AMD porting Mantle to Linux/SteamOS

[cageymaru]

That explains the current workflow we see today, involving only DirectX and OpenGL. It doesn't speak to any sort of fallback renderer for Mantle.


Right, so they're saying here that the port from DirectX to Mantle takes about four-to-six weeks. There's still no mention of any kind of fallback renderer for Mantle games on Linux.


So they'll spend four-to-six weeks to port to OpenGL (for everyone) and an additional four-to-six weeks to also port to Mantle (for users with AMD cards, who could already run their OpenGL port)?

The article doesn't actually state that, but even if it did, that seems like an odd division of effort.


They're still stuck porting to OpenGL if they want to support more than just a small sub-set of Linux users with Mantle-capable AMD graphics cards.

What is your statement about the fallback renderer supposed to mean? I took it to mean that you thought that Windows DX11 games that also happened to use a Mantle rendering path may not have an OpenGL path for Intel and Nvidia users. That's why I posted what I said before. No harm; no foul.

 

[Yakk]

Hmm, with Mantle on Linux, and Apple going towards Metal... not looking good for OpenGL.

 

[wonderfield]

50 games being developed have chosen to support it.

What is the list of these fifty games? Wikipedia lists ten. One of which will presumably never actually get a full implementation.

Plus one extra month of development time isn't going to run you out of business if you can sell your game to people running SteamBox computers that look like this. They come in Intel and AMD variants.

They can already sell to Brix owners.

OpenGL 4.0 is supposed to be the DirectX 12 equivalent for Linux from what I understand. If I'm wrong feel free to crucify me.

There is no OpenGL equivalent for DX12. The latter is a low-level API, while the latter is, for the most part, a high-level API. Extensions can get you relatively close to metal, but architecturally speaking, the core API is unlike Mantle or DX12.

If OpenGL was so fabulous then why aren't there tons of triple A titles using it?

Three main reasons:

A) Khronos isn't structured to give support to developers the way Microsoft does. Microsoft has a team that works with developers in Direct3D implementations. Khronos doesn't. For commercial development work, that's often important.

B) The driver situation is still erratic. NVIDIA is the only vendor who genuinely has their shit together (Intel actually has two entirely separate teams doing OpenGL work on Windows and *NIX platforms), and although NVIDIA dominates the discrete market, developers are unwilling to gamble on OpenGL driver issues on AMD and Intel hardware. id took the gamble on AMD's driver team pulling their weight and lost big with Rage. Things aren't so rosy for their last title on AMD hardware, either. Intel is in a better position, quite oddly, but Intel hardware still currently excels with Direct3D. GL perf and compatibility are still an issue.

Everyone does a good job on their D3D implementations, on the other hand. There's little risk in going down that path, and minimizing risk is simply sound business practice.

C) Some middleware does not provide support within an OpenGL context. This is becoming less of an issue.

AMD actually claimed that they would bring expose most of Mantle's features through GL vendor extensions, but they seem to have made no progress on that from what I can tell. Much as there's been no apparent progress on Mantle standardization.

 

[KazeoHin]

Wow, for months Unknown-One and Prime1 have been saying "It's never coming to Linux!" "It's a closed API, it will never break out of windows"

Now that news has been released and can actually FEEL the anger steaming off of them.

InB4 Unknown-one follows this up with "that's not what I said."

 

[jbltecnicspro]

Wasn't direct X supposed to be just that ? Look now ,the industry is being parted on dx11 already, Nvidia is pushing gameworks to make newer effects for their cards only.While dictating how developers input dx11 code.

AMD's received builds of games a month before release and optimizing to have black box .dlls changed just before release. Game developers are contractually obligated to implement any dll change nvidia provides,so you can't blame the developers for that. It's a war on resources and we all know amd has less resource. It's brilliant Win win .Just not good for gamers. (See watchdog release ran like crap on both lol)

You don't get any of that with mantle..,tho they can... as there was talks bout implementing additional rendering techniques.

