The AMD Mantle Thread

[3DVagabond]

That's a good analysis, except that Mantle has an even more uphill battle since it has to contend with Direct3D, specifically it having not only a massive install base, but that Microsoft having the capability and the will to continue to improve it by reducing it's overhead.

Direct3D may never have as low an operating overhead as Mantle (due to necessarily more abstraction), but it will likely come close enough to negate Mantle's biggest benefits, making it unnecessary and even counterproductive..

PhysX has competition as well, but it remains dominant and viable due to it's technical advantages such as full GPU acceleration and unique effects not found elsewhere.

And this is all AMD can hope for in the short term, unless Intel and NVidia adopt the technology as well, which seems very unlikely.

Honestly, I hope it does allow AMD to sell more cards, as it's better for consumers if AMD is a stronger competitor for both NVidia and Intel.

If M$ has the ability to substantially reduce DX overhead I wish they would do it? Also, let the devs see what's going on behind the curtain so they can better optimize. Instead we have DX vs Games with the drivers playing marriage counselor and nobody really being willing to communicate. It's heaps of extra work and expense that we don't need and it hurts performance. Worse part is we're the ones footing all the bills and getting expensive poorly performing games in the end.

Edit: PhysX isn't similar in any way.

 

[3DVagabond]

*This may have been discussed before, but this is a long thread so forgive me if it has, and I missed it. I watched the video from 'NVIDIA "The Way It's Meant to be Played" tech day', where John Carmack, Tim Sweeney and Johan Andersson talk NVIDIA G-Sync, AMD Mantle and Graphics Trends. It was a long video, but near the end someone in the audience brought up Mantle. Carmack & Sweeney were mostly opposed, primarily from the standpoint of..."we don't need another API, and what if NVIDIA and Intel responded with their own APIs. What a mess." Johan Andersson, on the other hand was already knee deep by this time(October,) but it was still early days and he wouldn't/couldn't speak about benchmarks and such. He did sound optimistic however. Now keep in mind that this was part of the big announcement of G-Sync, so later someone asked what I would call a loaded question, something to the effect of "There's clearly a lot of enthusiasm in the panel for G-Sync, would that be saying that you would be playing the games that your developing on NVIDIA hardware at this point?" Johan Andersson was not willing to agree with that. He said mutliple times that he wanted Mantle as well, so his final answer was both. As an invited guest at an NVIDIA event, talking about their new product, I think this says something about the appeal of Mantle, and I have seen a subsequent video where he discusses Mantle in more detail, and my impression is that he has gotten more excited about Mantle, for more reasons than just a low level API.

Bottom line is whoever brought up Mantle at an nVidia G-Sync event was trolling.

 

[yulgrhet]

So while console games generally have higher initial sales due to a greater volume of units, PC games make up for it in the end by having a FAR longer sales cycle.

So these are MASSIVE advantages, and are not lightly cast aside in the name of progress.

'Lightly cast aside' by whom?

 

[DamnedLife]

Also one of the key aspects of mantle is how developers can command what gets in the gcn architecture and how. One of the slides showed off 4 different versions of rendering through gcn with different levels of compute shading and rasterising each with different size of multi threading overheads. Now this actually comes completely from the consoles, the main to the metal idea of rendering. So especially Sony influenced Amd while they were creating the jaguar apus for them. This also reflects on where amd puts 8 Asynchronous Compute*Engines that can queue up to 8 commands each for a total of 64 on their 290 series exactly like it is on ps4. So amd takes cues from both amd on software and sony on hardware departments. Call them influencing hints if you like but this definitely cements my belief in amd to push mantle with great determination. Amd was in SoC business a long time but now their efforts comes bearing fruits. Sony engineers perfects the way of gcn and amd in turn uses same idea on their next gen gpus. Mantle is actually a lot closer to ps4 api than it is to xbone api and thats one of the main reasons why. Mantle will make it easier to port games between ps4 and pc. And also mantle may bring out of order rendering techniques of consoles to their in order gcn architecture. This way they may do aa aside from all other rendering going on in different thread while these are all actually sent to gcn in order not in threads.

For further reading: http://www.anandtech.com/show/7457/the-radeon-r9-290x-review/2

 

[VulgarDisplay]

I've seen certain people in this thread scoff at backward compatibility as if it were nothing important, and easily sacrificed at the Mantle altar....but that's just crazy talk if you ask me......much like that lunatic AMD executive that claimed there will be no DirectX 12..

