Bouncing Zabaglione Bros.
Legend
Maybe AMD were telling the truth when they said they made Mantle because it's what the developers kept asking them for..?
AMD Mantle API [updating]
- Thread starter MarkoIt
- Start date
What would be the developers and publishers Holy Grail?
My guess -> Write once with rapid low cost ports to every platform able to extract maximum performance from all hardware.
Take another look at Johan Andersson's GPU13 and APU13 presentations.
He not only wants Mantle operating on Windows, Linux, Apple and Android OSes and AMD, Nvidia, Intel and ARM hardware but, and here's the critical kicker, he says Mantle has the capability to do exactly that.
Mantle is extensible across OSes and all modern APUs, GPUs and CPUs and it can accommodate future architectures. All these points are contained in those presentations. Presuming of course AMD opens Mantle up to all comers. Incorporate Mantle into your engine and add extensions as applicable to accommodate new hardware, OSes and platforms.
That HOLY GRAIL is now in sight - it's DOABLE.
EA is looking into the future, say six years out, when Mantle runs on everything being sold and it can easily and cheaply port it's, by then highly optimized, PS4 console code to every platform out there ... except maybe Microsofts closed ecosystem that will be using the Xbox One's DX based code. ALL developers and publishers are in the process of grasping this reality and what it could mean to their future bottom line.
In six years, if Mantle is opened to other vendors in a timely manner, and the vendors get on board, Mantle ready hardware will be in the majority of gamers devices and developers and publishers can start pushing Mantle only ports.
That's the realizable potential, whether it will be realized is a different matter, but it's obviously a potential the developers and publishers will find sufficiently attractive to act to bring it to fruition.
I've run out of red bull. Can I have some of yours?
DavidGraham
Veteran
I must say, I read the article and it got me really excited for Mantle, the possibilities are endless. Less loading times and stutter, less memory footprint and more texture resolution, more draw distances, less CPU requirements and more multi-threaded utilization. I really want to see it expand if it can deliver on these promises.
An interesting part from the article :
An interesting part from the article :
Looks like things are not that easy on that front.
Last edited by a moderator:
If you're that advanced a coder, you'd probably be better off writing your own engine. 
Unknown Soldier
Veteran
One thing that is pissing me off is that it doesn't support older tech i.e. 5x00 and 6x00 series of AMD GPU's. I mean if they can make it support Nvidia and Intel GPU/CPU's then why not the older generation GPU's.
UniversalTruth
Veteran
Because they are still working on it and it is still some time to go until it's finished...
Anyway, regarding the CPU bottleneck and scaling. If a game runs with the same FPS at 2 GHz and 3.5 GHz, is this actually a good or bad thing if you see no scaling at all with faster CPUs? Someone can clarify?
Surely that means there is more CPU performance spare for the developers to exploit with better; sound, physics, AI, simulations, etc.
Because AMD stays in business by selling new GPUs and developing Mantle isn't cheap?
I didn't see where it was said which of those architectures they might hypothetically fit into Mantle--if any currently existing ones actually can, or have a way of doing it in a reasonably performant manner.
Older designs that are feature-equivalent to VLIW4 and below might have the same incompatibility.
If it turns out that the older AMD GPUs could be twisted to work in Mantle, but only with massive amounts of work and poor reliability, it wouldn't be up to AMD to say whether another GPU vendor is going to bother with heroics for its obsolete tech.
Older AMD GPUs don't support the descriptor sets so they handle textures completely differently. Old GPUs from other vendors wouldn't work either. For Nvidia GPUs this likely means any GPU that doesn't support there bindless texture extension.
Business wise it's a nice balance for Mantle to accommodate GCN based GPUs and Bulldozer based CPUs and APUs while providing a 'free' incentive to upgrade from older architectures.
Frostbite, and presumably other engines, will already far better utilize x86 multi-core CPUs, Mantle will just provide optimizations on top of that.
Mantle doesn't make any particular claims related to the CPU architecture, and shouldn't given Intel's dominance. Until GPUs start really caring about which CPU they link to, why should Mantle?
I don't see the relevance to the question of expanding Mantle to cover Intel or Nvidia GPUs.
I don't see the relevance to the question of expanding Mantle to cover Intel or Nvidia GPUs.
Slide 20 of DICE's GPU13 presentation teased a lot more:
FUTURE
Super excited about Mantle!
♦Will share more in the upcoming months
Tons of ideas going forward
♦Low-latency multi-GPU rendering for VR?
♦Dynamic game worlds living on the GPU?
♦...
Mantle is looking to be a combination of a Swiss Army Knife + Holy Grail for developers.
firstminion
Newcomer
It means that in your example the bottleneck is somewhere else.
What I'm really looking forward to with the release of BF4 Mantle closely followed by the release of Kaveri is AMD rigs once more becoming relevant on review sites!!!
It will also be a blast watching how the review sites handle the myriad relevant permutations that arise.
GF4 Mantle performance on any combination of Intel CPUs, AMD CPUs, AMD non-HSA APUs, AMD Kaveri, AMD GCN GPUs, Nvidia GPUs.
It will also be a blast watching how the review sites handle the myriad relevant permutations that arise.
GF4 Mantle performance on any combination of Intel CPUs, AMD CPUs, AMD non-HSA APUs, AMD Kaveri, AMD GCN GPUs, Nvidia GPUs.
