Richard Huddy - Gaming Scientist Interviewed

[triss]

I was trying to post snippets from that video discussing things that were being mentioned in the thread at the time. For some reason though it didn't work.

That was before Mantle hit any titles and before DX12 was even announced. At this time there was only one new API Mantle. I think AMD were hoping it could become an industry standard, why not as DX12 was nowhere to be seen at this time. They say for that to happen Nvidia would need to adopt it, but they're not sure if they would adopt it. Unless im missing something? AMD have said many times that they hoped Mantle would be the spark for change and clearly it has as sometime later DX12 popped out from nowhere. Mantle can still survive after DX12 as long as devs continue to use it and develop for it. If we're to believe what Huddy says in the video, then the majority of devs in the beta program want Mantle to continue after DX12 launches.

EDIT

I think what he was trying to say is, ultimately there can only be one API that supports everything. Mantle is not going to be that API, though they may have thought that at the time because DX12 didn't exist then.

Thanks , that the sort of insight i was looking for i probably didnt explain myself quite how i was thinking in my head ( common issue for me sorry )

 

[humbug]

I cant see Nvidia adopting mantle or AMD buying a license for gameworks LT but for either yes the option is there




http://www.techradar.com/news/compu...g-api-to-become-the-industry-standard-1218560


So once DX 12 hits and its close to mantle as devs say , Amd will close mantle down i presume as there can be only one?
Just asking btw as i thought you might have more insight into that LT

Huddy addressed this in the Video, Mantle will keep going after DX12 it currently has 47 developers using it, 30 of those have said to AMD that they want them to keep Mantle going after DX12.

Its not just a low level API, it is capable of more than that now and is still evolving, its has advantages over DX12 that will enable developers to add extra IQ, better memory management and so on.

 

[triss]

Huddy might say the pigs fly but Robert Hallock had said something else its why i asked LT for his take on it, which made sense when he explained it

I'm not disagreeing with you humbug but if its more then a API now what is it?

 

[LtMatt]

Yep no hard feelings triss. The messages were a tad contradicting but as i say DX12 was not around then and Mantle was in its infancy.

 

[humbug]

Huddy might say the pigs fly but Robert Hallock had said something else its why i asked LT for his take on it, which made sense when he explained it

I'm not disagreeing with you humbug but if its more then a API now what is it?

The article you linked is 4 months old, its no longer valid

Since then they have had feedback from the developers, 30 of which say they want to continue using Mantle after DX12.

As it stands now Mantle will keep going after DX12, it may even be in other platforms by then.

 

[GoogalyMoogaly]

There can be only one, unless it's not us, there there will be two!

I think Huddy's statements that Nvidia and Intel should see the benefits of the reduced CPU overhead suggests that they might not see the full benefits of Mantle. Plus AMD retaining control seems like the number 1 reason not to touch it, in my opinion. Especially now that DX12 has been announced and if Nvidia/Intel don't get the full benefit of Mantle then DX12 might work just as well for them.

All AMD have to do in introduce something new on their GCN 1.2 architecture that isn't compatible with Nvidia's architecture, put this into Mantle, call it Mantle 1.1 and then label all their cards as Mantle 1.1 compatible while Nvidia and Intel can only claim Mantle 1 compatibility. Then AMD just market the hell out of that fact.

Plus there's the obvious fact that if Nvidia and Intel use Mantle it's more likely to become mainstream than an optional extra. Which I don't think Intel, Nvidia or Microsoft would want.

 

[humbug]

There can be only one, unless it's not us, there there will be two!

I think Huddy's statements that Nvidia and Intel should see the benefits of the reduced CPU overhead suggests that they might not see the full benefits of Mantle. Plus AMD retaining control seems like the number 1 reason not to touch it, in my opinion. Especially now that DX12 has been announced and if Nvidia/Intel don't get the full benefit of Mantle then DX12 might work just as well for them.

All AMD have to do in introduce something new on their GCN 1.2 architecture that isn't compatible with Nvidia's architecture, put this into Mantle, call it Mantle 1.1 and then label all their cards as Mantle 1.1 compatible while Nvidia and Intel can only claim Mantle 1 compatibility. Then AMD just market the hell out of that fact.

Plus there's the obvious fact that if Nvidia and Intel use Mantle it's more likely to become mainstream than an optional extra. Which I don't think Intel, Nvidia or Microsoft would want.

We can find reasons all day about why Intel / Nvidia wont be using Mantle, I think one thing we can say for certain is that Nvidia will chose not to.

Intel may, they are already licencing multiple AMD technologies, clearly they are not adverse to using AMD technologies if its of benefit to them, Mantle is no different.

And lets not forget others, like Qualcomm. they also have no reason to ignore Mantle if its of benefit to them.

