Talk:Mantle (API)

Mantle API Other Vendor Support "Disadvantage"
Mantle API is a low-level API which means it is an API providing communication with the hardware, the driver alongside it a light-weight, which means it has basically no game-specific code as the DX or OGL drivers do.

This is not comparable to DirectX or OpenGL at all. It potentially has no use for nvidia or intel. They have their own architectures and would need to develop their own low level driver for the use of an low level API, they may be able to use the Mantle API as a starting point for their own, but without cooperation there would not be a standard established, The Mantle API to work would require the API to have code for all of the vendors involved.

This is not a disadvantage of the Mantle API it self, the saturation of the APIs in the industry is a potential disadvantage of the developers and customers.

Therefore, to AMD, there is no disadvantage if other vendors do not support this API, actually it is an advantage.

The real disadvantage of the mantle API is: Architecture Specific, Requires more development time, Requires more developer skill, Stability and Performance depends on game developer skill and effort.

However the architecture specific will not be added, it is currently only supported on GCN; but the point of the API is to support on other architectures for future Radeons, however older mantle games will work under the newer versions of mantle, while the driver would have to be replaced. The details are yet to be seen so this cannot be decided now.

Then, low-level is not a specific feature that needs to be software developed, any vendor can unlock their proprietary drivers and API for developers to assume full control of the hardware at any time they wish.

Finally, I have added some of the advantages that "comparable to console environment" that are direct result of the use of low-level hardware access compared to DX and OGL Xowets (talk) 00:17, 11 October 2013 (UTC)


 * This is good information (and is congruent with my knowledge of the subject). I (as a graphics card enthusiast) really think that this should be added to the article in some shape or form - people shouldn't have to go to the talk page to find very pertinent descriptions of the topic like this (in my opinion, the crucial other half of the 'advantages' section). I understand that it might be a tiny bit difficult to find a good source because what company is going to talk about its own disadvantages, but surely some leeway for rational deduction exists. (PS: My first edit. No doubt I did something wrong, so my apologies.) --110.175.244.197 (talk) 00:14, 5 January 2014 (UTC)

Disadvantages - citation issues
People keep removing the disadvantages section, this is a new thing and ofcourse there aren't some interview yet with a developer that would explain this.

I personally don't in cases like these, don't give a rats butt about wikipedia's source rules, the information is known now, why suppress it, you guys should revert back to disadvantages with citation needed because this is a stub and I think it should have laxer rules on stuff like this, unlike the bigger and more established articles.

I will include the disadvantages back, but comment them out because I am very positive they will be mentioned by 3rd parties sooner rather than later.

For those people who wish to know what this API does, that's what are the disadvantages, but to some developers and consumers are worth it (acceptable), there is no reason for me to make this information up. The people who originally removed it don't know if it's accurate or not, because they commented it as "guesswork". I wouldn't be putting this up without citations, if I didn't know for sure. Xowets (talk) 09:17, 15 October 2013 (UTC)


 * Everything on wiki can be cited. You can cite the person you talked to, the video game level you played, the book you read. But everyone creates a cite for a url it is overdone on wiki. So in short wherever you got your quote from you can cite it, everyone here doesn't know where it came from or if it even true.Cky2250 (talk) 13:14, 15 October 2013 (UTC)


 * Also try to not list things, start writing this like a history section on any other wiki page. You can mention advantages and disadvantages without listing them off.Cky2250 (talk) 13:18, 15 October 2013 (UTC)


 * Well I'm not english so I might have coherency and fluidity problems making complex explanations for formal use so I use to avoid that; but there's another problem and I need to address this right now before people start taking that out of context.
 * There is a new article out where AMD explains some of the questions, and let me point out AMD is not the the best to answer these, you need to get details like that from actual developers.
 * AMD's "exclusivity" explanation is here http://pastebin.com/6X2Ai0eH


 * In that, they claim "improves time to market;" and "reduces development costs," - Let me explain because they mean something else, it is not the context you think it is.


 * They said these to things in relation to the coincidental console connection (Radeons on XOne and PS4), when developers would be able to simply copy paste their sophisticated hand-made optimization code from the console version to the PC version of a game, because the API supports this transfer, so it will be faster, which will mean they won't need to have more developers and separate them on each version in terms of GPU optimization, it will also cost less because they don't need to work extra months, this is a big deal for developers, but this has however NOTHING to do with the API it self, this is only a possibility that has opened it self for this SPECIFIC Architecture, if Nvidia creates their own equivalent API they will not have thes advantages, because there's no nvidia hardware in the Consoles.
 * Therefore, these 2 things are irrelevant for PC exclusives, only for multiplatforms. And this will all be over when AMD has a new architecture for PC, this transfer won't be possible anymore.