Future looks bleak ,divided either way ,even more when developers give different visual experience per gpu brand. Maybe intel can come up with their own effects, some type of realistic BloodFX, Bloodinside....Blood+ . Runs best on intel hardware.

Looking forward to lower cpu overheard in dx12 , mantle , on windows or linux either way.

Yeah, you're right. So much bullshit either way you slice it. That said, I'm also looking forward to less CPU overhead and all the other techs and what they'll bring.

 

[Unknown-One]

Wow, for months Unknown-One and Prime1 have been saying "It's never coming to Linux!" "It's a closed API, it will never break out of windows"

Now that news has been released and can actually FEEL the anger steaming off of them.

InB4 Unknown-one follows this up with "that's not what I said."

That isn't what I said.

I never claimed anywhere on this forum that the API wouldn't become available on Linux. You're literally making shit up now.

 

[PRIME1]

Wow, for months Unknown-One and Prime1 have been saying "It's never coming to Linux!" "It's a closed API, it will never break out of windows"

It's still not on Linux, It's still not open. I'm still right.

 

[trick0502]

It's still not on Linux, It's still not open. I'm still right.

that makes you sound like a whiny bitch

 

[Brackle]

that makes you sound like a whiny bitch

Because it isn't Nvidia. Now if it was Physx and Gameworks he would be praising it!

 

[PRIME1]

Because it isn't Nvidia. Now if it was Physx and Gameworks he would be praising it!

Physx is available on Linux and is open (at least compared to Mantle).

Here is the publicly available SDK.

https://developer.nvidia.com/physx-sdk

Platforms: Win, OSX, Linux, XBOX®, PlayStation®, Android, ioS

 

[Unknown-One]

What is your statement about the fallback renderer supposed to mean? I took it to mean that you thought that Windows DX11 games that also happened to use a Mantle rendering path may not have an OpenGL path for Intel and Nvidia users.

I was just trying to draw attention to the fact that they don't seem to be addressing the issue of OpenGL at all (OpenGL being the only real option for a fallback renderer for Mantle under Linux).

If developers are forced to do an OpenGL port anyway (because 2 out of 3 vendors don't support Mantle), then Mantle doesn't serve to reduce any barriers when porting games to Linux... it just adds an additional codepath to maintain.

The performance boost Mantle brings with it is nice on the hardware that supports it, but it only helps Linux users who happen to have AMD cards... that's a very small fraction-of-a-fraction of customers.

 

[cageymaru]

I was just trying to draw attention to the fact that they don't seem to be addressing the issue of OpenGL at all (OpenGL being the only real option for a fallback renderer for Mantle under Linux).

If developers are forced to do an OpenGL port anyway (because 2 out of 3 vendors don't support Mantle), then Mantle doesn't serve to reduce any barriers when porting games to Linux... it just adds an additional codepath to maintain.

The performance boost Mantle brings with it is nice on the hardware that supports it, but it only helps Linux users who happen to have AMD cards... that's a very small fraction-of-a-fraction of customers.

But if AMD stops making regular desktop chips and only makes APU's, then having a Mantle path would allow for more sales to those users. I guess you could tell them to buy a video card like the Intel crowd. But personally I would try to capture that APU crowd. 4 - 6 weeks of more development for more sales sounds like a positive to me.

 

[Unknown-One]

But if AMD stops making regular desktop chips and only makes APU's, then having a Mantle path would allow for more sales to those users.

That's still catering to a fraction, of a fraction, of a fraction of the market.

A fraction that could still run an OpenGL version of the game, as long as it scales decently with lower graphics settings. Doesn't it make more sense to devote more time to making sure the OpenGL version (which everyone can run) will scale, than to devote time to Mantle (which only helps a fraction of customers)?

I guess you could tell them to buy a video card like the Intel crowd. But personally I would try to capture that APU crowd. 4 - 6 weeks of more development for more sales sounds like a positive to me.