 

[triss]

The article you linked is 4 months old, its no longer valid

Since then they have had feedback from the developers, 30 of which say they want to continue using Mantle after DX12.

As it stands now Mantle will keep going after DX12, it may even be in other platforms by then.

haha i'll remember that all articles 4 months or older are now invalid
Thats most the launch mantle slides out for starters
But getting back to mantle is now more than a low level api Humbug what is it now?
Even though it will evolve isnt it still just a API like DX?

 

[GoogalyMoogaly]

We can find reasons all day about why Intel / Nvidia wont be using Mantle, I think one thing we can say for certain is that Nvidia will chose not to.

Intel may, they are already licencing multiple AMD technologies, clearly they are not adverse to using AMD technologies if its of benefit to them, Mantle is no different.

And lets not forget others, like Qualcomm. they also have no reason to ignore Mantle if its of benefit to them.

Aren't Qualcomm mostly mobile though? And Huddy said they're not looking at doing a mobile API.

When Intel have licensed AMD tech before, how often has there been an alternative (that is likely to become the market leader) as another option?

I know people will continue to say that once it has it's official sdk Nvidia will be free to adopt Mantle, but I think that's like saying "You're free to cut your leg off with a hacksaw". More than likely doable, but seems highly unlikely and really quite a stupid idea.

 

[JamesM]

haha i'll remember that all articles 4 months or older are now invalid
Thats most the launch mantle slides out for starters
But getting back to mantle is now more than a low level api Humbug what is it now?
Even though it will evolve isnt it still just a API like DX?

If you watched the video Huddy goes onto explain that if, for example, AMD release a new card (or Nvidia for that matter) that includes a new feature set it could take weeks, months before Microsoft update DX12 to take full advantage of this.

In AMDs case with Mantle, if they include a new feature set on their card, they can, in theory, update Mantle the day the new card is released so it can take advantage straight away.

Mantle will, again going from what was said in the video, offer better performance compared to DX12 on AMD cards because Mantle was designed for their cards where as DX12 has to work on a range of cards. Yes the performance gap will be reduced but Mantle should still offer better performance on AMD cards.

Take F1 for example, Mercedes make engines and also have their own team. The Mercedes engine is the best by far yet the teams with the same engine are no where near close to the Mercedes F1 team. Why? Because the car was designed with the engine.

(hope that makes sense)

 

[GoogalyMoogaly]

If you watched the video Huddy goes onto explain that if, for example, AMD release a new card (or Nvidia for that matter) that includes a new feature set it could take weeks, months before Microsoft update DX12 to take full advantage of this.

In AMDs case with Mantle, if they include a new feature set on their card, they can, in theory, update Mantle the day the new card is released so it can take advantage straight away.

Mantle will, again going from what was said in the video, offer better performance compared to DX12 on AMD cards because Mantle was designed for their cards where as DX12 has to work on a range of cards. Yes the performance gap will be reduced but Mantle should still offer better performance on AMD cards.

Take F1 for example, Mercedes make engines and also have their own team. The Mercedes engine is the best by far yet the teams with the same engine are no where near close to the Mercedes F1 team. Why? Because the car was designed with the engine.

(hope that makes sense)

And this is the reason I think Nvidia and Intel should both avoid Mantle like the plague.

 

[LtMatt]

If you watched the video Huddy goes onto explain that if, for example, AMD release a new card (or Nvidia for that matter) that includes a new feature set it could take weeks, months before Microsoft update DX12 to take full advantage of this.

In AMDs case with Mantle, if they include a new feature set on their card, they can, in theory, update Mantle the day the new card is released so it can take advantage straight away.

Mantle will, again going from what was said in the video, offer better performance compared to DX12 on AMD cards because Mantle was designed for their cards where as DX12 has to work on a range of cards. Yes the performance gap will be reduced but Mantle should still offer better performance on AMD cards.

Take F1 for example, Mercedes make engines and also have their own team. The Mercedes engine is the best by far yet the teams with the same engine are no where near close to the Mercedes F1 team. Why? Because the car was designed with the engine.

(hope that makes sense)

^

Someone has good listening skills.

 

[triss]

If you watched the video Huddy goes onto explain that if, for example, AMD release a new card (or Nvidia for that matter) that includes a new feature set it could take weeks, months before Microsoft update DX12 to take full advantage of this.

In AMDs case with Mantle, if they include a new feature set on their card, they can, in theory, update Mantle the day the new card is released so it can take advantage straight away.

Mantle will, again going from what was said in the video, offer better performance compared to DX12 on AMD cards because Mantle was designed for their cards where as DX12 has to work on a range of cards. Yes the performance gap will be reduced but Mantle should still offer better performance on AMD cards.