As you say, backward compatibility (and cross architectural compatibility) are the biggest reasons why PC gaming has come so far. If PC gaming never had these attributes, it would have fragmented into oblivion long ago.

I mean, just look at consoles. Consoles completely switch hardware and programming models nearly every generation, whilst PC's basic architecture and programming model has remained consistent, but with evolutionary (and sometimes revolutionary) changes.

This allows PC gamers to play games on a wide range of hardware, and to play games that have long since ceased publication. Right now, I am playing Baldurs Gate 2 enhanced edition, an improved version of a game that came out almost 15 years ago.

So while console games generally have higher initial sales due to a greater volume of units, PC games make up for it in the end by having a FAR longer sales cycle.

So these are MASSIVE advantages, and are not lightly cast aside in the name of progress.

Why does mantle need to be backwards compatible? Nowhere in all of the publications on mantle has AMD said, "We are dropping Direct X entirely and your old games will no longer work." Certain posters that sensationalize would have you believe that. If they post anything about how mantle isn't backwards compatible they are just here to 100% derail the thread.

Of far more importance to Mantle is FORWARD compatibility. AMD has said they will continue to support direct x and mantle. Devs will continue to support direct X until such a time (though highly unlikely) that mantle completely replaces DX.

Having a GPU that runs mantle will not instantly make it stop running direct X games. That is just a completely incorrect argument against mantle. IF mantle is forward compatible (they already said it will be) then it will be backwards compatible for their future generations of GPUs that will continue to support direct X.

This argument makes no sense whatsoever.

 

[3DVagabond]

Its fake.

Can you supply the link for the quote, please? I'd like to read the entire context that the quote was taken from. Thanks.

 

[DiogoDX]

Can you supply the link for the quote, please? I'd like to read the entire context that the quote was taken from. Thanks.

Is in the video.

 

[DamnedLife]

Another fake video but the interesting part is that the guy managed to change regedit values of CCC to show the commented out portion of it and that part shows Mantle tab to either enable or disable it. http://www.youtube.com/watch?v=mHaYjVJaQqE#t=45

 

[Shivansps]

Engine support for Mantle its likely to happen on all comercial engines, they support OpenGL as well, that does not mean games running on a Mantle supported engine is going to make use of its features or even incluide as a option because of the extra work and support needed.

Until we find evidence of the contrary the only advantage for Mantle tiles will be reduced cpu requeriments, and that, to me, its not something to be hyped about.

 

[ASM-coder]

Until we find evidence of the contrary the only advantage for Mantle tiles will be reduced cpu requeriments, and that, to me, its not something to be hyped about.

But from what I've heard and read,CPU constraint is the problem. Remove that bottleneck, and FPS increases.That sounds worthwhile.

 

[VulgarDisplay]

Another fake video but the interesting part is that the guy managed to change regedit values of CCC to show the commented out portion of it and that part shows Mantle tab to either enable or disable it. http://www.youtube.com/watch?v=mHaYjVJaQqE#t=45

ROFL. Hilarious troll video haha. :awe:

 

[VulgarDisplay]

Engine support for Mantle its likely to happen on all comercial engines, they support OpenGL as well, that does not mean games running on a Mantle supported engine is going to make use of its features or even incluide as a option because of the extra work and support needed.

Until we find evidence of the contrary the only advantage for Mantle tiles will be reduced cpu requeriments, and that, to me, its not something to be hyped about.

We don't know what they will use the extra performance from mantle for yet. It could give players an advantage. Longer draw distances come to mind as a easy way to use more performance.

 

[Shivansps]

We don't know what they will use the extra performance from mantle for yet. It could give players an advantage. Longer draw distances come to mind as a easy way to use more performance.

Do you know that the vast mayority of games out there are not cpu bound on high end cpus and could support larger draw distances than today, but they just dont do it because is not worth the effort because only a minority of the player base will be able to take advantage of it?
Also dont forget that the gpu has to render it, meaning not all mantle GPU will be able to take any advantage of that, its not a "hey increase the draw distance, magic quality gain".