No reason the APU crowd can't run OpenGL. I'd rather see those 4-to-6 weeks spent on optimizing OpenGL (which helps everyone) rather than implementing Mantle (which helps a very small portion of users).

Also, the faster APU's get (and they are getting faster), the less Mantle matters. That means developing for Mantle fills a small niche that's only going to get smaller as hardware in that range of devices becomes better-able to soak up the overhead of more-classic (and more-widely-supported) graphics APIs.

 

[The Mac]

There are other advantages to mantle other than just speed improvements.

Asychonous memory access for example

 

[KazeoHin]

It's still not on Linux, It's still not open. I'm still right.

Ha! Oh man, you're like a cartoon character! Are you paid by AMD to make Nvidia fans look bad? At least Unknown-One knows when to shut up and stop replying when he has nothing to argue with.

 

[Unknown-One]

There are other advantages to mantle other than just speed improvements.

Asychonous memory access for example

Which doesn't really help Mantle's case. Devs also have to do an OpenGL port that doesn't have access to those advantages, and the OpenGL port is more important, as it will be running on more systems than the Mantle renderer.

Once again, you're talking about maintaining an additional codepath, while also ultimately doing nothing to reduce any barriers to porting games to Linux.

At least Unknown-One knows when to shut up and stop replying when he has nothing to argue with.

Uh, right... Because me pointing out that you're literally making-up things I've never said is somehow equivalent to "shutting up because I've got nothing to argue with." Get a clue man...

"Lulz, told you so! You said it would never happen"
"None of my posts said it would never happen"
"Lulz, he has nothing to argue with!"

Do you realize how ridiculous that exchange looks?

 

[cageymaru]

That's still catering to a fraction, of a fraction, of a fraction of the market.

A fraction that could still run an OpenGL version of the game, as long as it scales decently with lower graphics settings. Doesn't it make more sense to devote more time to making sure the OpenGL version (which everyone can run) will scale, than to devote time to Mantle (which only helps a fraction of customers)?


No reason the APU crowd can't run OpenGL. I'd rather see those 4-to-6 weeks spent on optimizing OpenGL (which helps everyone) rather than implementing Mantle (which helps a very small portion of users).

Also, the faster APU's get (and they are getting faster), the less Mantle matters. That means developing for Mantle fills a small niche that's only going to get smaller as hardware in that range of devices becomes better-able to soak up the overhead of more-classic (and more-widely-supported) graphics APIs.

No, AMD is doing ARM and x86 APU's from now on. So if you want to sell to people with an AMD chipset, w/o a video card, then you need to have the proper tools in your software so that it runs the best that it can on their hardware. I'm going to make the assumption that you own an Intel APU; not the Iris Pro. Can you run Battlefield 4, Civ 5, Thief on it without a video card and above 30fps? If you can then your statement is correct. If not then more work needs to be done to make that a reality. The AMD solution is called Mantle and it works very well.

 

[Unknown-One]

No, AMD is doing ARM and x86 APU's from now on. So if you want to sell to people with an AMD chipset, w/o a video card, then you need to have the proper tools in your software so that it runs the best that it can on their hardware.

What are you saying "No" to, exactly? You don't think AMD's APU's are going to get faster over time? If that's the case, it's a dead-end platform... Sorry, don't believe that. I'm pretty sure we'll continue to see faster and faster APU's from AMD, which means API overhead matters less-and-less over time.

Also, Mantle isn't required to do game-design for an APU. I assure you, they're capable of running OpenGL without a problem, just like Intel and Nvidia GPUs.

You're also not addressing the issue of Mantle only being applicable to a VERY small number of users. We're talking about Linux users, who have an AMD graphics chip, an AMD graphics chip that supports Mantle, in a system weak enough to need Mantle to get decent performance. A user has to meet all 4 of those criteria, simultaneously, for Mantle to actually matter.

