DirectX vs Mantle current and long term

[Despoiler]

Yes, Microsoft is on record saying that they are going to port the low runtime and tooling capabilities of the Xbox One API to Windows:

Source

I also find it very ironic that all of these Mantle die hards are bashing Direct3D, when Mantle burrows so much tech from Direct3D in the first place..

Built from the ground up my foot

Are you ever going to make any attempt to understand the topics you are commenting on? It's pretty clear you still haven't read the Dan Baker interview or if you did you didn't understand the tech. The link is below if you missed it.

http://www.maximumpc.com/AMD_Mantle_Interview_2014

 

[Gloomy]

AMD's DirectX drivers are their own work, not Microsoft's. I'm sure they used more than the HLSL bits. Why wouldn't they? :|

 

[Spjut]

That's a silly assertion, because we all know the kind of "proof" you're looking for is for some AMD exec to come out and make a public statement such as, "Yeah, we paid EA/Dice a ton of money for exclusive rights to BF4!"

Like they would ever give the public such information. Deals like that happen behind closed doors, but everyone knows it's a regular occurrence in the industry.

According to BF4 central, AMD paid EA up to 8 million for Battlefield 4 deal.

BF4 was a major PC game, that would drive hardware sales and adoption. AMD has exclusive bundling rights, and the BF4 17 minute trailer was done on AMD hardware. Also, the Frostbite 3 engine and nearly all of it's titles will get Mantle.

AMD knew exactly what it was doing when it ponied up that kind of cash for BF4 and Frostbite 3..

And not once is there a specific sum mentioned for Mantle. That 8 million dollars figure is supposedly a lump sum for everything.

Did AMD pay Dice for incorporating Mantle support? Very likely, but unless you work on AMD or Dice, you don't know how much it cost them. Heck, for all what we know it could be on the same level as the typical GE or TWIMTBP title.

 

[frozentundra123456]

I don't really care that much who paid who for what, or who asked to develop mantle. As the saying goes "it is what it is" and will have to stand or fall on its own merits. Unless Intel and nVidia adopt it, it will at best remain an adjunct to direct x. And for both technical and political reasons, I can't see either adopting it. When you think about it, AMD is in a very strange position in this regard. Unless Intel and nVidia adopt mantle, it will never become an overall API like direct x. But if they were to adopt it, AMD would lose all or part of their competitive advantage, depending on how well mantle ran on non-gcn hardware.

And even if all 3 did adopt it (very unlikely IMO) you will still need dx to play all the games not designed for mantle.

 

[hungtran]

AMD would love it if Intel and Nvidia adopted Mantle. It would make their CPUs a lot more competitive for gaming, but that's also a reason why Intel won't ever adopt it.

 

[frozentundra123456]

AMD would love it if Intel and Nvidia adopted Mantle. It would make their CPUs a lot more competitive for gaming, but that's also a reason why Intel won't ever adopt it.

Theoretically yes, but so far, we have only seen that happen with really weak apus paired with a powerful discrete gpu. According to the latest BF4 multi player results I have seen, FX, i5, and i7 cpus all surprisingly show similar benefits. Certainly if mantle were to become universal, it could affect the choice of a gaming CPU however.

 

[sontin]

I dont think Mantle will chance anything. Intel processor are still much faster than AMD's product:

http://www.pcgameshardware.de/AMD-R...ecials/AMD-Mantle-Test-Battlefield-4-1107754/

Mantle is not solving the per core performance. It allows more draw calls because it uses less calls than DX for the same work but in the end you still need a processor with a very good per core performance.

 

[hungtran]

People upgrade CPUs mainly for gaming, and Mantle can potentially close the gap between Intel and AMD when considering gaming performance/$. The enthusiasts will always get the fastest CPU for an extra 20 FPS, but that's not the majority of the market, certainly not for the OEMs.

At the other front, AMD wants APUs to be their future. Their goal is probably to target all Mantle games to run at above 30 FPS on medium settings on their APUs eventually. If Intel wants to compete there, they'll need to undercut their higher-end offerings with their CPU+iGPU pricing. In any case, Mantle adoption will increase AMD's CPU market share. You can argue that Mantle won't change a thing, but AMD's position without Mantle is a lot worse. You can't deny that.