 * So don't put these 2 things into Advantages section, this is not a Mantle feature, this is only possible because consoles have the same architecture for now, this will hold only temporairly until new architecture for PC arrives and this won't work anymore unless AMD makes some kind of specific program to convert or at least partially transfer the code, but this is a guess, a new architecture would need work from scratch, converting and adapting won't truly be programming to the metal.


 * Furthermore, the claim about "improves time to market" which means faster release date is also a bit inaccurate, the developer is not going to leave out Nvidia support and other APIs, so it won't matter if the Mantle code is complete, they will release the game when they have all the other APIs complete for the same platform, ofcourse technically they could release the game only for AMD cards at first, but this is highly unlikely. Xowets (talk) 00:25, 16 October 2013 (UTC)

and what's with the page notice? i expect more details will come out once Mantle is implemented, but as of now, AMD only announced it, what, a month ago? not that stuff shouldn't be cited, but I don't see a particular problem on this page, no more so than any other, anyhow. Aunva7 (talk) 18:23, 20 November 2013 (UTC)

other games to support
Sniper elite v3 to support mantle by Rebellion

http://www.legitreviews.com/rebellion-announces-will-using-amd-mantle-games_129404

Majinsnake (talk) 08:40, 23 December 2013 (UTC)

IMPORTANT: Missing Citation!
The livestream for AMD has been permantanly deleted according to the link. I'm not going to remove it, because im a brand-new user and im not sure if i should or not haha. Someone, please chime in here!.

Note, i havent seen the cite prior to it's deletion, so ive no idea what it had contained!

Sometime editor always a lurker (talk) 12:27, 29 January 2014 (UTC)

DSA, Mantle, and citations
A few days ago, I removed the claim that "The most significant functionality of Mantle—the low overhead state changes—has subsequently been made available (in multi-vendor form) in OpenGL 4.5 as Direct State Access". The citations for the claim are this Khronos PDF file and this article from ArsTechnica. It was subsequently reinstated by User:Carewolf.

I submit that neither of these citations supports the claim that DSA is equivalent to Mantle's low state change overhead. The PDF in particular was written before DSA entered core. And while DSA has been an EXT extension for over half a decade, the PDF never talks about it. So it is irrelevant to that statement.

As for the ArsTechnica article, the article's title is "OpenGL 4.5 released—with one of Direct3D’s best features". This is in reference to DSA, and it suggests that DSA is in some way similar to Direct3D 12. But that's not the case, if you actually read the article. For example, it clearly states that "...compared to Microsoft's Direct3D, which has supported DSA for many years." What the title was referring to was not specifically D3D 12, but most versions of D3D.

Furthermore, the ArsTechnica article never once makes the claim that DSA improves "overhead state changes". Indeed, the part of the article that talks about DSA only talks about it in reference to how users interact with the API. They talk about the advantage with respect to middleware, but not to the overhead of state changes.

In fact, the only part of the ArsTechnica article that talks about Mantle is in a paragraph early on, and in the bottom section. And both of those are compared to Khronos's new, low-level API still in development. At no time does the ArsTechnica article compare DSA to Mantle.

So there is no foundation for this claim. So I see no problem in removing it. Korval (talk) 15:44, 19 January 2015 (UTC)


 * I don't believe DSA is equivalent of Mantle, but it is part of a set of OpenGL features that when used together can reduce or eliminate most API overhead (batched, indirect and bindless). I mainly reinstated it because I believed the pdf https://www.khronos.org/assets/uploads/developers/library/2014-gdc/Khronos-OpenGL-Efficiency-GDC-Mar14.pdf was addressing this. Neither of the references are the best, but several experts such as John Carmack has stated modern OpenGL usage (4.4 at the time with some extensions) are similar to Mantle http://www.dsogaming.com/news/john-carmack-nvidias-opengl-extensions-rival-mantle-8gb-on-consoles-not-a-significant-change/ see also http://linustechtips.com/main/topic/117504-modern-opengl-vs-mantle-from-nvidias-talk-at-steam-dev-days/ Carewolf (talk) 10:47, 20 January 2015 (UTC)


 * But DSA is not mentioned in any of those sources. Carmack speaks specifically of NVIDIA's OpenGL extensions, which DSA never was (even as an EXT extension, it is credited to many non-NVIDIA people). I'm guessing that, since it was 2013, Carmack was talking about NV_bindless_texture and NVIDIA's other bindless stuff; since then, bindless_texture became an ARB extension (though not the other bindless stuff).