Look magic does not exist and the devs target the higher player base as possible, they need to consider that calculating the time/cost to see if it worth it to do it at all. Just because a few devs that are AMD partners tell that Mantle its impressive and that they gona use it does not change the reality of the gaming industry.

Devs are not going to target crazy draw calls numbers and a 290x, period.

Unless DX is dropped as the main API, if that happen we may start to see an improvement, but we all know thats not gona happen.
But it does not change the fact that probably all comercial engines are gona support it, so there is a high change that most game come with Mantle support as probably they will not have to do anything for it, its the useness of that kind of mantle support that is questionable.

 

[krumme]

Do you know that the vast mayority of games out there are not cpu bound on high end cpus and could support larger draw distances than today, but they just dont do it because is not worth the effort because only a minority of the player base will be able to take advantage of it?
Also dont forget that the gpu has to render it, meaning not all mantle GPU will be able to take any advantage of that, its not a "hey increase the draw distance, magic quality gain".

Look magic does not exist and the devs target the higher player base as possible, they need to consider that calculating the time/cost to see if it worth it to do it at all. Just because a few devs that are AMD partners tell that Mantle its impressive and that they gona use it does not change the reality of the gaming industry.

Devs are not going to target crazy draw calls numbers and a 290x, period.

Unless DX is dropped as the main API, if that happen we may start to see an improvement, but we all know thats not gona happen.
But it does not change the fact that probably all comercial engines are gona support it, so there is a high change that most game come with Mantle support as probably they will not have to do anything for it, its the useness of that kind of mantle support that is questionable.

There is plenty gpu oriented optimization. Go look eg the gpu section of Johan presentation of apu13 on youtube. And there is plenty more to the cpu utilization than drawcall. For that see the cpu section

Btw I hope anandtech and ryan gives us a more user friendly explanation. They use to do. Looking forward to it. The non benchmark part is often really detailed an easy to understand at AT.

 

[krumme]

Amd have said they have 10 times as many ressources working on dx as mantle right now.

That might change in the future. But it goes to show how backwards compatability and fragmentation is prioritized.

 

[3DVagabond]

Amd have said they have 10 times as many ressources working on dx as mantle right now.

That might change in the future. But it goes to show how backwards compatability and fragmentation is prioritized.

FWIU, once implemented, Mantle doesn't require much support from the hardware vendor.

 

[Pottuvoi]

Unless DX is dropped as the main API, if that happen we may start to see an improvement, but we all know thats not gona happen.
But it does not change the fact that probably all comercial engines are gona support it, so there is a high change that most game come with Mantle support as probably they will not have to do anything for it, its the useness of that kind of mantle support that is questionable.

At least developers can now develop for AMD GPUs as Mantle as main API.
Meaning they do not need to deal with the normal driver shit for AMD cards anymore. (writing things differently within DX to get optimal performance for AMD, just Intel/nvidia cards now.)

One thing I have been wondering about is how Mantle will change the user forced settings in games.
I doubt we will be able to change so many things anymore as driver layer should be a lot thinner and graphics under developer control.

 

[Carfax83]

If M$ has the ability to substantially reduce DX overhead I wish they would do it?

Microsoft has had an iron grip on PC gaming for how many years now? They've grown fat and lazy, with no incentive to do anything different from what they've been doing.

Mantle now gives them that incentive, as it's a genuine threat to their monopoly on PC gaming.

 

[Carfax83]

'Lightly cast aside' by whom?

By AMD, who else...

 

[bystander36]

At least developers can now develop for AMD GPUs as Mantle as main API.
Meaning they do not need to deal with the normal driver shit for AMD cards anymore. (writing things differently within DX to get optimal performance for AMD, just Intel/nvidia cards now.)

One thing I have been wondering about is how Mantle will change the user forced settings in games.
I doubt we will be able to change so many things anymore as driver layer should be a lot thinner and graphics under developer control.

They still need to optimize for AMD cards as believe it or not, many AMD users have non GCN cards.

 

[bystander36]

Microsoft has had an iron grip on PC gaming for how many years now? They've grown fat and lazy, with no incentive to do anything different from what they've been doing.

Mantle now gives them that incentive, as it's a genuine threat to their monopoly on PC gaming.

You are assuming a single API, capable of being compatible with every card made for the past 10 years, can easily be improved.

It would be easy, relatively speaking, to make an API work well on a single architecture. Making one work well for many different architectures is another thing.