 

[frozentundra123456]

People upgrade CPUs mainly for gaming, and Mantle can potentially close the gap between Intel and AMD when considering gaming performance/$. The enthusiasts will always get the fastest CPU for an extra 20 FPS, but that's not the majority of the market, certainly not for the OEMs.

At the other front, AMD wants APUs to be their future. Their goal is probably to target all Mantle games to run at above 30 FPS on medium settings on their APUs eventually. If Intel wants to compete there, they'll need to undercut their higher-end offerings with their CPU+iGPU pricing. In any case, Mantle adoption will increase AMD's CPU market share. You can argue that Mantle won't change a thing, but AMD's position without Mantle is a lot worse. You can't deny that.

The problem with that reasoning is the lack of games that support mantle. One cannot base a cpu buying decision based on mantle, because right now only one game supports it, and what, maybe at best 4 or 5 games by the end of 2014? If one is building a system, it makes no sense to base a cpu buying decision on capabilities in a vast minority of games.

So far, mantle shows most of its benefits with cpu limited situations, and APUs are graphics and bandwidth limited. The data I have seen with APUs on the igp is that mantle shows around a 10% or less benefit. If they could get mantle games designed for still lower end APUs like kabini, which still have decent graphics but really weak cpus, they might show more benefit.

 

[3DVagabond]

Well it just goes to show that Mantle isn't only about reducing CPU overhead..

That's an extremely expensive form of motion blur, and given the performance hit, no developer worth their salt would ever implement it in their game..

Games aren't movies, they're just games..

I'm pretty sure you actually believe that, but why should I think you know? I'm only trying to understand the knowledge base you are working from to make this statement. Are you a game dev? Are you saying that Dan Baker "isn't worth his salt as a dev"? That is a very definitive and damning statement you've made which I think you need to back up with some facts and/or a background statement about yourself that shows your expertise to make that claim.

 

[3DVagabond]

AMD would love it if Intel and Nvidia adopted Mantle. It would make their CPUs a lot more competitive for gaming, but that's also a reason why Intel won't ever adopt it.

It does nothing to make AMD CPU's any faster relative to Intel CPU's. All it's doing is distributing the workload across the CPU cores more efficiently. Last I checked, Intel made multi core CPU's as well, so it will make their CPU's more efficient just like it does AMD CPU's.

 

[VulgarDisplay]

I dont think Mantle will chance anything. Intel processor are still much faster than AMD's product:

http://www.pcgameshardware.de/AMD-R...ecials/AMD-Mantle-Test-Battlefield-4-1107754/

Mantle is not solving the per core performance. It allows more draw calls because it uses less calls than DX for the same work but in the end you still need a processor with a very good per core performance.

Its funny how mantle won't change anything because Intel CPUs are faster. It must be a fluke that my minimums in bf4 went from 74 fps on dx11 with a 4.5ghz 2500k to being locked at my frame rate cap of 120fps. 46 fps higher minimums than DX 11 is a HUGE deal.

I play at cpu limited settings to help maintain smooth gameplay, but that doesn't change the fact that mantle can give me 46 more fps when I really need it.

 

[hungtran]

Because the CPU bottleneck gets smaller for gaming, and that improves performance/$. Not talking about straight up IPC comparisons because AMD CPUs would get destroyed. You have to take into consideration AMD APUs will be below the $100 price point, maybe closer to $50 or less for OEMs. Intel can't compete against AMD's integrated graphics performance at those prices without gutting their margins, which becomes more pronounced with Mantle.

So you reduce the CPU gap and hope for more developer adoption of Mantle and HSA. We are at beta 1 stages right now with the first game. Things will only improve from here on out for Mantle, but if and only if more developers see the value proposition. It'll be feast or famine I think. If the ball ever gets rolling with adoption, it will snowball fast. There'll be a lot of politics going on with Nvidia and Intel with the developers and OEMs for sure in the coming months. Who know what'll happen, it'll be interesting to observe nonetheless.

 

[OCGuy]

Intel can't compete against AMD's integrated graphics performance at those prices without gutting their margins, which becomes more pronounced with Mantle. .

Any data in particular that shows Mantle giving a tangible benefit on a APU?

