Mantle Performance Preview: LOOKS AWESOME!

Video: http://www.youtube.com/watch?v=QIWyf8Hyjbg

Mantle heavily reduces the CPU’s workload

R9-290X + FX 8350 Base Build:

As the speaker discloses after the demo (at 37:18), their CPU is actually hardly even being used. The developer presented the audience with a before-and-after frame analysis of how Mantle affects the system’s workload in comparison to DirectX. In the frame analysis taken from the game running with DirectX, we see a chunk of the CPU being used by the driver thread. In the second (the one where Mantle was used), that thread is simply gone.

Not only did it free up a lot of CPU time when switched to Mantle, but Mantle itself took only a fraction of CPU time to function as opposed to DirectX’s required driver thread. This means that developers have the option to go one of two ways if they use Mantle: they can utilize that unused CPU time to improve their game, or they can simply drop the CPU requirement. The viability of lower CPU performance requirements was later verified even further (at 39:55) when the speaker disclosed that even when underclocked to 2GHz, the FX-8350 was still waiting for the GPU to finish on any given frame. Typically, we would see things the other way around.

http://wccftech.com/amd-mantle-demo-game-gpubound-cpu-cut-2ghz/#ixzz2o4EAfAgn

Great! Can't wait to see it in BF4.

Maybe have Nvidia license Mantle? Unless this only works with the GCN architecture.

This means the end of CPU hog games such as Metro etc.?

Goodbye DirectX!

all talk until we see actual real world performance.

I don't understand why AMD would not keep this exclusive. Nvidia has no interest in making gsync available for AMD cards.

Gsync isn't the same thing.

So i might keep that phenom II of mine a couple more years

Interesting

I know that.

Why not make it open source?

Devs aren't going to alienate their user base like that.

Because AMD isn't interested in holding the industry back with proprietary BS.

That or it's because they know they can't get away with it like Nvidia can.

Either way it's best if it's open. If it's as good as the thread leads us to believe then you'd certainly want it to see more widespread use. That's better than it ending up like some of the gimmicky crap that only appears in a couple of games every year. *cough* TressFx, PhysX *cough*

Coding close to the metal sounds good until you realize that this is nonsense, its backwards. When we first got 3d hardware that's what we did, 3dfx glide? It was a fragmented market and It was an awful mess. Higher level api allowed universal development and pc hardware evolution just picked up the slack, and it still works that way. Coding close to the metal assumes no change, and that's just not how things are on the pc. Will a few titles feature it? Sure, so it happened with glide, in the long run it just won't matter.

Let alone with Gpu computing rendering CPU even less important on games any intel advantage on that front will be lost and that is a good think for AMD's cpu department and i believe that is also part of the reason for not keeping this exclusive

Apparently it could work on Nvidia cards if Nvidia chooses to adopt it. http://wccftech.com/amd-mantle-api-require-gcn-work-nvidia-graphic-cards/ Not a proprietary API. This is good for all gamers!

indeed, nvidia is the only one jumping on proprietary bs. BTW, mantle doesn't need to be adopted by the pc industry, if it takes off for consoles which all feature AMD gpu's and AMD cpus then it would transition over to the PC side very easily. So there is no separate costs with deving for mantle api and direct x. Smart move by AMD.

If they have a PS4 build, the dev can get that build and recycle it for Mantle.

As mentioned in the lecture, Mantle was designed to have enough abstraction layer for AMD's future GPU architecture which leads to Mantle could work with NVIDIA's modern GPUs.

My body is ready.

This is one reason I'm not upgrading my GPU until later next year.

I want to see where developers are going to go with Mantle thrown into the mix. At the moment I'm still planning on going Maxwell but that can change if AMD come up with something competitive and Mantle proves to be a winner.

i don't know much about Mantle... but if this actually works as good as it seems, you'd practically never have to upgrade your cpu again and just do gpu upgrades? AMD is gonna clean house if they keep this only for their video cards... only reasons i've been sticking it with nvidia is better drivers, physx, and gsync (who knows if this will pay off). but i guess if it works on nvidia also its good both ways.

this could even put hurt to intel.. this makes AMD cpus just fine for gaming if the GPU does all the work.

upgrading my video card will wait for sure now...

wow all the years my 2500k@4.2 will last me now. Will this be available to 800 series from nvidia?if not i really dont see the point of getting an nvidia card from now on (if mantle is actually used in more than 1/10 games)

no nvidia support unless nvidia decides to go ahead and take advantage of AMD's offer to use it (mantle can work on both, nvidia just needs to say yes). So for now its only AMD 7xxxx and 2xx

stupid ass nvidia... ill miss physx in BL3 but if performance boost is as big with mantle, there's no point of getting the already overpriced nvidia gpu's

I smell a little PR tbh.... These kinds of numbers are usually inflated of course, but I'm more curious about them making it available to Nvidia.

Something tells me, it's somewhat more complicated/less beneficial for Nvidia to implement, than AMD would like us to think.

Besides, does'nt Nvidia already have something similar in NVAPI? Maybe that's why they're not falling over themselves to adopt Mantle.

Agreed on all accounts.

NVAPI still relies on Direct3D.

I'm skeptical about Mantle not requiring GCN to run. The benefits from going to a low-level API stem from the fact that commands and functions are hardware-specific and optimized.

Although, the part where it acts as the layer between driver and card might work with Nvidia cards as well. If I understood the presentation correctly, Direct3D does a lot of redundant checks and handles threading poorly whereas Mantle removes those checks and does thread assignment much more intelligently reducing the amount of time it takes to actually get data to the GPU to draw a frame.

It's an interesting new tool and if it takes off I can see Direct3D and OpenGL trying to follow and make similar optimizations available. Either way, it's good for the industry.