• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

DX12 CPU Performance Leaked, Way Faster Than DX11 – Threading Model Resembles AMD’s Mantle

You really don't get it, Mantle effectively bundles up commands and passes them to the driver, Nvidia and Intel are free to make Mantle drivers, it IS compatible with both companies should they make drivers. The compatibility with an API comes driver side, not with the API. It's entirely in Nvidia/Intel's hands if they choose to offer compatibility with it, not AMD's.

If that is all it is doing where does the massive VRAM usage @4K come from or the lack of CF support when it is available in DX11 ?
 
Quick question (Kaapstad)

Which games don't support Crossfire when Mantle API is used? Genuinely interested in this.

If AMD were that bothered about supporing Mantle that would do away with the V-ram mirroring in Crossfire mode, which Mantle was supposed to be able to do. Effectively we're supposed to have access to 8GB V-ram and not 4GB mirrored.
 
Does it really matter if it has its roots in Mantle or not?
If DX12 gives developer's more scope to push their games harder and faster, and it works on any compatible GPU regardless of vendor then that's a win in my book and if it had anything to do with Mantle or not is irrelevant.
 
Quick question (Kaapstad)

Which games don't support Crossfire when Mantle API is used? Genuinely interested in this.

If AMD were that bothered about supporing Mantle that would do away with the V-ram mirroring in Crossfire mode, which Mantle was supposed to be able to do. Effectively we're supposed to have access to 8GB V-ram and not 4GB mirrored.

The two that spring to mind at the moment for multi GPUs are SEV3 and Beyond Earth, the game devs do tend to get there in the end but it can take a long time.

In theory it is possible to stack VRAM with Mantle but from what I understand of it, it is a lot easier to use mirrored VRAM.
 
If that is all it is doing where does the massive VRAM usage @4K come from or the lack of CF support when it is available in DX11 ?

The game dev has to support such features, again the driver and api are working to give control to the game dev to do what they want. Game dev's are using engines that for a game today that will have started to be built anything from 3 to 15 years ago(obviously heavily modified). Game design needs to change a little because DX11 never required a game dev to offer such support, even though much of what engines do like memory management is required in game engines these days and was duplicated under DX11 because these engines work on either consoles or through high level DX, game engine design naturally follows those design goals.

Do you think a game dev engine built 4 years ago being reused was focused on providing xfire support when consoles don't have xfire, when DX took care of it on windows, even then games did need to help AMD/Nvidia offer xfire/sli scaling.

With DX12, Mantle, game design will change to included these features, we've already seen for instance Civ Earth offer a very different type of xfire that AMD doesn't provide at all under DX11.

4k, again, beta API, games that didn't explicitly think about Mantle, 4k or efficiency of memory at 4k through design, in the future these things will be factored in. There is not a game out there in which Mantle(and DX12) won't be patched in, when games are designed from the ground up for these things, things improve, welcome to the games industry.

My point exactly

When DX12 is available it will have features that the current Mantle build does not.

When DX12 is available Mantle will have features that the current build doesn't offer, what is the point here? Regardless of if MS build DX12 on top of Mantle or not, having more features doesn't mean it can't be using Mantle, having the same feature set as Mantle doesn't mean it IS using Mantle.

However, in a couple of decades MS has shown no interest in a low level API on PC's, performance has been neutered for a long time. AMD made Mantle, they announced it then they showed it to the two biggest industry API makers and said use it as you will AND for free. Here's a hint, when someone offers to save you money very few companies say no. Refuse to use Mantle, add a couple years of extra work which costs man hours and therefore money or use it as a base, save time, save money, reduce time to market by a couple of years. I know which I think is significantly more likely, everyone is free to believe what they want.

Fact is it doesn't matter and without viewing the source code side by side no one will know unless MS admitted it.
 
Assumption that will probably never be clarified so he isn't right or wrong.

However haven't heard a single MS employee saying that it hasn't lifted Mantle techniques though and said AMD were talking ****.

I have not heard a single MS employee saying that DX12 doesn't give you Ebola either. Best stock up on hazmat suits people.
 
The two that spring to mind at the moment for multi GPUs are SEV3 and Beyond Earth, the game devs do tend to get there in the end but it can take a long time.

In theory it is possible to stack VRAM with Mantle but from what I understand of it, it is a lot easier to use mirrored VRAM.

Mantle in Beyond Earth works with Multi GPU, its just that its limited to 2.
 
Last edited:
The logical VRAM pool thing was touted (in vague terms) as a future possibility. When you can write HSA code in something other than Java (and when HSA actually gets here) I think you might see this happen. The latest CUDA supposedly allows this (unified ram), as an aside. Although I'm clueless as to how any implementation would get around PCIe bus latency.

And to all the people smugly shouting down the "dx12 = mantle" people, guess what... you can't prove it either. In fact there is more evidence for DX12 being a panic reaction to Mantle than not.
 
