(DSOG via MaxPC) Nvidia Finally Officially Speaks About AMD’s Mantle

[3DVagabond]

Are you sure this was meant for me? I haven't accused anyone of anything, let alone vendor lock-in.

All I'm saying is that Mantle is about thinning the intermediate layer between GPU and game to increase efficiency. DX12 will do the same.

Once both APIs are sufficiently thin, there's little reason to choose one over the other, so why not choose something that's not vendor controlled. (Just like what happened with Cg.)

If your next GPU purchase is for longer than a year, then Mantle should not be a major factor in your decision.

Sorry. quoted wrong post.

 

[SirPauly]

Specifically you accusing AMD of vendor locking it. Not whether other IHV's decide on their own not to use it. We'll see if they have a flag in there that says if there's an nVidia GPU as the main render then Mantle isn't allowed to run. That would then be the same as PhysX.

PhysX is a middleware -- and a more apples-to-apples is Cuda:

Though it has been submitted to no outside standards body, it is in fact completely free to download the specs and write CUDA apps, and even completely free to write a CUDA driver to allow your company’s hardware (CPU, GPU, whatever) to run apps written in the CUDA environment.

Nvidia “owns” and controls the future of CUDA, so it’s not open in the “open source” definition, but it’s certainly free. Nvidia tells us it would be thrilled for ATI to develop a CUDA driver for their GPUs.

Gee, sounds like what AMD will do with Mantle!

 

[3DVagabond]

PhysX is a middleware -- and a more apples-to-apples is Cuda:



Gee, sounds like what AMD will do with Mantle!

CUDA was never mentioned in the post I quoted.

Mantle is no more "open" than G-Sync, hardware PhysX, or Gameworks is. It's controlled and designed by a corporation that has an inherent conflict of interest to make it truly open. To think otherwise is a fantasy.

We'll see if it's "a fantasy" and Mantle is "no more open" than these nVidia technologies. I really believe that it will be much more open, but we need to wait and see. No point in going back and forth about it now.

 

[blackened23]

Well that depends on how silly someone wants to get with semantics, right? Huddy already stated that Mantle is theirs and AMD will maintain full control over it to optimize for their GPUs.

The problem is, it isn't truly an open standard unless it's a consortium based body that has the interest of all parties in mind. A body where all can modify and enhance the API. This is not happening. Fact of the matter is, AMD is maintaining control and nobody can mess with it. And that's fine. It's AMD's value add. Nobody cares. But don't call Mantle what it isn't. It's AMD's proprietary value add, period. And that, again, is perfectly fine.

Honestly, nobody cares. What's annoying is seeing Mantle portrayed as something it isn't. But AMD can have Mantle as their value add, and go forth to do good things. It will never be the standard since it addresses 35% of the dGPU market.

 

[3DVagabond]

Well that depends on how silly someone wants to get with semantics, right? Huddy already stated that Mantle is theirs and AMD will maintain full control over it to optimize for their GPUs.

The problem is, it isn't truly an open standard unless it's a consortium based body that has the interest of all parties in mind. A body where all can modify and enhance the API. This is not happening. Fact of the matter is, AMD is maintaining control and nobody can mess with it. And that's fine. It's AMD's value add. Nobody cares. But don't call Mantle what it isn't. It's AMD's proprietary value add, period. And that, again, is perfectly fine.

Honestly, nobody cares. What's annoying is seeing Mantle portrayed as something it isn't. But AMD can have Mantle as their value add, and go forth to do good things. It will never be the standard since it addresses 35% of the dGPU market.

Completely irrelevant to the discussion at hand. Which I will repeat one more time, "No point in arguing about it. We won't know until it's released."

 

[Noctifer616]

The problem is, it isn't truly an open standard unless it's a consortium based body that has the interest of all parties in mind. A body where all can modify and enhance the API.

You mean like the Khronos Group and OpenGL? How is that working out?

 

[SirPauly]

CUDA was never mentioned in the post I quoted.

Obviously -- its called engaging the point and adding more to it!

You also added this gem:

Did nVidia offer AMD hardware PhysX licence free? Just put it out there for AMD or anybody else to use with no strings attached?

about physx:

Nvidia maintains that the licensing terms are extremely reasonable—it would work out to less than pennies per GPU shipped.

 

[Cloudfire777]

Why should Nvidia support Mantle when they got similar results for Nvidia systems with a driver update that worked across all games?

Nvidia is pushing out CUDA6 and getting more unified control of the whole system, CPU+GPU, not just GPU like they did before, so their drivers will eventually get more control of the CPU overhead later anyway.