Take F1 for example, Mercedes make engines and also have their own team. The Mercedes engine is the best by far yet the teams with the same engine are no where near close to the Mercedes F1 team. Why? Because the car was designed with the engine.

(hope that makes sense)

While that makes perfect sense its still just a api right? A updateable api but still a api?
Didn't dx 11 come out then get updated to 11.2? i know over a slower timeframe but still, Dont get me wrong i like mantle i just dont see how its More than a API
GM does raise a very valid point though if this happened what about intel/nvidia if they accept mantle as the future

 

[JamesM]

And this is the reason I think Nvidia and Intel should both avoid Mantle like the plague.

Out of interest, why?

AMD have already said they will be releasing the SDK for Mantle in the future and in theory, again going from what was said in the video, Mantle should work just as well on Nvidia cards (within reason) as they both essentially do the same thing.

This whole thing does stink a bit of Egos and such like which is annoying as anything that offers a better experience for the gamer should be implemented imo.

Will be interesting to see what performance DX12 gives and just how close it will be to Mantle.

 

[JamesM]

While that makes perfect sense its still just a api right? A updateable api but still a api?
Didn't dx 11 come out then get updated to 11.2? i know over a slower timeframe but still, Dont get me wrong i like mantle i just dont see how its More than a API
GM does raise a very valid point though if this happened what about intel/nvidia if they accept mantle as the future

Yes it's an API, simple as that and DX11 came out in what, 2009? There have been two updates since then, not exactly a lot.

The thing is no one knows how Mantle works in depth. There may be part of the AMD GPU that works better if you do it a certain way but doing it this way may have a performance effect on a Nvidia card. DX12 has to cover all bases so, as mentioned in the video, although the performance difference will be less (Mantle & DX11 vs Mantle & DX12) Mantle should still have a performance edge with AMD cards.

It's up to the games developers to decide what path they want to take.

Lets not forget also that Mantle offers a stepping stone to console and possible mobile platforms too so it has that going for it aswell.

 

[triss]

Thanks james , thats what i thought and why i asked humbug about

Its not just a low level API, it is capable of more than that now and is still evolving

Yes it will be great if it gets on linux aswell and users aren't tied to windows unless they want

 

[GoogalyMoogaly]

Out of interest, why?

AMD have already said they will be releasing the SDK for Mantle in the future and in theory, again going from what was said in the video, Mantle should work just as well on Nvidia cards as they both essentially do the same thing.

This whole thing does stink a bit of Egos and such like which is annoying as anything that offers a better experience for the gamer should be implemented imo.

Will be interesting to see what performance DX12 gives and just how close it will be to Mantle.

As you said, if AMD release a new card with an extra feature they can update Mantle to work with it straight away.
If Nvidia release a card with an extra feature, they have to go to AMD with details of it and get AMD to add it into the API so that it can be used? What if AMD say no?
What if AMD say yes, they'll release the update for it in 6 months time?
What if they release an update but purposely make it slightly awkward for Nvidia to implement?
What if AMD start work on Mantle 2.0 but don't let anyone else have sight of it until it's officially released meaning AMD have cards supporting Mantle 2.0 by the time you've even got access to the spec. Now you have to start reworking it to work with your cards while AMD market all their cards as Mantle 2.0.

To use the F1 example, if Red Bull were competing with Mercedes for the title and each race victory, would they want to have to ask Mercedes to implement each of the changes they request? Give Mercedes insight into what they're doing?
Or they could use the new more generic engine by (for example) Audi that don't have a team of their own. Now the more teams that use Audi's engine the more even the footing is. If enough teams use the Adui engine and stop using Mercedes then it might become too expensive for Mercedes to keep R&Ding and producing their own engine if nobody else is buying it. So the Audi engine becomes standard, things are even and you've not placed control of your engine in the hands of your main rival.
Maybe I shouldn't have done the F1 example...

 

[GoogalyMoogaly]

Yes it's an API, simple as that and DX11 came out in what, 2009? There have been two updates since then, not exactly a lot.

The thing is no one knows how Mantle works in depth. There may be part of the AMD GPU that works better if you do it a certain way but doing it this way may have a performance effect on a Nvidia card. DX12 has to cover all bases so, as mentioned in the video, although the performance difference will be less (Mantle & DX11 vs Mantle & DX12) Mantle should still have a performance edge with AMD cards.

It's up to the games developers to decide what path they want to take.

Lets not forget also that Mantle offers a stepping stone to console and possible mobile platforms too so it has that going for it aswell.

You mean like the XBOX ONE console that with be running DX12?

The other thing is that, as I understand it, Mantle doesn't do what DirectX 11 does, it does what Direct3D v11 does.