That's an EXTREMELY small niche. Why would a developers want to maintain an entire additional codepath just for such a small subset of their customers?

I'm going to make the assumption that you own an Intel APU; not the Iris Pro. Can you run Battlefield 4, Civ 5, Thief on it without a video card and above 30fps? If you can then your statement is correct.

Drop the settings enough, and sure, those games will run at 30 FPS on Intel Integrated graphics.

That's hardly applicable to the situation with Mantle, though. Biggest advantage of Mantle is that it reduces CPU overhead, but an Intel chip running integrated graphics isn't CPU-limited. Even if it supported Mantle, it wouldn't help much.

If not then more work needs to be done to make that a reality. The AMD solution is called Mantle and it works very well.

I don't see how Mantle solves anything for the vast majority of users. It just seems to create extra work for developers to service an extremely tiny fraction of potential customers. I'd rather see them spend extra time optimizing the OpenGL port than maintaining Mantle.

 

[wonderfield]

I'm going to make the assumption that you own an Intel APU; not the Iris Pro.

I'm not sure I understand the purpose of 'Iris Pro' the qualifier here.

The AMD solution is called Mantle and it works very well.

"AMD solution" is a good term to use, yes. It is an AMD solution, and one that works quite well as an AMD solution — it's seemingly a good API. There are, however, solutions that can be called "solutions" without the "AMD" qualifier, at varying levels of availability on current consumer hardware and drivers.

Mantle works well, but so do GL vendor extensions: they provide a way to target vendor-specific hardware features through the context of an open platform, and vendor extensions have a path toward becoming ARB extensions depending upon their success (in which case, they become widely-adopted by multiple vendors and can be eventually mainlined).

AMD expressed intention to go down that path. Actual progress has been seemingly non-existent.

 

[Pieter3dnow]

Wondering why people are persisting that OpenGL extensions solve anything on the same level as Mantle does?

The other argument about making a niche product for a nice market does not fly, since the code for Mantle will be easy to port to DX12.

Mantle solves the problems that developers have with both DX and OpenGL even if they have to do some extra work they are seeing the value by making their "engine" run on Mantle which makes porting easier and works on multiple games using the same engine.

Developers asked for Mantle. If any developers could do the same in OpenGL they would have never bothered with Mantle.

 

[KazeoHin]

Which doesn't really help Mantle's case. Devs also have to do an OpenGL port that doesn't have access to those advantages, and the OpenGL port is more important, as it will be running on more systems than the Mantle renderer.

Once again, you're talking about maintaining an additional codepath, while also ultimately doing nothing to reduce any barriers to porting games to Linux.


Uh, right... Because me pointing out that you're literally making-up things I've never said is somehow equivalent to "shutting up because I've got nothing to argue with." Get a clue man...

"Lulz, told you so! You said it would never happen"
"None of my posts said it would never happen"
"Lulz, he has nothing to argue with!"

Do you realize how ridiculous that exchange looks?

I wasn't referencing this thread in particular. You do know when to 'bow out' when an argument makes AMD look a bit too good. I was trying to make that sound like a good thing..

Also my Linux/mantle reference: I generalise all Nvidia fanboys in one big clump. So you actually may not have made reference to AMD 'never porting Mantle to Linux', but one or more of the Nvidia die-hards did, and did so adamantly, using that one piece of speculation as the keystone in many arguments. Thus I just assumed you were among them. I honestly couldn't be bothered searching back for quotes at this time so you win. Sorry if I offended you with that generalisation.

But you have NOTHING on the stalwart denial that PRIME1 demonstrates. Take that as a bad thing or a good thing.

 

[cageymaru]

What are you saying "No" to, exactly? You don't think AMD's APU's are going to get faster over time? If that's the case, it's a dead-end platform... Sorry, don't believe that. I'm pretty sure we'll continue to see faster and faster APU's from AMD, which means API overhead matters less-and-less over time.

