Back to Article

  • Gloomy - Wednesday, February 26, 2014 - link

    But the people with Nvidia cards told me Mantle was unnecessary and pointless? Surely this is unnecessary also? Reply
  • shadowofthesun - Wednesday, February 26, 2014 - link

    I'm pretty sure those people (who certainly are not restricted to Nvidia users) were complaining about the addition of yet another graphics API, and this article at least seems to indicate that those people were right and an entirely new API is not required.

    I'm pretty sure the people complaining about having low level access only exist in your imagination.
  • Kevin G - Wednesday, February 26, 2014 - link

    DirectX: Evolving Microsoft's Graphics Platform sounds like a session where they're going to re-engineer DirectX to become lower level. Note the last time MS did this (DX9 -> DX10) the driver model radically changed with the massive API change. The driver roll out with Vista was rocky to put it politely.

    I'd be a bit skeptical of the Direct3D Future session. Much was said about DX10 reducing overhead as Vista previews proclaimed. As things turned out, there never was a massive performance leap with DX10 (though image quality did improve).

    The OpenGL session is interesting as diving down to the metal is possible with various OpenGL extensions. The interesting development here would doing so through ARB extensions. I'm quite curious as to what comes out of this session.
  • Gigaplex - Wednesday, February 26, 2014 - link

    "As things turned out, there never was a massive performance leap with DX10 (though image quality did improve)."

    Quite the opposite. At least for the first couple of rounds of titles to switch to DX10, performance dropped and image quality didn't change much at all. I believe Crysis had a Very High graphics setting that only worked in DX10 mode, but there were config file hacks that let you run at Very High on DX9 with much higher framerates and it didn't look all that different.
  • FaaR - Thursday, February 27, 2014 - link

    You're compairing apples and oranges. DX9, DX10 = graphics APIs, Crysis = computer game. You can't draw conclusions about the merits of a graphic API by looking at one single piece of software using that API. Reply
  • chizow - Wednesday, February 26, 2014 - link

    Yep, exactly. Reply
  • Frenetic Pony - Thursday, February 27, 2014 - link

    Mostly, Mantle was a kick in the ass for Krohnos and Miscrosoft. Finally get them up and do something like it, before Mantle becomes popular, an API controlled by not them. Reply
  • xdrol - Friday, February 28, 2014 - link

    This. Even Mantle in its own is pointless, its value is freaking huge in pulling MS and Khronos out of their comfort zone to start working. Reply
  • HisDivineOrder - Wednesday, February 26, 2014 - link

    People with nVidia cards told you Mantle was unnecessary and pointless because it was a third API that doesn't support any product except GCN-based ones. That means basically a smidgen of the APU market (Kaveri-based APU's from Jan or newer) or 7xxx series GPU's or later. Relatively speaking, that's a small percentage (GCN) of a small percentage (AMD) of a small market (PC gaming discrete GPU's).

    Meanwhile, the lion's share of desktop gaming is owned by Intel and nVidia right now. Neither vendor supports or even suggested they'll support Mantle.

    That's what has always made Mantle unnecessary. We would all have welcomed Mantle's improvements if AMD had bothered to implement them in OpenGL like nVidia did. Instead, they felt the need to throw out a new API in an effort to seize by force what they could not win by straight up competition with nVidia.

    They tried to rig the game in their favor by changing the playing field, but the problem is no one's going to show up on their field.

    Better for Intel and nVidia--once AMD went it alone with Mantle--to leave AMD hangin' in the wind, having to support three API's while nVidia and Intel buckle down and focus on only two.

    The improvements were great, especially for draw calls and CPU usage. The problem was the proprietary nature of Mantle. And it is proprietary as long as AMD keeps it locked down and NOT a standard, not brought forward as a standard, and not even having said they'd MAKE it a standard.

    They suggested they wouldn't be adverse to the idea, which is a far, far cry from actually saying they're making it an open standard.

    I suppose Mantle pisses other users off (including AMD users who don't have a GCN-based product like 6xxx series cards or earlier, Llano APU's, Trinity APU's, or Richland APU's) because it promises improvements that are entirely NOT reliant upon GCN, but ties those improvements directly to GCN regardless.

