AMD Mantle Support possible?

[xendance 3]

Not really, the biggest fps drops are because of the simulation, not directx overhead. If you have a big fight in a city you can increase viewdistance without losing performance.

The directx overhead is very much real in all games. You can't deny that.

[nicolasroger 11]

Also I think it had something to do with the draw call wich are made by the cpu. Mantle is said to help a lot on this aspect.

On this quick review here: http://www.custompcreview.com/news/amd-catalyst-14-beta-driver-enables-mantle-frame-pacing-fix-hsa-support/19954/

we can see that mantle boost performance most on mid range CPU with high end GPU.

[calin_banc 19]

The percentage of machines that could utilise the mantle API isn't high enough to warrant development time.

The time is better spent improving the simulation engine and fixing bugs.

You only need about 1 or 2 people for roughly 2 months. So yeah, little effort, great results.

Not really, the biggest fps drops are because of the simulation, not directx overhead. If you have a big fight in a city you can increase viewdistance without losing performance.

The overhead is BIG! I got very low GPU usage ( about ~65%) even on multidisplay scenario. Tested by myself, not seen on the interwebs.

Edited January 30, 2014 by calin_banc

[Furret 0]

You only need about 1 or 2 people for roughly 2 months. So yeah, little effort, great results.

From what I've read that figure was obtained from a team making a new engine from scratch, thus had the ability to make fundamental changes to suit the new API, things become a great deal more difficult for currently existing engines.

That figure was also given out during an AMD promotional video/advert, so I wouldn't trust it until some unbiased individuals have tried it.

[griffz 1]

Still, the problem with Arma is that it require so much cpu? Is't mantle only for gpu?

it's EXACTLY the opposite. Mantle aim to reduce CPU overhead , and increase draw calls/sec

Long story short, mantle is a low level interface , built to control GPU function in a faster way than what direct3D can do . It can remove a CPU bottleneck , and bring back AMD to an Intel performance level... in a first time.

later , it can DRASTICALLY improve performances in some situation

Edited January 30, 2014 by griffz

[calin_banc 19]

From what I've read that figure was obtained from a team making a new engine from scratch, thus had the ability to make fundamental changes to suit the new API, things become a great deal more difficult for currently existing engines.

That figure was also given out during an AMD promotional video/advert, so I wouldn't trust it until some unbiased individuals have tried it.

You read it wrong then. The render was originally designed for D3D and Mantle was added later on by ONE mane in TWO MONTHS, without great documentation on how to do it, not the entire team.

[mamasan8 11]

The percentage of machines that could utilise the mantle API isn't high enough to warrant development time.

The time is better spent improving the simulation engine and fixing bugs.

http://www.dsogaming.com/news/amds-mantle-does-not-require-gpus-with-gcn-architecture/

According to one report, AMD has 36% of the market, not really all that small.

And I assume Nvidia can choose to support Mantle, if they want.

AMD = Open source

Nvidia = Closed source.

[CrazyBaron 10]

http://www.dsogaming.com/news/amds-mantle-does-not-require-gpus-with-gcn-architecture/

According to one report, AMD has 36% of the market, not really all that small.

And I assume Nvidia can choose to support Mantle, if they want.

AMD = Open source

Nvidia = Closed source.

and how many of that 36% support mantle?

[calin_banc 19]

7xxx and upwards to the future. Anyway, it will still be ignore even if it's like 1 month work on the idea of "better to focus on something else", just like 64 bit support and other perf. improvements.

[sultanofswing 10]

it's EXACTLY the opposite. Mantle aim to reduce CPU overhead , and increase draw calls/sec

Long story short, mantle is a low level interface , built to control GPU function in a faster way than what direct3D can do . It can remove a CPU bottleneck , and bring back AMD to an Intel performance level... in a first time.

later , it can DRASTICALLY improve performances in some situation

How do you expect Mantle to reduce a CPU bottleneck? If the CPU cannot process what the video card is asking for there really is not much you can do.

ARMA 3 does not need Mantle, ARMA 3 needs a new engine, point blank.

[calin_banc 19]

How do you expect Mantle to reduce a CPU bottleneck?

By using as many cores as you need or available to send data to the GPU; easy. http://youtu.be/DYcjhJGabjs

Oh, and that's the way a modern, next gen engine should look like/be.

Edited January 31, 2014 by calin_banc

[froggyluv 2130]

7xxx and upwards to the future. Anyway, it will still be ignore even if it's like 1 month work on the idea of "better to focus on something else", just like 64 bit support and other perf. improvements.

Not trying to be argumentative, but have you tried VBS 2 64bitexe? It performs worse than Arma 3

[calin_banc 19]

That's way the devs need to do their job right. Resembles the time when one of them and some defenders of Arma's current state, argued that concurrency isn't the target (while true), when it's obvious the lack of concurrency (and in general, a more up to date way of "doing things") is exactly why the game is performing bad. :)

[griffz 1]

How do you expect Mantle to reduce a CPU bottleneck? If the CPU cannot process what the video card is asking for there really is not much you can do.