Also, Mantle isn't required to do game-design for an APU. I assure you, they're capable of running OpenGL without a problem, just like Intel and Nvidia GPUs.

You're also not addressing the issue of Mantle only being applicable to a VERY small number of users. We're talking about Linux users, who have an AMD graphics chip, an AMD graphics chip that supports Mantle, in a system weak enough to need Mantle to get decent performance. A user has to meet all 4 of those criteria, simultaneously, for Mantle to actually matter.

That's an EXTREMELY small niche. Why would a developers want to maintain an entire additional codepath just for such a small subset of their customers?


Drop the settings enough, and sure, those games will run at 30 FPS on Intel Integrated graphics.

That's hardly applicable to the situation with Mantle, though. Biggest advantage of Mantle is that it reduces CPU overhead, but an Intel chip running integrated graphics isn't CPU-limited. Even if it supported Mantle, it wouldn't help much.


I don't see how Mantle solves anything for the vast majority of users. It just seems to create extra work for developers to service an extremely tiny fraction of potential customers. I'd rather see them spend extra time optimizing the OpenGL port than maintaining Mantle.

Oh could you benchmark your I3 Intel chip without a video card installed and let us know how far above 30 fps in BF4 you can maintain. If you could do the same with Civ 5 on the largest map, please post that also. Just use the onchip graphics processor. Wouldn't take you long at all.

And Mantle is designed to allow ease of porting for GCN in Windows and Linux.. So I still don't see your argument unless you're saying that AMD doesn't sell enough APU's or video cards to matter in the big scheme of things. In that case then why do you care about a developer spending time with a tiny company such as AMD? They aren't even a blip on Intel or Nvidia's radar going by what you're saying.

Why are you so against a niche company making software that makes their products run faster?

 

[wonderfield]

Developers asked for Mantle. If any developers could do the same in OpenGL they would have never bothered with Mantle.

If AMD offered the extensions, they might. AMD not doing that precludes them from doing so.

I never claimed it would be "the same". I already described OpenGL and Mantle as being different APIs from a conceptual perspective. I said that problems can be solved using OpenGL. Naturally, vendor-specific extensions dictate...vendors making extensions. Which AMD isn't doing.

 

[The Mac]

it has some, but not to the extent that nvidia does. AMD has said in not so many words they dont like extensions.

There was a blog post from one of the code monkys at Valve that basically said OGL is infinitely more screwed up than dx, which is why very few people use it. He said Kronos could fix it by starting from scratch if they wanted to, but they probably wont.

 

[wonderfield]

AMD has said in not so many words they dont like extensions.

While simultaneously claiming that they're going to make them. Rather contradictory messaging, don't you think?

He said Kronos could fix it by starting from scratch if they wanted to, but they probably wont.

They've already effectively broken clean a number of times. They will continue to do so so long as they keep running things the same way they have been.

 

[Unknown-One]

Wondering why people are persisting that OpenGL extensions solve anything on the same level as Mantle does?

I never said they did, but they don't need to.

The fact that OpenGL runs on all three vendor's hardware is enough to make it the primary renderer deserving of a dev's attention.

The other argument about making a niche product for a niche market does not fly, since the code for Mantle will be easy to port to DX12.

I don't see how ease-of-porting has any bearing on the argument that Mantle only helps a small niche of customers. Especially since it doesn't make a game any easier to port (unless the dev totally skips implementing an OpenGL fallback, which they won't do)

All points from my previous post still apply:
"You're also not addressing the issue of Mantle only being applicable to a VERY small number of users. We're talking about Linux users, who have an AMD graphics chip, an AMD graphics chip that supports Mantle, in a system weak enough to need Mantle to get decent performance. A user has to meet all 4 of those criteria, simultaneously, for Mantle to actually matter."

My question still stands:
"That's an EXTREMELY small niche. Why would a developers want to maintain an entire additional codepath just for such a small subset of their customers?"

