your settings in game you can get nearly a 100% increase in frames from mantle where it counts the most: minimums.

 

[Despoiler]

I don't think Mantle is the problem with the memory management as control over that got transferred to the application. Either Dice has more work to do to better support various cards or there is a higher emphasis on the user determining optimal settings.

 

[skipsneeky2]

Thoughts on how a 2gb gtx770 and a i5 2500 non k would run BF4 and at what settings at 1080p?

Better on windows 8.1 over 7 or when gpu is maxed there's no difference?Difference no matter the settings or even if 2gb is cutting it for ultra?

Been out of the loop on this game since November,had so many issues with this game i got a quick refund but honestly wanna get back into this game but know little to nothing on this games current situation performance wise or stability wise.

 

[wand3r3r]

Thoughts on how a 2gb gtx770 and a i5 2500 non k would run BF4 and at what settings at 1080p?

Better on windows 8.1 over 7 or when gpu is maxed there's no difference?Difference no matter the settings or even if 2gb is cutting it for ultra?

Been out of the loop on this game since November,had so many issues with this game i got a quick refund but honestly wanna get back into this game but know little to nothing on this games current situation performance wise or stability wise.

This seems to be a mantle testing thread, but check out Tech Buyers Guru article on Win8.1 vs 7. Are you are thinking about buying a card for BF4?

http://www.thetechbuyersguru.com/windowsgaming.php

 

[skipsneeky2]

This seems to be a mantle testing thread, but check out Tech Buyers Guru article on Win8.1 vs 7. Are you are thinking about buying a card for BF4?

http://www.thetechbuyersguru.com/windowsgaming.php

Eventually a 290, when prices settle back to msrp or close to it. VulgarDisplays 780ti vs 290 post stunned me and honestly was trying to see how the 770 compares to those 2.

Then i wonder with how well 8.1 seems to run the game,does the os matter when you have mantle enabled with a mantle compatible card or does it only matter when DX11 is only running?

 

[omeds]

Only DX - it makes use of features only exposed under win 8.1. You should be good to go with win7 and Mantle afaik.

 

[wand3r3r]

Only DX - it makes use of features only exposed under win 8.1. You should be good to go with win7 and Mantle afaik.

Interesting, I never thought about that but it should be built into the game/AMD driver so you can benefit from 7 too. BF4 has been a game which has been demonstrating benefits from 8.1 but if you have mantle and windows 7 you may not need 8.1 (for this single game).

I'd like to see some tests on that!

 

[NTMBK]

Only DX - it makes use of features only exposed under win 8.1. You should be good to go with win7 and Mantle afaik.

Not entirely true; Windows 8/8.1 has improvements to the task scheduler which help with heavily multithreaded workloads, like a Mantle game.

 

[omeds]

Excellent. Would be great to see some comparisons.

 

[VulgarDisplay]

Eventually a 290, when prices settle back to msrp or close to it. VulgarDisplays 780ti vs 290 post stunned me and honestly was trying to see how the 770 compares to those 2.

Then i wonder with how well 8.1 seems to run the game,does the os matter when you have mantle enabled with a mantle compatible card or does it only matter when DX11 is only running?

Just so you know it was a gtx780 and not a Ti. Either way, that spot on the map shows a hard CPU limit, and a gtx780ti would have the same framerate as the other cards with my CPU on windows 7.

Anyone able to dual boot windows 7 and 8.1 and run some tests on Mantle? Do the improvements in 8.1 also translate into Mantle?

 

[DiogoDX]

DICE GDC 2014 presentation (slides) about mantle.

Rendering Battlefield 4 with Mantle
March 20 2014, By Johan Andersson

In this technical presentation from Game Developers Conference 2014, Johan Andersson shows how the Frostbite 3 game engine is using the low-level graphics API Mantle to deliver significantly improved performance in Battlefield 4 on PC and future games from Electronic Arts. He will go through the work of bringing over an advanced existing engine to an entirely new graphics API, the benefits and concrete details of doing low-level rendering on PC and how it fits into the architecture and rendering systems of Frostbite. Advanced optimization techniques and topics such as parallel dispatch, GPU memory management, multi-GPU rendering, async compute & async DMA will be covered as well as sharing experiences of working with Mantle in general.

http://www.frostbite.com/2014/03/rendering-battlefield-4-with-mantle/

What cauth my attention was their initial tests with async compute in the tile base lighting with up to 80% faster.

 

[HWfreak]

We are going to see a lot more from Mantle, we have only seen the tip of the ice berg.

 

[Phynaz]

"Not fully reliable"
"Pretty small part of frame time"

Ugh. What's the point then?

 

[DarkKnightDude]

They're demonstrating it can be done I guess. Just need some dev time from what I can tell.