 * The video is focused on the cost of state changes. But the video does not talk about the cost of calling the functions to change that state. It talks about the cost of binding new objects with new state (for the most part). DSA only affects the functions that do the state changing (though it does mean you don't have to bind simply to change an object's state).


 * Also, the forum you link to lists a number of extensions that contribute to this performance boost. Yet EXT_direct_state_access is conspicuously absent from that list. Korval (talk) 18:55, 20 January 2015 (UTC)


 * So again, there seems to be little evidence for DSA being a significant component of OpenGL achieving Mantle-like performance. I'm not necessarily saying that DSA isn't an important part of achieving Mantle-like performance on OpenGL. I'm saying that nothing you've cited states this to be the case. Korval (talk) 18:53, 20 January 2015 (UTC)


 * I am not defending DSA == Mantle here. DSA is only one part of the bindless API, but is not the central part of this modern OpenGL usage. All I am saying is that OpenGL has extensions that tries to achieve the same as Mantle.Carewolf (talk) 10:12, 21 January 2015 (UTC)


 * The statement in question is not whether DSA is a part of "the bindless API". Nor does the statement in question state that OpenGL has extensions that try to achieve the same effects as Mantle. The statement in question states: "The most significant functionality of Mantle—the low overhead state changes—has subsequently been made available (in multi-vendor form) in OpenGL 4.5 as Direct State Access." That statement is not supported by any of the references you have cited. Korval (talk) 07:41, 22 January 2015 (UTC)
 * As I have said multiple time. I have no problem with you correcting or removing that part. What I reverted was a removal of a much bigger paragraph that included a lot more than DSA, and also the AZDO reference. I am defending the reversion of your deletion. I don't care what kind of bad excuse you make up for your deletion, or if one small part of what you deleted is inaccurate. You deletion was WRONG. If you want to improve the paragraph or remove the part about DSA being the most important part. Then remove that, or correct that, but please stick to that. Whenever you remove more than is justified in the summary you may end up with revert. Carewolf (talk) 22:26, 22 January 2015 (UTC)

Wrong claims for Battlefield 4
Under Benchmarks, there is the following claim:

"Up to 45% faster than Direct3D in Battlefield 4"

According to the following link, which is listed as a reference for the claim above

http://www.pcgamesn.com/battlefield/amd-claim-performance-boosts-45-using-mantle-over-directx

" A couple of important caveats: the 45% figure comes not from Battlefield, but Oxide Games’ Starswarm demo - a 64-bit engine test first shown in December that attempts to fill space with thousands of ships on-screen at once. And it doesn’t apply to just any generic AMD card - rather the A8-7600, a new APU from the Kaveri line the manufacturer is showing at CES. A very expensive AMD card. "

The misleading figure originates from the misleading cover photo, combining both Battlefield 4 artwork and the "45% faster" claim in the same slide. — Preceding unsigned comment added by Wessamfathi (talk • contribs) 14:31, 27 November 2015 (UTC)
 * I altered the wording to remove the 45% number. Sizeofint (talk) 04:34, 28 November 2015 (UTC)
 * sounds good, but that source is dubious at best. i mean, for one, they called the a8 an graphics card, and for another, they called it "very expensive". reality is, it's not even the top-line APU. how reliable can that source really be? in any case, the 45% is a very optimistic figure, probably created by the marketing team. -- Aunva6talk - contribs 06:33, 28 November 2015 (UTC)

External links modified
Hello fellow Wikipedians,

I have just added archive links to 1 one external link on Mantle (API). Please take a moment to review my edit. If necessary, add after the link to keep me from modifying it. Alternatively, you can add to keep me off the page altogether. I made the following changes:
 * Added archive http://web.archive.org/web/20130927180942/http://www.livestream.com/amdlivestream?t=637320 to http://www.livestream.com/amdlivestream?t=637320

When you have finished reviewing my changes, please set the checked parameter below to true to let others know.

Cheers.—cyberbot II  Talk to my owner :Online 16:06, 26 February 2016 (UTC)

No longer actively developed?
Amd's official Mantle page, AMD's Revolutionary Mantle Graphics API, "In the News" section, has no news beyond 2014, the year Mantle was announced. I assume this means it is no longer actively developed by AMD? Is this enough evidence to mention in the article, as confirmation that AMD has switched their attention to Vulcan, to which they contributed the Mantle code? ToolmakerSteve (talk) 19:57, 4 October 2017 (UTC)