 

[3DVagabond]

Obviously -- its called engaging the point and adding more to it!

You also added this gem:



about physx:

So, no then. You need to license it. Again, AMD is saying this won't be the case with Mantle. We'll see when it's out.

 

[SirPauly]

imho,

I believe Mantle will be free -- similar to Cuda! Mantle will be owned and controlled by AMD - so open may not be the right word but simply free to use!

 

[Noctifer616]

imho,

I believe Mantle will be free -- similar to Cuda! Mantle will be owned and controlled by AMD - so open may not be the right word but simply free to use!

Free but not open. Just like DX, except not tied to Windows (assuming they bring Linux support).

 

[PPB]

Why should Nvidia support Mantle when they got similar results for Nvidia systems with a driver update that worked across all games?

Because people dont like to play Battlefield 4 on empty servers? No magical driver saved Nvidia from low frametimes on action packed fights in real MP scenarios. Only a fool can buy there is parity on NV with DX11 and AMD with Mantle on that title.

 

[Cloudfire777]

Because people dont like to play Battlefield 4 on empty servers? No magical driver saved Nvidia from low frametimes on action packed fights in real MP scenarios. Only a fool can buy there is parity on NV with DX11 and AMD with Mantle on that title.

I`d like to see some documentation that there is lag on the Nvidia system vs AMD system on battlefield 4.

 

[Keysplayr]

He can't produce anything like that. He can probably produce something that he read that "sounds" like what he posted, or he chose to interpret it that way. Extreme exaggeration about empty servers because anything less wouldn't even come close to making his point believable. Not that it was anyway. So I agree, documentation not only showing low frametimes (in other words unplayable) and also that only Nvidia owners can play on empty servers only. Because that is how "I" interpreted his post.
Embellish much?

 

[MathMan]

Free but not open. Just like DX, except not tied to Windows (assuming they bring Linux support).

DX development is a cooperative process where Microsoft, AMD, Nvidia, Intel, and other parties bounce off ideas. Microsoft obviously has the last say, but it is not a one way street (it can't be: they're not the ones making the HW after all.)

With AMD still holding the API hostage and not disclosing it to Intel, and AMD claiming they will keep tight control (read: one way street), it should be obvious that they are far from being 'just like DX'.

Nobody in their right mind would assign development resources to follow such an open spec. AMD didn't implement Cg either for the same obvious reason.

 

[Noctifer616]

DX development is a cooperative process where Microsoft, AMD, Nvidia, Intel, and other parties bounce off ideas. Microsoft obviously has the last say, but it is not a one way street (it can't be: they're not the ones making the HW after all.)

So, no one suggested multy-threading over the years? What about new features? AMD had tesselation in their hardware for years before DX adopted the features. What about the fact that DX11 didn't have a major update for years and thus is lacking stuff like Asynchronous Compute?

I do understand the fear of adopting Mantle which is AMD controls, but MS hasn't been doing a great job with MS. On top of that DX 12 is going to be way behind Mantle. We already have Mantle games, yet DX 12 ones are coming late next year. If the DX peace continues, won't Mantle overtake DX 12 in features quite fast? What if AMD brings out a new graphics architecture with new features and enables those in Mantle ASAP while MS waits a year or more with DX?

 

[blackened23]

What about new features? AMD had tesselation in their hardware for years before DX adopted the features.

Now that's just flat out not true. AMD intro'ed the 5870 in September 2009, and DX11 was launched a month later. Interestingly enough, the 5870 and 6970 were dogs at tessellation so it probably benefited AMD to not have many games using tessellation or DX11. The 5870 had its strong points but DX11 was an extreme weak point for Evergreen as was tessellation. I don't see tessellation benefiting your argument even despite the fact that DX11 and AMD tessellation hardware were released at roughly the same time. Let's roll back history and AMD fans actually wanted games to use DX9 when it became apparent that tessellation performance on then evergreen was pretty poor. So much so that this why you see "AMD optimized tessellation" in the control panel (I assume it's still there) to this day....................because of evergreen and cayman. Evergreen and Cayman were both dogs at tessellation (compared to Fermi), and Tahiti went a long ways towards addressing that.

 

[Noctifer616]

Now that's just flat out not true.

http://en.wikipedia.org/wiki/TeraScale_%28microarchitecture%29#Hardware_tessellation

Since 2006 AMD cards had tessellation units.

