Mantle and Frame Latency

shankly1985 · 3 Oct 2014 at 13:21

I see posts around here about Mantle in some games showing less frame rate. mantle isn't all about pushing Max or avg FPS its about how they get delivered to the users.
Mantle Offers much less, latency over DirectX and its this reason meny of us say Mantle feel much more smoother.

So if you have Thief results show DirectX min 30 - Max 100 - avg 60 and then Mantle Min 30 - max - 100 avg 40

mantle will still have the best motion and smoothness over DirectX, because of the frame latency.

Digging up post from when Mantle was coming out and am sure other have results from BF4 or other ways on testing frame latency do post.

But one thing we can check, independently of frame rate, is frame smoothness. How much does BF4 improve the sometimes-rocky experience of playing games in Crossfire, over and above those improvements introduced by the XDMA engine of the Hawaii GPU family?

http://www.extremetech.com/wp-content/uploads/2014/02/BF4-FrameTime-Crossfire1.png

No hotlinked images, if you want to post an image then please rehost using your own webspace. Thank you.

This graph shows the 1% worst frame times for the AMD A10-7850K and the Haswell Core i7-4770K. Note that while performance on the dual-GPU R9 290X was flat for AMD, 1% frame times still show significant improvement. The smaller the initial number, the less noticeable the improvement is, but the figures here illustrate that latencies can improve even when frame rates hold steady.
http://www.extremetech.com/gaming/1...iggest-innovation-in-gaming-since-directx-9/3

Lower Latency

Thanks to Mantle, apps with lower framerates still feel faster because the latency is better. The team is also able to add features they weren't sure would be possible for performance reasons such as film quality motion blur and other film-style techniques. The studio also provided a demo showing several battleships in space loaded with turrets, and hundreds of fighter ships shooting lasers at fleets of enemy ships. The rendering was in real time: you could zoom out and view the whole battle while keeping an incredibly fluid framerate.

http://www.tomshardware.com/news/amd-apu-developer-conference-mantle,25079.html

Ok as you can see Battlefield 4 run @Ultra settings 1080p FOV 80 with One GPU @977/1250, Mantle Frame Latency is so much better. The game felt much better, and motion looked more smooth.

Here is a Video of the run I did, ignore the overlay I recording with Dxtory and to a slow 5400RPM Drive.
If any you guys with Nvidia would like to post your results feel free.

https://www.youtube.com/watch?v=aLsOslY7Ot8

humbug · 3 Oct 2014 at 13:30

At last some tangible evidence to what is otherwise just conjecture.

Does BF4 feel more responsive with Mantle? Yes.
Is your hit reg better? Yes.
Do more of your kill attempts result in kills running Mantle? Yes.

shankly1985 · 3 Oct 2014 at 13:35

More information that people can feel the Frame Latency Reduced when running mantle.

James J · 3 Oct 2014 at 13:37

Crossfire doesn't work with mantle yet though so surely those are single card vs multi gpu anyway?

VincentHanna · 3 Oct 2014 at 13:41

James J said:
Crossfire doesn't work with mantle yet though so surely those are single card vs multi gpu anyway?

Yes it does...

shankly1985 · 3 Oct 2014 at 13:41

James J said:
Crossfire doesn't work with mantle yet though so surely those are single card vs multi gpu anyway?

Crossfire Works in all Mantle games apart from new SE3 that just got Mantle.

The reduced frame latency isn't just for Crossfire its how mantle works.

Thief for example I played it with out crossfire because mantle xfire wasn't working yet. The gameplay over DirectX even though DirectX showed better benchmark. mantle was so much smoother.

mmj_uk · 3 Oct 2014 at 13:52

These results are meaningless.

Our evaluation used the 64-player version of the Golmud Railway map. All benchmarks were performed on the same server with a full group of people.

Since when was a game of multiplayer an accurate benchmark method? you can play 1000 different games and get 1000 different results because no two games are the same. Their results even show that the i3 4330 performed better with a single GPU than in Crossfire, no doubt because there was a lot of action in that particular game.

The only way to get accurate results is to use a fixed rail benchmark where everything that happens is the same between runs.

Rusty0611 · 3 Oct 2014 at 13:53

I get marginally better frame rates with Mantle but I can't say that there's any difference for me that is perceptible but I'm a high end CPU / single GPU user so I don't expect that I was supposed to benefit that much from it.

shankly1985 · 3 Oct 2014 at 14:05

mmj_uk said:
These results are meaningless.

Since when was a game of multiplayer an accurate benchmark method? you can play 1000 different games and get 1000 different results because no two games are the same. Their results even show that the i3 4330 performed better with a single GPU than in Crossfire, no doubt because there was a lot of action in that particular game.

The only way to get accurate results is to use a fixed rail benchmark where everything that happens is the same between runs.

Its not about frame rate so playing online wont make a difference to the frame latency.

GoogalyMoogaly · 3 Oct 2014 at 14:11

Oh good, another Mantle thread, it's been about 6 minutes so I guess another one was bound to turn up!

I seem to recall a time when the 7970 didn't have framepacing, but did have the big patch that improved framerates. I also recall this frametime thing being mentioned then. Although a good portion of the AMD users said that even though the graphs showed it they couldn't noticed any problems and it all seemed smooth to them.