 

[hungtran]

Due to the simple fact that AMD can improve their graphics performance faster than they can improve their CPU performance on their APU's. Sooner than later, the CPU performance of the APU will be holding back its graphics performance in games. Mantle is obviously not meant for just the current gen APU's.

 

[bystander36]

Due to the simple fact that AMD can improve their graphics performance faster than they can improve their CPU performance on their APU's. Sooner than later, the CPU performance of the APU will be holding back its graphics performance in games. Mantle is obviously not meant for just the current gen APU's.

If gaming can be greatly improved by more cores, easily, there is hope that CPU's can progress in the same manner that GPU's do. They can start increasing cores and net real improvements. This is assuming that the multithreading that Mantle or future improved API's can scale to as many cores as there are present.

 

[Carfax83]

Are you ever going to make any attempt to understand the topics you are commenting on? It's pretty clear you still haven't read the Dan Baker interview or if you did you didn't understand the tech. The link is below if you missed it.

http://www.maximumpc.com/AMD_Mantle_Interview_2014

What does that have to do with what I said?

If DirectX was stagnant, then maybe you and Dan Baker would have a point. But DirectX will continue to evolve and might very well come close to equaling Mantle in terms of overhead and threading..

It's already gets close on NVidia's D3D pathway, a long with DX11 MT.

 

[Carfax83]

I'm pretty sure you actually believe that, but why should I think you know? I'm only trying to understand the knowledge base you are working from to make this statement. Are you a game dev? Are you saying that Dan Baker "isn't worth his salt as a dev"? That is a very definitive and damning statement you've made which I think you need to back up with some facts and/or a background statement about yourself that shows your expertise to make that claim.

No I'm not a developer, but common sense dictates that any IQ enhancing feature that causes a 400 to 600% performance decrease to unplayable levels is too expensive to include in any game where performance affects quality of the experience.

Star Swarm is a benchmark, so it's understandable in this case, but if this form of motion blur makes it's way into any games based on the Nitrous engine, then I'll be absolutely shocked.

And it's not like it's even improving IQ to a substantial degree either. More like degrading it.

 

[Will Robinson]

Almost like oceans of tessellation ne?

 

[el etro]

... Anandtech speculates that AMD's new Mantle API might be a direct copy of Xbox One's low level

Is speculation.

 

[gdansk]

Furthermore, even if it were a direct copy of the Xbox One API, Microsoft & AMD cooperated on it. It's not like AMD can walk in and copy an API, they have to implement each function within the API or arrange a license for the implementation. That doesn't make much sense considering they have mentioned it will be an open specification.

If we're speculating,
I don't think Microsoft is much involved in Mantle itself.
I think it's AMD taking their own hardware abstraction API, streamlining it, and making it for available other developers target it is as well. AMD will use it to rewrite their OpenGL/Direct3D drivers in Mantle. Then for all future GCN architectures, they simply port Mantle then both OpenGL/Direct3D will work by doing Mantle API calls.
/endspeculation

 

[BrightCandle]

Actually they can just copy the api. That is what the whole Google v oracle lawsuit was all about. An api is public knowledge and is required for compatibility, it would actually be anticompetitive to stop a compatible copy from being made. Now normally it's very hard to implement the same api as the hidden bits can be tricky to get right but its absolutely allowed under normal USA practices and law, they do not need a licence for that.

 

[Carfax83]

Almost like oceans of tessellation ne?

That was already debunked by Crytek themselves. In wireframe mode, there was no culling which meant you could see everything that was tessellated, including the ocean. Only uninformed people perpetuate the idea that Crysis 2 was "over tessellated."

Crysis 2 tessellation implementation is actually one of the best in any game..

 

[sontin]

Despite draw calls unnecessary work (control points and triangles) gets culled by the Hull Shader and Early-Z units. So you cant "over tessellate" anything.

 

[cytg111]

.. assertion...

I have no assertion but the written words by you, I make no assumptions at what you may *really* mean or anything in between, so assertion is on you cowboy.
As it may be you have asserted yourself into trolldom.
That, frankly, was the only objective, cause I'd just hate for new readerships to come read your gospel and take it as truth.
You're kind of the bibel guy that just wont go away. Later.

edit:

That was already debunked by Crytek themselves.

Which is kind of like you exonerating yourself of being "creative" with the facts.
Yea, the rest of the world dont move like that.