Mantle solves the problems that developers have with both DX and OpenGL

Which doesn't matter, because very few customers need the additional performance on their machines (and of those that do, even fewer have GPU's that support Mantle). If you want to port a game to Linux and have it run on the majority of systems, you have to do an OpenGL port and it has to work right.

That leaves Mantle as a secondary (and fairly optional) thing to support.

even if they have to do some extra work they are seeing the value by making their "engine" run on Mantle which makes porting easier

Mantle doesn't actually make porting any easier. The game must still be ported to OpenGL so it can run on all systems, so Mantle doesn't remove that hurdle.

Instead, supporting Mantle becomes an ADDITION to the existing workflow of porting.

Developers asked for Mantle. If any developers could do the same in OpenGL they would have never bothered with Mantle.

Doesn't matter, they'll have to make OpenGL work anyway if they want their games to actually run on more than just a small number of recent AMD GPUs.

 

[Unknown-One]

Oh could you benchmark your I3 Intel chip without a video card installed and let us know how far above 30 fps in BF4 you can maintain. If you could do the same with Civ 5 on the largest map, please post that also. Just use the onchip graphics processor. Wouldn't take you long at all.

Again, what does this have to do with Mantle?

Biggest advantage of Mantle is a reduction in CPU overhead... which means you need to be CPU limited to see any real benefits. Running modern games on intel integrated graphics on a Core i3 or Core i5 = severely GPU-limited, which a Mantle-like API doesn't help.

All such a test would show is that most Intel integrated graphics chips are slow, which we already knew. No low-level API is going to help that, but games that scale properly sure will.

Edit: Also, Battlefield 4 scales REALLY well to low-end GPUs. Here's someone running it, at 1080p, on a Core i7 4770k (that has an intel HD4600 on-board, which isn't iris pro) https://www.youtube.com/watch?v=ldrjn3mEQOc

And Mantle is designed to allow ease of porting for GCN in Windows and Linux.. So I still don't see your argument unless you're saying that AMD doesn't sell enough APU's or video cards to matter in the big scheme of things.

Re-read what I said. How many customers fit all 4 criteria I outlined simultaneously? Not many.

And, once again, Mantle doesn't make porting a game any easier. The developer still has to port to OpenGL (just like they always have) if they want to release a game on Linux. In-so-doing, that also means they support almost all AMD cards by-default (so why bother implementing Mantle? For the tiny cross-section of users that actually need it AND have hardware that can run it?)

Why are you so against a niche company making software that makes their products run faster?

Huh? Never said that. I'm against a developer expending resources to support adding an ENTIRE additional renderer to their game that helps almost nobody. Making sure the OpenGL renderer is solid should be top-priority, since that's what most people will use.

If you'd read the thread, I said that the extra performance Mantle affords to the hardware that supports it is great and all, but that doesn't mean it makes sense for a developer to bother implementing it.

Just think about it for a second, here's what happens when porting to OpenGL (spit-balling the percentage, but you'll get the idea):
- The game supports AMD, Nvidia, and Intel graphics.
- The game runs acceptably for almost all customers.
- The bottom 0.5% of customers can still run the game, but it might be too slow for them

And when we add an additional 4-to-6 weeks of dev time to implement Mantle, in addition tot he already-developed OpenGL renderer?:
- The game still supports AMD, Nvidia, and Intel graphics.
- The game still runs acceptably for almost all customers.
- The bottom 0.4% of customers can still run the game, but it might be too slow for them

How does implementing Mantle make sense? Even when implemented, it doesn't even help everyone with a slow machine (just those with a very-recently-purchased slow machine).

 

[InternationalHat]

You realize you've written an essay about a graphics API that really seems to have no impact on you.

 

[cageymaru]

You realize you've written an essay about a graphics API that really seems to have no impact on you.