    I hoped DirectX and OpenGL would bring over the improvements that were relevant from Mantle without the baggage. It appears they are. That's great.
  • SleepyFE - Wednesday, February 26, 2014 - link

    How do you know that Mantle isn't a modified version of OpenGL. Maybe that's what the talk is about. They will finaly make it available. And Intel and Nvidia are jumping on board. Why else would AMD be there (Yes we have a better solution, but we decided to help our competitors beat it to the ground.)
  • przemo_li - Thursday, February 27, 2014 - link

    Where did You seen HLSL in OpenGL ? O.O

    No. Mantle is not OGL. OGL REQUIRE big bad ass GPU driver that handle all the traffic to the monolithic GPU. Mantle is about spliting that GPU into separate execution units, and trimming down that driver to be as thin as possible (game/app will have to do the job of GPU driver, but some already do it)
  • SleepyFE - Thursday, February 27, 2014 - link

    I said modified!!
    They may not have started from scratch.
  • chizow - Wednesday, February 26, 2014 - link

    Spot on. Reply
  • Sushisamurai - Thursday, February 27, 2014 - link

    Didn't they already cover this in a pipeline story? Nvidia and intel are able to access mantle (if not now then in future plans, don't recall exactly)... For an undisclosed fee. With the over guys hoping on the bandwagon, they'll compete directly with AMD's low level offerings and take a piece of that low level licence fee pie Reply
  • chizow - Thursday, February 27, 2014 - link

    Don't think there was a pipeline story about it, just a lot of noncommittal back and forth depending on whom you asked, and when.

    AMD said Mantle was proprietary and would remain so for the foreseeable future, but maybe long-term someone could write low level API for their own hardware (lol ya think?)
    “The plan is, long term, once we have developed Mantle into a state where it’s stable and in a state where it can be shared openly [we will make it available]. The long term plan is to share and create the spec and SDK and make it widely available. Our thinking is: there’s nothing that says that someone else could develop their own version of Mantle and mirror what we’ve done in how to access the lower levels of their own silicon. I think what it does is it forges the way, the easiest way,” explained Mr. Corpus."

    And Johan Andersson in one of his many AMD-funded speaking engagements had a slide saying anyone could theoretically support Mantle (again, write their own low-level API for their own hardware), trying to back away from the whole proprietary stigma that AMD has criticized repeatedly over the years.

    It's just lip service so that AMD didn't sound hypocritical, and to help improve media/PR image of Mantle and it's chance for adoption. Looks like it doesn't really matter now though, if given a choice between supporting low-level DX extensions or supporting Mantle, it's really going to be a no-brainer for IHVs (Nvidia/Intel) and ISVs (game devs), especially when the XB1 code should be a near seamless port after MS brings the XB1 DX11.x enhancements they mention to the PC.
  • chizow - Wednesday, February 26, 2014 - link

    Mantle is pointless, and now Microsoft has just reaffirmed this reality. Another vendor-specific API on a platform with already 2 established APIs that support all vendors is clearly unnecessary and pointless. Mantle will end up a footnote in PC gaming history, nothing more. Maybe it had a hand in forcing Microsoft to act, maybe not, but in the end it will fall by the wayside. Reply
  • przemo_li - Thursday, February 27, 2014 - link

    4 months ago, MS asked DIRECTLY about future of DX could only talk about future of TOOLS surrounding DX.

    Progress is astounding if You ask me.
  • chizow - Thursday, February 27, 2014 - link

    Huh? No, they explicitly said they were planning to bring the low level, lightweight DX runtime from the XB1 to the PC:
    "We’re also working with our ISV and IHV partners on future efforts, including bringing the lightweight runtime and tooling capabilities of the Xbox One Direct3D implementation to Windows, and identifying the next generation of advanced 3D graphics technologies."
  • przemo_li - Friday, February 28, 2014 - link

    This is after mantle.

    I'm talking about a bit earlier period. When AMD (as usual) was making fuss about lack of innovation in DX (even OGL is super set of DX11.1 now). And MS had no comments.