 

[Carfax83]

Why does mantle need to be backwards compatible? Nowhere in all of the publications on mantle has AMD said, "We are dropping Direct X entirely and your old games will no longer work." Certain posters that sensationalize would have you believe that. If they post anything about how mantle isn't backwards compatible they are just here to 100% derail the thread.

So AMD doesn't intend Mantle to be an eventual replacement for Direct3D? I mean, how long can developers, and even AMD itself continue to support two APIs?

If Mantle isn't intended to one day replace Direct3D, then they're just wasting their time. And if Mantle is intended to replace Direct3D, then compatibility (both software and hardware) will eventually have to be confronted.

Of far more importance to Mantle is FORWARD compatibility. AMD has said they will continue to support direct x and mantle. Devs will continue to support direct X until such a time (though highly unlikely) that mantle completely replaces DX.

Well I'm glad you can at least admit that the prospects for Mantle replacing Direct3D is highly unlikely no matter how good it is.

Having a GPU that runs mantle will not instantly make it stop running direct X games. That is just a completely incorrect argument against mantle. IF mantle is forward compatible (they already said it will be) then it will be backwards compatible for their future generations of GPUs that will continue to support direct X.

How is it an incorrect argument if Mantle is intended as a replacement for Direct3D? What you're suggesting even if you don't know it, is that Mantle and Direct3D will somehow co-exist forever in some kind of harmonious relationship.....which is hogwash to be frank.

As long as developers continue to develop PRIMARILY for Direct3D, Mantle will never gain any traction; even for future games.

 

[Carfax83]

You are assuming a single API, capable of being compatible with every card made for the past 10 years, can easily be improved.

I didn't say easily, but I believe it definitely can. Microsoft has been doing exactly this since DX10, but it has been a slow and methodical process.

Direct3D is no longer upgradeable without an OS update, due to the changes in driver models that have been taking place. So they are definitely doing it, just very slowly.

Mantle should speed things up though.. And Microsoft is already on record as expressing the desire to port the Xbox One's low level API to Windows.

 

[BFG10K]

To those who say Microsoft isn't innovating DirectX because it has a monopoly - that's absolute rubbish. They have to improve things, if for no other reason than the fact that their compositing window manager is essentially a full screen Direct3D application with multiple buffers. This is yet another reason why Mantle will never supplant DirectX.

But let's assume Mantle takes off and Direct3D is dropped. The reality is that neither nVidia nor Intel is going to support Mantle. That means nVidia will push their own API and so will Intel.

Do we want AMD sponsored games to use only Mantle and nVidia sponsored to only use NVAPI? This is what some of you argue in the name of some nebulous performance improvement which has yet to be substantiated.

Do we want to go back to buying a different GPU for each game we play like the 1990s? Do some of you enjoy the prospect of having to physically swap hardware just to play certain games? Because that's where this path leads.

I mean we already had what Mantle supporters want: it was called DOS. Devs could bypass the OS completely and directly program the hardware. Is anyone going to sit here with a straight face and claim DOS was superior to the benefits of the abstraction provided by Windows and DirectX?

This abstraction is what allows me to play 15 year old games like Descent 3 on my Titan.

 

[ShintaiDK]

This abstraction is what allows me to play 15 year old games like Descent 3 on my Titan.

Good thing you didnt say Descent 2. That game was Glide only. No OpenGL/Direct3D. Only alternative was the software renderer

Classic example of the problem.

Some more games with the same issue. Much more than these btw:
Blood
Carmageddon
Descent 2
Dreams to Reality
EF2000
EF2000: TactCom
Elder Scrolls Adventures: Redguard
Extreme Assault
Gex 3D
Grand Theft Auto
Independence War: The Starship Simulator
Need For Speed 2 SE
Outlaws
Pandemonium 2
Red Baron 2
Screamer 2
Sentinel Returns
Shadow Warrior
Starfighter 3000
Test Drive: Off-Road 2
Tomb Raider
Turok
Whiplash
Croc: Legend of the Gobbos
NIRA Intense Import Drag Racing
XCar: Experimental Racing
Extreme 500
Manx TT Superbike
F-16 Multirole Fighter
MiG-29 Fulcrum
Joint Strike Fighter
Fighter Pilot
Nuclear Strike
Screamer 2
Screamer Rally