Agreed. I could show him Apples in the form of an AMD APU running BF4 single player without Mantle or an onboard graphics solution to back it up. Or I can show him Apples in the form of AMD APU w/o a video card under Mantle in BF4 multiplayer. He'll then link me Oranges showing a i7-4770K with a separate onboard solution doing the same. The whole discussion was about SteamOS / Linux solutions and whether Mantle should exist as an alternative rendering path to pure OpenGL. It seems to me the the Intel solutions need some type of to the metal rendering path such as *cough* Mantle so they too can run the game without the use of extra graphics solutions.

You also have to take into account that we're talking about SteamOS SteamBox designs that literally fit in the palm of your hand. The more junk you can leave out of the design the faster you can shrink it to fit within a Smart TV. Since one solution requires an extra video solution to compete then it's obvious that these hardware are dissimilar enough that their developers would desire a different rendering path to make them shine. I hear that OpenGL 5.0 promises to bring it closer to Mantle type speed, but who knows as it's not already out yet.

I still don't see where tacking on 1 month of extra development time to make your software shines for people that want a SteamBox that is literally small enough to fit in the palm of your hand is so bad. You've spent 5 - 7 years developing an engine and game. How is another 30 days of development time going to hurt you when the tools are already integrated into your game engine? Look at BF4 Hardline from Visceral. It already has a working stable Mantle rendering path within the game less than 6 months after Mantle launched for the very first time. Same thing will happen for all CryEngine games and other gaming engines that integrate Mantle into the game engine. In the end, the tools will be there for the developer to choose to use or ignore. And having choices in software development seems like a good thing to me.

 

[maboblivion]

be careful of heat strokes in summer. cover your head, drink plenty of fluids

 

[Unknown-One]

Agreed. I could show him Apples in the form of an AMD APU running BF4 single player without Mantle or an onboard graphics solution to back it up. Or I can show him Apples in the form of AMD APU w/o a video card under Mantle in BF4 multiplayer. He'll then link me Oranges showing a i7-4770K with a separate onboard solution doing the same.

Not seeing the issue here. It's integrated graphics vs. integrated graphics. They can both run the game without Mantle. You just further proved my point that it's unnecessary to devote resources towards it (especially when you have a much-more-important OpenGL port to look after)

And I don't know what you mean by a "separate onboard solution." The i7 4770k is a single die which includes graphics, just like AMD's APUs. This was an Apples-to-Apples comparison.

It seems to me the the Intel solutions need some type of to the metal rendering path such as *cough* Mantle so they too can run the game without the use of extra graphics solutions.

What are you talking about? I posted a video showing intel integrated graphics running Battlefield 4 without any form of "metal rendering" solution.

Clearly intel doesn't *need* a new API since the 4770k's integrated graphics was able to handle the game without a problem.
Clearly intel doesn't need extra graphics solutions, because the on-die one worked just fine.

You also have to take into account that we're talking about SteamOS SteamBox designs that literally fit in the palm of your hand. The more junk you can leave out of the design the faster you can shrink it to fit within a Smart TV.

And there are Intel systems the same size. Your point?

Since one solution requires an extra video solution to compete

*Buzzer* Nope. There was no discreet graphics card in-play in the video I posted. The HD 4600 is built into the Core i7 4770K.

then it's obvious that these hardware are dissimilar enough that their developers would desire a different rendering path to make them shine.

You're making no sense. What's dissimilar about an on-die GPU vs. and on-die GPU, where both vendors have solutions that video into PC's that fit into the palm of your hand?

Why would the developers desire an alternate rendering path when even Intel integrated graphics can run games like Battlefield 4 at 1080p without a low-level API?

I still don't see where tacking on 1 month of extra development time to make your software shines for people that want a SteamBox that is literally small enough to fit in the palm of your hand is so bad.

Me either, so they should spend that extra month polishing their OpenGL implementation to make sure it scales properly for those machines (some of which wont support Mantle either, in spite of their form factor).