AMD Mantle could get you 20% more performance
Says Johan Andersson from DICE. Andersson said creating a Mantle version of the Frostbite 3 engine took them about two months to complete. Mantle support will be added by an update in late December, nearly two months after the game's release. Jorjen Katsman of Nixxes, the firm porting Thief to the PC, added that AMD's API significantly reduces overhead. The Mantle API has just 8 percent overhead whereas DirectX 11 has an overhead of around 40 percent. Katsman says it's not unrealistic to receive 20 percent extra GPU performance in games with Mantle support.
As reported by tech report:
Andersson didn't bring up performance estimates, but other developers who discussed Mantle at APU13 did. Jorjen Katsman of Nixxes, the firm porting Thief to the PC, mentioned a reduction in API overhead from 40% with DirectX 11 to around 8% with Mantle. He added that it's "not unrealistic that you'd get 20% additional GPU performance" with Mantle.
But the "pink elephant in the room," as he called it, is multi-vendor support. Andersson made it clear that, while it only supports GCN-based GPUs right now, Mantle provides enough abstraction to support other hardware—i.e. future AMD GPUs and competing offerings. In fact, Andersson said that most Mantle functionality can work on most modern GPUs out today. I presume he meant Nvidia ones, though Nvidia's name wasn't explicitly mentioned. In any event, he repeated multiple times that he'd like to see Mantle become a cross-vendor API supported on "all modern GPUs."
I've gleaned more details about Mantle, and I'll share those with you guys when I'm not scurrying between keynotes and meeting rooms. The sense I get from the developers AMD invited to APU13, though, is that Mantle yields considerable benefits in terms of development flexibility and performance, and it's worth implementing even in its current, vendor-locked state. Andersson wasn't the only developer to express a desire for multi-vendor support.
There's no telling yet whether Mantle will ever become a cross-vendor, cross-platform standard, or whether the future holds something different, such as a competing Nvidia API or a future version of DirectX with some of the same perks. One thing is clear, though: Mantle looks set to shake up the industry in a very real way.
Senior Member
Posts: 341
Joined: 2012-12-31
Really curious how is Nvidia gonna reply to Mantle. Gsync isn't gonna be the thing though it's a great idea that someone should have gotten years ago. Anyway they can't afford to let AMD gain a lead in AAA games just thanx to an API with worse hardware.
Senior Member
Posts: 7416
Joined: 2006-09-24
g-sync would be great if it wasn't tied to nvidia and to asus monitors (not even all asus monitors will support it ofc)
Senior Member
Posts: 5368
Joined: 2006-12-22
Does AMD have worse hardware ?
Senior Member
Posts: 827
Joined: 2004-04-03
Hopefully this should give my 7950 a nice boost

Senior Member
Posts: 243
Joined: 2013-03-09
Impossible. Even a simple operation like copying some memory from main RAM to the GPU will involve the CPU. Commands will still be sent to the GPU as well.
I think what AMD is getting at here is that modern GPUs don't need as much "preparation" of the data before letting it go to work on the render. DirectX hides a lot of hardware functionality behind software abstractions which helped developers work with the fixed-function nature of older GPUs. Now those abstractions are just standing in the way of the sophisticated general-purpose nature of recent GPUs.
On a side note, my take on this 20% claim is a best-case scenario. People are screaming for numbers on this Mantle thing and I think this guy let a number slip that is probably just going to disappoint in most situations.
I am more interested in the other render possibilities that may be opened up by allowing engines to use many-but-simple draw calls to achieve other effects than what we've seen over the past several years. I, for one, am ready for new games that are more than just a twist on the Unreal DX9 engine. I'd like to see some new and exotic rendering possibilities take hold - like photon mapping or metropolis light transport for example.