Why is it that now AMD have better framerates that these people that couldn't notice the problem before can suddenly see a big difference and it's game-changing?

Ayahuasca · 3 Oct 2014 at 14:11

As Rusty has said I've used both, DX on my nvidia rig and Mantle on a mates rig and it didn't feel any smoother on either. Maybe it's just a problem for some AMD users in BF4 on DX? I've ran the overlay before to check frame times and the graph was perfectly stable with the odd huge spike just like you see in that Mantle graph above.

shankly1985 · 3 Oct 2014 at 14:16

GoogalyMoogaly said:
Oh good, another Mantle thread, it's been about 6 minutes so I guess another one was bound to turn up!

I seem to recall a time when the 7970 didn't have framepacing, but did have the big patch that improved framerates. I also recall this frametime thing being mentioned then. Although a good portion of the AMD users said that even though the graphs showed it they couldn't noticed any problems and it all seemed smooth to them.

Why is it that now AMD have better framerates that these people that couldn't notice the problem before can suddenly see a big difference and it's game-changing?

Yes another Mantle thread, only this time putting some facts across... mantle frame rate even at lower FPS is much smoother.

Thief I could tell the difference right away.. With DirectX when frame rate is low you get a motion judder effect on Mantle I have never once had this.

mmj_uk · 3 Oct 2014 at 14:18

shankly1985 said:
Its not about frame rate so playing online wont make a difference to the frame latency.

It doesn't matter, game engines respond differently to different situations which could all affect the rendering latency.

To get accurate benchmarks you need to keep variations down to a minimum which is why artificial benchmarks are fixed, benchmarking using an online multiplayer server with 64 players all doing entirely different things between runs is just plain silly.

James J · 3 Oct 2014 at 14:20

shankly1985 said:
Crossfire Works in all Mantle games apart from new SE3 that just got Mantle.

The reduced frame latency isn't just for Crossfire its how mantle works.

Thief for example I played it with out crossfire because mantle xfire wasn't working yet. The gameplay over DirectX even though DirectX showed better benchmark. mantle was so much smoother.

Ahhhh I didn't know it had to be patched in like a profile for each game. My mistake.

shankly1985 · 3 Oct 2014 at 15:05

mmj_uk said:
It doesn't matter, game engines respond differently to different situations which could all affect the rendering latency.

To get accurate benchmarks you need to keep variations down to a minimum which is why artificial benchmarks are fixed, benchmarking using an online multiplayer server with 64 players all doing entirely different things between runs is just plain silly.

Ok is they anyway we can monitor Frame latency say in Thief? Any tools for Frame latency work on Mantle?

pgi947 · 3 Oct 2014 at 15:18

Does fraps benchmark tool work in mantle? If so you can record frametimes and view them in a graph.

Frametimes outputted by the gpu are becoming irrelevant anyhow, with gsync/async managing frame delivery perfectly.

shankly1985 · 3 Oct 2014 at 15:21

pgi947 said:
Does fraps benchmark tool work in mantle? If so you can record frametimes and view them in a graph.

Frametimes outputted by the gpu are becoming irrelevant anyhow, with gsync/async managing frame delivery perfectly.

Nope
I guess could run BF4 singleplayer mission, once on both. BF4 has its own frame latency monitoring tool.

More images

shankly1985 · 3 Oct 2014 at 15:52

Anyone have a game save profile for BF4? I still not completed it. I need a profile where I can select the mission I want to play.

Going to run some tests over the weekend.

H@mmer · 3 Oct 2014 at 16:08

I'm getting horrible performance with Mantle in BF4.
Was hoping the new BF4 patch would sort it but it's just the same.
Massive CPU spikes which I don't get running DX11. Game freezing with Mantle errors. Windows running out out memory. :mad:

I'm even on the newest Cat's 14.9 running Win 8.1 x64. I7 [email protected] and 16g of ram. Maybe AMD have forgotten about us 7990 owners.
I bought this card last October for use with Mantle and BF4. A year on it still doesn't deliver.

The 7990 is a great card and at £400 nothing else can touch it. I'll be keeping it for a while as I get my 120fps in DX11 @1080 but I do feel as though I've been mislead with all the Mantle talk.

Maybe Mantle is just for GCN 1.1 and newer not for GCN 1.0. The new Cats 14.9 just seem to boost FPS with the 290 series of cards.
DICE don't seem to have fixed the memory leak in Mantle with the Fall Update.
So it looks like I'll be playing on DX11 for a while longer.

Gerard · 3 Oct 2014 at 16:13

humbug said:
At last some tangible evidence to what is otherwise just conjecture.

Does BF4 feel more responsive with Mantle? Yes.
Is your hit reg better? Yes.
Do more of your kill attempts result in kills running Mantle? Yes.

Does mantle come to a grinding halt in some situations on bf4 eyefinity? Yes.

Particularly if i'm in a tank and I engage the aps and a tankbuster is strafing me, the close up explosions can dive bomb the game to single digits, whereas on dx it'll stay around 35-40fps.

Its a pretty common situation while driving a tank around and currently mantle doesn't do a good job of providing a good experience at 5760x1080.