Gotta love AMD, EA/Dice invent Mantle with thanks to MS for DX. Huddy then takes someone elses invention to MS like they were the ones that brought it to the table, do they have no shame ?
 
Would MS commit millions and an army of coders to create something that does the same thing as something that already exists and can use?

Or would they simply make what already exists work on all their intended vendors?

Reinvent bread?
 
It's interesting how the narrative here regarding Mantle has changed. Originally the AMD faithful were telling all and sundry that Mantle was based on the Xbone low level API, and how that would bring a glorious new world of super optimised console ports. Yet we now know that not only is the Xbone API currently based on evil old DX11, but Microsoft refuted the claims regarding Mantle.

Yet now the same people who once said Mantle was based on essentially DX11 (wink.gif) are now saying DX12 is based on Mantle. Personally I'd be more likely to believe that both have been underwraps for a while with development on both rather separate beasts. AMD shipped Mantle, in a rather unready state, many bugs, and support for even slightly different variations of GCN is problematic. Microsoft cannot afford that with DX, we know it is in some kind of working state as Intel had a working demo months ago, but it will need much longer in the oven to avoid the issues that (still?) plague Mantle.

Would MS commit millions and an army of coders to create something that does the same thing as something that already exists and can use?

Or would they simply make what already exists work on all their intended vendors?

Reinvent bread?

Mantle seems designed explicitly for GCN, even slight variations of which cause performance problems. Microsoft would be crazy to build the foundations of a hardware agnostic API on such a platform specific base.
 
Last edited:
Mantle seems designed explicitly for GCN, even slight variations of which cause performance problems. Microsoft would be crazy to build the foundations of a hardware agnostic API on such a platform specific base.

Would MS commit millions and an army of coders to create something that does the same thing as something that already exists and can use?

Or would they simply make what already exists work on all their intended vendors?

Reinvent bread?
 
My point being that Mantle is so focussed on one platform, that trying to make it hardware agnostic would be a bigger job than starting from scratch. Microsoft have decades of experience in creating and developing graphics API's (even low level ones as shown with the Xbox).
 
The game dev has to support such features, again the driver and api are working to give control to the game dev to do what they want. Game dev's are using engines that for a game today that will have started to be built anything from 3 to 15 years ago(obviously heavily modified). Game design needs to change a little because DX11 never required a game dev to offer such support, even though much of what engines do like memory management is required in game engines these days and was duplicated under DX11 because these engines work on either consoles or through high level DX, game engine design naturally follows those design goals.

Do you think a game dev engine built 4 years ago being reused was focused on providing xfire support when consoles don't have xfire, when DX took care of it on windows, even then games did need to help AMD/Nvidia offer xfire/sli scaling.

With DX12, Mantle, game design will change to included these features, we've already seen for instance Civ Earth offer a very different type of xfire that AMD doesn't provide at all under DX11.

4k, again, beta API, games that didn't explicitly think about Mantle, 4k or efficiency of memory at 4k through design, in the future these things will be factored in. There is not a game out there in which Mantle(and DX12) won't be patched in, when games are designed from the ground up for these things, things improve, welcome to the games industry.



When DX12 is available Mantle will have features that the current build doesn't offer, what is the point here? Regardless of if MS build DX12 on top of Mantle or not, having more features doesn't mean it can't be using Mantle, having the same feature set as Mantle doesn't mean it IS using Mantle.

However, in a couple of decades MS has shown no interest in a low level API on PC's, performance has been neutered for a long time. AMD made Mantle, they announced it then they showed it to the two biggest industry API makers and said use it as you will AND for free. Here's a hint, when someone offers to save you money very few companies say no. Refuse to use Mantle, add a couple years of extra work which costs man hours and therefore money or use it as a base, save time, save money, reduce time to market by a couple of years. I know which I think is significantly more likely, everyone is free to believe what they want.

Fact is it doesn't matter and without viewing the source code side by side no one will know unless MS admitted it.

You really don't get it, Mantle effectively bundles up commands and passes them to the driver, Nvidia and Intel are free to make Mantle drivers, it IS compatible with both companies should they make drivers. The compatibility with an API comes driver side, not with the API. It's entirely in Nvidia/Intel's hands if they choose to offer compatibility with it, not AMD's.

So Mantle is doing a lot more than you implied here.:)

As to Mantle saving the Devs money, you must be joking. Whether the devs use Mantle or not they still have to go to the same expense writing the DX11/12 version and the cost of supporting Mantle is extra.
 
My point being that Mantle is so focussed on one platform, that trying to make it hardware agnostic would be a bigger job than starting from scratch. Microsoft have decades of experience in creating and developing graphics API's (even low level ones as shown with the Xbox).

It actually the other way round, it far easier to add in the vendor necessary code path, it can be done with a much smaller team in much less time.
 
yeh dx12 has to support a ton of different hardware, if mantle helped or not it still has to be a tough job

hope it makes it into windows10 release, im liking the tech preview so far!
 
Back
Top Bottom