    But if You read carefully that text, than again You will see how little change there is for API itself. MS would not even force Tiled Resources and other such stuff onto Nvidia/AMD/Intel and just left it as "may support" bits.

    DX (And OpenGL to some extend, though, Khronos have better way to announce what should be in the next gen hw), lack clear vision of what should land in next gen hw, which would be supported by all GPUs. DX12 instead of DX11.3...
    (And MS do have incentive to stick to the DX11 for a while now. XboxOne will not support DX12...)
  • encia - Thursday, March 06, 2014 - link

    You missed the word CONSOLE in the following Microsoft's statement

    "However, you asked us to do more. You asked us to bring you even closer to the metal and to do so on an unparalleled assortment of hardware. You also asked us for better tools so that you can squeeze every last drop of performance out of your PC, tablet, phone and CONSOLE."

    Microsoft's current game CONSOLE is Xbox One and it runs on AMD GCN based GPU.
  • SirMaster - Thursday, February 27, 2014 - link

    No... Mantle is pointless because of this.

    The people who said Mantle is pointless figured that just from the fact that Mantle was about to exist meant that this would naturally follow.

    At least that what I've always figured. It's basic competition.
  • Krysto - Thursday, February 27, 2014 - link

    This is EXACTLY why I supported Mantle. Because I knew that if it becomes popular it forces the others to respond and either join AMD's Mantle project and all support it, or do something about their more popular but bloated APIs and make them more inline with Mantle.

    Win-win for gamers.
  • nevdawg - Wednesday, February 26, 2014 - link

    Good on AMD for igniting innovation in the industry Reply
  • Wreckage - Wednesday, February 26, 2014 - link

    You may want to google "3DFX" and "GLIDE" Reply
  • Mathos - Wednesday, February 26, 2014 - link

    And you sir may want to google Nvapi.

    Nvidia has been doing the same thing for years, that AMD is starting now.

    And 3dfx didn't die because of of Glide. 3dfx died, because they got greedy, and decided it was better to buy out one of their board manufacturers and go single source, rather than selling their chips to 2nd party board makers like everyone else. Which caused prices to skyrocket, and innovation in aftermarket pcb designs to die out for their products. Glide support died, because 3dfx ceased to exist, and eventually was bought out by Nvidia.
  • The_Assimilator - Wednesday, February 26, 2014 - link

    The difference is that nVidia has never touted NVAPI as the holy grail of performance improvements. Reply
  • Flunk - Wednesday, February 26, 2014 - link

    Or Cuda. Reply
  • chizow - Wednesday, February 26, 2014 - link

    CUDA is completely different, it was an API designed for compute because there was no compute API in place. Can't fault Nvidia there for innovating and creating an industry and ecosystem from nothing for lack of better tools. Reply
  • Krysto - Friday, February 28, 2014 - link

    I guess. It wouldn't kill them to join the growing HSA Foundation, now, though. If I were them I'd do it only to single out Intel, before they bring their MIC stuff into mainstream processors and start marketing the hell out of it.

    HSA also sounds pretty damn awesome, so it's not like they have to dump CUDA for a much inferior "open" spec. Besides, they could probably continue to support CUDA on top of HSA anyway.
  • TheJian - Sunday, March 02, 2014 - link

    Cuda 6 has the same type of stuff in software, and is waiting on maxwell to use it. It would be unwise to join HSA while you have a better option in Cuda that owns 90% share already. Kill HSA with Cuda is the smart business solution (whether I like that or not doesn't matter, talking WISE business moves here). Back down when forced when you have so much invested in cuda.

