[TR] A quick look at Mantle on AMD's Kaveri APU

[NTMBK]

http://techreport.com/review/26042/a-quick-look-at-mantle-on-amd-kaveri-apu

Looks promising, but those annoying latency spikes kill it a bit. Hopefully the next patch will fix that.

 

[Red Hawk]

I wonder what effect, if any, Mantle can have on the memory bandwidth bottleneck inherent to integrated graphics.

 

[Silverforce11]

Useless review, like we needed more singleplayer testing for BF4...

 

[OCGuy]

2 FPS? Not good.

 

[SPBHM]

if the game is extremely limited by GPU performance and Mantle is mostly about reducing CPU dependency, it's natural that it wouldn't give the 7600 a huge boost... it's just 384 sps at 720MHz and 128bit DDR3, don't expect miracles,

 

[ShintaiDK]

I wonder what effect, if any, Mantle can have on the memory bandwidth bottleneck inherent to integrated graphics.

None. You need to move the same assets. Mantle is mainly CPU related.

 

[Red Hawk]

2 FPS? Not good.

A solid 2 FPS when you're already lower than 30 FPS can make tangible difference. The reviewer also remarked that the better frametimes (overall, disregarding the spikes) made the game feel better to play, though he also admitted that he's become more sensitive to frametimes after all the frametime-focused reviews and analysis he's done.

 

[frozentundra123456]

Less than 30 FPS at low settings, 768p- sign me up.

 

[Red Hawk]

Less than 30 FPS at low settings, 768p- sign me up.

Give me $500 to buy a new laptop and I'll sign up.

 

[Arkadrel]

Less than 30 FPS at low settings, 768p- sign me up.

How much is a A8-7600? Around 100$ or so?

Honestly not bad for its price.

I know people here can probably find a CPU+GPU combo for less than 100$
that can do more right?

 

[Dresdenboy]

None. You need to move the same assets. Mantle is mainly CPU related.

I wonder if DirectX does memory copies when having to switch/convert between memory buffer types.

 

[Erenhardt]

There was an mantle slide from APU event that said mantle helps optimize memory usage. Developer can stage memory writes etc.
Can't find it now...

 

[OCGuy]

How much is a A8-7600? Around 100$ or so?

Honestly not bad for its price.

I know people here can probably find a CPU+GPU combo for less than 100$
that can do more right?

Can find one quite easily that smokes it for pretty much exactly $100. But if your name isnt Sony or Microsoft, and you arent buying 10,000,000. of them, AMD wont sell it to you. They also dont need mantle.

 

[frozentundra123456]

All right, I will apologize for my original answer, as it was a bit snide. However, no matter how cheap it is, I can't imagine anyone wanting to play BF4 at such compromised settings, in single player, no less. And while gameplay is smoother under mantle, except for the lag spikes which one assumes will be eventually solved, it only increases raw fps less than 10%, and is still under the generally accepted standard for playability *at the absolute lowest settings*. I don't even understand why AMD would provide the hardware for such a test, they must have known the game would be basically unplayable.

It is too bad they don't give any indication if multi-player is any better, but I would expect it would be even worse, because it is more demanding, and for sure requires a higher framerate than single player to be competitive.

 

[sontin]

The CPU performance of AMD's APUs is so bad that literally nearly everyone has a CPU with the same CPU performance at home. That means they only need to change the GPU. And for $100 you get easily twice the performance.

 

[Arkadrel]

All right, I will apologize for my original answer, as it was a bit snide. However, no matter how cheap it is, I can't imagine anyone wanting to play BF4 at such compromised settings, in single player, no less.

From that review:

The Singapore beach landing takes place outdoors, and it's quite demanding from a graphical standpoint.

Basically one of the most demanding area's in the game is where he tested it.

1366x768 @low settings and ~28 fps avg.
In the most demanding area in SP.

For a A8-7600 (not a A10-7850k), with only DDR3-2133 (@ slow timeings).

The CPU performance of AMD's APUs is so bad that literally nearly everyone has a CPU with the same CPU performance at home.

And yet alot of people dont, and/or have slower/worse performance IGPs.



Watch this:

A10-7850k , 1920x1080 + low settings multi player game of BF4 (mantle):

http://www.youtube.com/watch?v=j172NtbjxqU

FPS(In Game FPS Calculator)
Min: 16
Max: 56
Avg: 42

 

[NTMBK]

I wish they had repeated the experiment with the A10-7850K- the more powerful GPU might move some of the bottleneck onto the CPU, making Mantle more useful.

 

[3DVagabond]

I wish they had repeated the experiment with the A10-7850K- the more powerful GPU might move some of the bottleneck onto the CPU, making Mantle more useful.

Yes! I wonder why they didn't do that? :sneaky:

 

[Shivansps]

I recall that DICE tested it on 7850K and they where claming 4fps, from 39 to 43.

 

[Will Robinson]

The increase in FPS Mantle is generating seems disproportionate to the increase in "smoothness" that everyone seems to be experiencing.Now we know that frame latency and not just frames per second is the key measure of perceived smoothness in game.
The frame pacing is working so well other than that spike(which apparently is mostly fixed with the latest build) that the Mantle API needs to be judged on its overall impact and not just some random increase in frames.
When Mantle is working well there's been a very positive reaction to the feel of gameplay...that's the most important thing.

 

[Obsoleet]

I'd take any gains where I can get them. +1 for Mantle.

 

[frozentundra123456]

From that review:

Basically one of the most demanding area's in the game is where he tested it.

1366x768 @low settings and ~28 fps avg.
In the most demanding area in SP.

For a A8-7600 (not a A10-7850k), with only DDR3-2133 (@ slow timeings).




And yet alot of people dont, and/or have slower/worse performance IGPs.


Watch this:

A10-7850k , 1920x1080 + low settings multi player game of BF4 (mantle):

http://www.youtube.com/watch?v=j172NtbjxqU

FPS(In Game FPS Calculator)
Min: 16
Max: 56
Avg: 42

That is the point, isn't it? I want my system to be able to handle the most demanding areas, not just the easy parts. And yes, the 7850k, is more playable, but I was referring to the results of the specific configuration that was tested. The results almost seem like worst case scenarios for both of amds new showcase technologies: an apu choice that giver very poor performance, and a scenario that shows minimal improvement from mantle.