ARMA 3 does not need Mantle, ARMA 3 needs a new engine, point blank.

By using as many cores as you need or available to send data to the GPU; easy. http://youtu.be/DYcjhJGabjs

Oh, and that's the way a modern, next gen engine should look like/be.

yeah , console already have this kind of low level access from what i read . in this regards, console are even faster than high end PC.

AMD showcased this change. swarm of thousand of spaceship fighting each other.

how that apply to arma3 ? i dont know.

Edited January 31, 2014 by griffz

[CrazyBaron 10]

console are even faster than high end PC.

what? they don't even outperform $600 PC

look at BF4 or any other game

xbox1: 720p, 40-60fps

ps4: 900p, 40-60fps and pretty sure they not on ultra settings....

i3+r9 270x just own them, 70fps, 900p, high, dx11 / 60fps, 1080p, high, dx11

and don't even say consoles have potential it's a joke...

Edited January 31, 2014 by CrazyBaron

[calin_banc 19]

He was probably speaking about efficiency, where the same hardware runs better under console environment. Oh, and btw, that pc rig you got there, won't hold 60fps on 64 man servers anyway, not on ultra. ;)

[CrazyBaron 10]

He was probably speaking about efficiency, where the same hardware runs better under console environment. Oh, and btw, that pc rig you got there, won't hold 60fps on 64 man servers anyway, not on ultra. ;)

i3 and r9 270x have around same hardware performance as ps4

didn't say on ultra for 900p and 1080p and tell me when consoles will use ultra settings or stable 60+ fps because i don't see it

also pretty sure it will handle it with xbox's 720p and ultra 60+\- ofc it will drop sometimes but once again console crap drops fps too

Edited January 31, 2014 by CrazyBaron

[griffz 1]

what? they don't even outperform $600 PC

look at BF4 or any other game

xbox1: 720p, 40-60fps

ps4: 900p, 40-60fps and pretty sure they not on ultra settings....

i3+r9 270x just own them, 70fps, 900p, high, dx11 / 60fps, 1080p, high, dx11

and don't even say consoles have potential it's a joke...

i was actually speaking about EFFICIENCY of console api. the pixel and texel rate of pc GPU are monster if you compare to previous console generation. i agree.

but it didn't has the bottleneck you have with directX.

On consoles, you can draw maybe 10,000 or 20,000 chunks of geometry in a frame, and you can do that at 30-60fps. On a PC, you can't typically draw more than 2-3,000 without getting into trouble with performance, and that's quite surprising - the PC can actually show you only a tenth of the performance if you need a separate batch for each draw call.

[clydefrog 3]

He was probably speaking about efficiency, where the same hardware runs better under console environment. Oh, and btw, that pc rig you got there, won't hold 60fps on 64 man servers anyway, not on ultra. ;)

A PS4 can't hold 60fps on BF4 64 man servers either, it drops below. And they don't use the equivalent of PC Ultra settings either, there is pretty much no anti-aliasing from the looks of it and generally worse image quality.

Edited January 31, 2014 by clydefrog

[scarecrow398 43]

Still, the problem with Arma is that it require so much cpu? Is't mantle only for gpu?

Mantle reduces the overhead on the CPU which aids in reducing the workload and improving performance on the CPU side of things... something quite important with Arma...

http://www.extremetech.com/gaming/175676-amds-mantle-finally-emerges-turns-out-its-actually-for-boosting-low-end-cpus-not-gpus

[Leon86 13]

But it doesn't help the battle simulation. The performance is lowest when the battle simulation puts one core on 100%. Lower cpu rendering overhead would probably be nice to enable insane viewdistances, but probably won't improve performance a lot.

[mamasan8 11]

But it doesn't help the battle simulation. The performance is lowest when the battle simulation puts one core on 100%. Lower cpu rendering overhead would probably be nice to enable insane viewdistances, but probably won't improve performance a lot.

What part of 'freeing up CPU' don't you understand? Of course Mantle would help with battle simulation as well. Since CPU spends less time on draw calls and more time on the rest.

[Leon86 13]

What part of 'freeing up CPU' don't you understand? Of course Mantle would help with battle simulation as well. Since CPU spends less time on draw calls and more time on the rest.

the core simulation isn't multithreaded apparantly, so unless you're using a dualcore it will not increase performance.

[xendance 3]

the core simulation isn't multithreaded apparantly, so unless you're using a dualcore it will not increase performance.

Umm... what? What do you mean? Are you saying that if you speed up a part of serially executed code, the whole code path doesn't get a speed boost?

[Leon86 13]

No, I'm saying it seems that the core simulation part is almost always the bottleneck if performance is very poor, it maxes out one core, the directx overhead / audio spreads over the rest of the cores, not maxing them out.