 

[JamesM]

As you said, if AMD release a new card with an extra feature they can update Mantle to work with it straight away.
If Nvidia release a card with an extra feature, they have to go to AMD with details of it and get AMD to add it into the API so that it can be used? What if AMD say no?
What if AMD say yes, they'll release the update for it in 6 months time?
What if they release an update but purposely make it slightly awkward for Nvidia to implement?
What if AMD start work on Mantle 2.0 but don't let anyone else have sight of it until it's officially released meaning AMD have cards supporting Mantle 2.0 by the time you've even got access to the spec. Now you have to start reworking it to work with your cards while AMD market all their cards as Mantle 2.0.

To use the F1 example, if Red Bull were competing with Mercedes for the title and each race victory, would they want to have to ask Mercedes to implement each of the changes they request? Give Mercedes insight into what they're doing?
Or they could use the new more generic engine by (for example) Audi that don't have a team of their own. Now the more teams that use Audi's engine the more even the footing is. If enough teams use the Adui engine and stop using Mercedes then it might become too expensive for Mercedes to keep R&Ding and producing their own engine if nobody else is buying it. So the Audi engine becomes standard, things are even and you've not placed control of your engine in the hands of your main rival.
Maybe I shouldn't have done the F1 example...

I get your points and i'm no expert in any means about all this stuff.

Firstly, whats stopping Nvidia creating their own API? AMD made Mantle to offer game developers a better way of accessing the raw power of both CPUs and GPUs.

Secondly, Huddy in the video said AMD will be releasing the SDK for Mantle when they feel it's ready. Why can't Nvidia take that SDK and tweak it for their own feature sets if need be?

Thirdly I don't believe for a second that any games developer would release a game coded purely for Mantle. It would for lack of a better word be suicide. I can see games developers either coding for solely DX12 or DX12 and Mantle.

AMD aren't making anyone use Mantle, developers are choosing too once they see the benefits it offers.

Yes I fully agree that AMD could, if they wanted to, be awkward sods and restrict updates if they choose too but I hope they don't. Again we can't really discuss this as we just don't know what will happen.

Yes DX11 may be usable on the XBOX ONE but any console that implements AMD hardware should see a benefit to using Mantle. Same reasoning as a previous post, AMD designed API for AMD hardware is going to be better than a generic DX12/11

In regards to the F1 example, teams are competing within the rules. If one team betters the other because they find some golden key to performance why should the other team demand the same treatment? In other words it's up to Nvidia now to show their cards and see what they can do. Plus lets face it, you don't have to share floor plans for a car for an engine update.

 

[GoogalyMoogaly]

I get your points and i'm no expert in any means about all this stuff.

Firstly, whats stopping Nvidia creating their own API? AMD made Mantle to offer game developers a better way of accessing the raw power of both CPUs and GPUs.

Secondly, Huddy in the video said AMD will be releasing the SDK for Mantle when they feel it's ready. Why can't Nvidia take that SDK and tweak it for their own feature sets if need be?

Thirdly I don't believe for a second that any games developer would release a game coded purely for Mantle. It would for lack of a better word be suicide. I can see games developers either coding for solely DX12 or DX12 and Mantle.

AMD aren't making anyone use Mantle, developers are choosing too once they see the benefits it offers.

Yes I fully agree that AMD could, if they wanted to, be awkward sods and restrict updates if they choose too but I hope they don't. Again we can't really discuss this as we just don't know what will happen.

Yes DX11 may be usable on the XBOX ONE but any console that implements AMD hardware should see a benefit to using Mantle. Same reasoning as a previous post, AMD designed API for AMD hardware is going to be better than a generic DX12/11

In regards to the F1 example, teams are competing within the rules. If one team betters the other because they find some golden key to performance why should the other team demand the same treatment? In other words it's up to Nvidia now to show their cards and see what they can do. Plus lets face it, you don't have to share floor plans for a car for an engine update.

Nothing's stopping Nvidia creating their own API. It'd fragment the market, but I don't think either AMD or Nvidia are too worried about that. Now DX12 has been announced Nvidia may just put their eggs in that basket. If it hadn't been for DX12, who knows how things would've gone.

I'm sure Nvidia can tweak it, but the fact that AMD retain control makes me believe that anything significant needs to be actioned by AMD. New functionality, that can't be accessed through DX would suggest it's more than a tweak. If Nvidia could just add all this in without AMD's say so, then I don't see how AMD are retaining control?

I'd imagine that if the console are both using a to-the-metal type API then they should convert to both Mantle and DX12 fairly easy. You were suggesting that Mantle offered a stepping stone that DX12 didn't. If the XBOX ONE is running (or will be) DirectX 12 (not DX 11) then you'd think that's a better stepping stone than a non-Mantle API (however similar) to Mantle. DX12-varient -> DX12-varient or some-close-to-the-metal-api -> Mantle?