 

[ShintaiDK]

So, no one suggested multy-threading over the years? What about new features? AMD had tesselation in their hardware for years before DX adopted the features. What about the fact that DX11 didn't have a major update for years and thus is lacking stuff like Asynchronous Compute?

I do understand the fear of adopting Mantle which is AMD controls, but MS hasn't been doing a great job with MS. On top of that DX 12 is going to be way behind Mantle. We already have Mantle games, yet DX 12 ones are coming late next year. If the DX peace continues, won't Mantle overtake DX 12 in features quite fast? What if AMD brings out a new graphics architecture with new features and enables those in Mantle ASAP while MS waits a year or more with DX?

Even AMD havent implemeted all DX features. And wrongfully claim their cards are DX11.2 for example. Not to mention their lack in exactly multithreaded DX drivers.

Adoption of DX12 in games is a very rapid move. Simply because developers dont have to weight in the pros and cons for supporting an API for a niche set of potential buyers.

Also AMD have shown no intention of sharing Mantle. They have shown the exact opposite. And in terms of DX12 advancement. Perhaps the question is if AMD have tried to sabotage the progress in favour of Mantle. I think we all remember when AMD said there would be no DX12. Yet DX12 was in full development.

And in any API war, marketshare is everything. Something AMD doesnt have. So think twice before you ask for every IHV to supply their own optimized API that developers have to choose from. Not to mention where companies would pay for the exclusives. Just as we see on consoles. because not only would it be a complete disaster for the gaming industry. It would also put those companies with weak marketshare and small pockets out of business very fast.

 

[ShintaiDK]

http://en.wikipedia.org/wiki/TeraScale_%28microarchitecture%29#Hardware_tessellation

Since 2005 AMD cards had tessellation units.

Didnt those tessellation units turn out to be so weak that they was essentially useless? It sure took AMD a long time to bring useful tessellation performance to the PC instead of just a checkbox marker.

 

[blackened23]

edit: nevermind

 

[ShintaiDK]

Anyway, 5870 was the first AMD part with full hardware tess.

And the performance there was so bad, that AMD even had tessellation performance in slides for the HD6xxx.

 

[blackened23]

And the performance there was so bad, that AMD even had tessellation performance in slides for the HD6xxx.

Actually, i'll have to correct myself. It appears that TruForm did have dedicated tessellation hardware, but it was not the same as what was required by Shader Model 5.0. So I guess, he is halfway correct. It wasn't the same type of "hardware" tessellation that is broadly adopted today, but there was minimal hardware for Truform in older ATI GPUs. Shader Model 5.0 had different requirements for hardware tessellation and it was not the same as ATI's TruForm. This is why, IIRC, that AMD's drivers lock out tessellation on older than 5870 parts.

This changes nothing though. The first flagship part using Shader Model 5.0 was the 5870 and it was, quite frankly, terrible at tessellation.

 

[Noctifer616]

I think we all remember when AMD said there would be no DX12. Yet DX12 was in full development.

Mantle talks started in early 2012. Actual development in early 2013. By the end of 2014 we will have the final version out while games with Mantle support are already out. Are you telling me that MS needs more than 2 years to finish DX 12 when AMD can do so in less time with their limited resources?

If DX 12 was in full development back then we would have games with DX 12 right now.

With that said, at the time of the interview DX 12 didn't exists, it might have been talked about by MS, but actual development probably didn't start. I would imagine that MS found out about AMD working on Mantle sometimes in mid 2013 which gave them enough time to make an early alpha version of the API and demonstrate it at GDC.

 

[ShintaiDK]

Mantle talks started in early 2012. Actual development in early 2013. By the end of 2014 we will have the final version out while games with Mantle support are already out. Are you telling me that MS needs more than 2 years to finish DX 12 when AMD can do so in less time with their limited resources?

If DX 12 was in full development back then we would have games with DX 12 right now.

With that said, at the time of the interview DX 12 didn't exists, it might have been talked about by MS, but actual development probably didn't start. I would imagine that MS found out about AMD working on Mantle sometimes in mid 2013 which gave them enough time to make an early alpha version of the API and demonstrate it at GDC.

Mantle is still beta, correct? Mantle may first be release ready when DX12 is.

DX doesnt ship as beta. It ships as a final product.

See the difference?

And DX12 have been in development for years. Unless you say MS, Intel, nVidia and Qualcomm is lying.

But funny enough, AMD likes DX12 so much it can barely get its hands down.

Even 3DMark already runs DX12:

Not to mention the Forza demo ran on nVidia. And the 3DMark on Intel Iris Pro.

Who is missing?