    Like MS with directX, you don't help OpenGL by supporting it in Xbox1 until forced with so much tied up in DirectX. If mobile getting desktop gpu stuff (k1, m1, v1 etc) causes a ton of stuff to go OpenGL (which PS4 uses, not DX, and so far sold ~2x more than xbox1) then MS will be forced to go there, but until then why help them kill your platform? I think MS loses this war (just due to OpenGL's portability to everything else, and Intel starting to move into android bigger) but I understand it was their best "business" move. If you want a game to run everywhere easily you avoid DX which is WINTEL x86 only.
  • BrightCandle - Wednesday, February 26, 2014 - link

    You couldn't be more wrong. NVAPI doesn't do gaming interfacing to the card at all. You should consider it more like an API that contains all the things the Nvidia control panel contains, like the ability to override antialiasing or resolution settings. It does not include rendering commands. Reply
  • chizow - Wednesday, February 26, 2014 - link

    Overall agree that NVAPI isn't as obtrusive as Mantle, but it does actually include rendering commands. You can write any low level shader code you like in ASM via NVAPI, it's just not exposed to the application, but you can get to low level ASM code by hooking in through the driver. This is how Nvidia's 3D Vision works, and how brilliant tools like HeliX mod literally change shader code to correct Stereo 3D images. Reply
  • ppi - Wednesday, February 26, 2014 - link

    Keep in mind 3dfx had no response to nVidia products after ~Voodoo3. Especially with Voodoo5 they were late to market, and as I recall the benchmarks, a less expensive GeForce2 did easily beat it and GF3 was final nail in the coffin. Reply
  • Mathos - Thursday, February 27, 2014 - link

    Thats because the Voodoo 3 and 5 were done in the time frame after 3DFX had bought out STB. In my time as a gamer, and PC enthusiast, I've owned a Diamond card based on the S3 Virge, as well as one based on the Rendition Verite chipset. I owned a Diamond Monster 3D during the Voodoo 1 days. I owned an STB Black Magic 2 during the Voodoo 2 days. I owned a Guillomat (spelling?) Voodoo Banshee card. I've owned Nvidia Riva TNT, and TNT2 cards. And I even had a Voodoo3 for some time, as well as access to a voodoo5, they were actually faster, but they were limited to 16bit color, which was their major flaw. A Geforce 3 Ti 450 I can't remember the brand of. I had a Geforce 4 Ti 4600 for a time. And then Had a Geforce FX 5600 Ultra for some time. Changed to ATI/AMD after the HD3000 series, and more or less just upgrade every other generation depending on performance. Still may go Nvidia if the card they have performs best at the price point I can afford at the time.

    The use of Nvidia Nvapi started in 2002, used for games that were specifically designed to run better on Nvidia cards of the time, aka Geforce 3 and 4. Thats when the Nvidia the way it's meant to be played logo popped up. Guess what... the Xbox, which used an Nvidia geforce 3 based gpu, had launched just before that, in late 2001, November to be exact.... Coincidence anyone? Contrary to popular belief, it did stir up bad reactions from people back then, saying it was competing with OGL, and DX. Even though in reality, it works the same way mantle does, and it's proprietary and specific to nvidia's hardware. This is why NTWIMBP titles have always performaned better on Nvidia cards, than competing ATI cards even when ATI/AMD had clear performance advantage in none nvidia mtb games.

    Fast Forward to Today. AMD/ATI GPU's are currently in all. AMD is simply doing the same thing. But on the bright side, it looks like it was enough of a fire being lit under someones rears to start improving the main graphics API's. Mantle doesn't ignore DX, it uses code specific to AMD cards to improve or enhance performance while running DX. Which is the same thing Nvapi does.

    This is a direct quote from Nvidia's own website "NVAPI is NVIDIA's core software development kit that allows direct access to NVIDIA GPUs and drivers on all windows platforms. NVAPI provides support for categories of operations that range beyond the scope of those found in familiar graphics APIs such as DirectX and OpenGL."
  • chizow - Wednesday, February 26, 2014 - link

    NVAPI exposed some low level functions for Nvidia GPUs when DX and OpenGL functionality was limited or not fully supported, but it ALWAYS sat below the DX or OpenGL APIs. It never tried to usurp them, instead leaving the existing API hierarchy and support chain intact. Same cannot be said for Mantle. Surely you see the difference right?

    Just look at it from the other perspective, if Nvidia decided to develop and promote NVAPI as a low level API to replace existing HLSL, would you in favor of it? For -5 to 5% scaling in commonly used gaming scenarios? Why bother?
  • JlHADJOE - Thursday, February 27, 2014 - link

    I thought it was more because they got really complacent and fell so far behind in terms of features, and then eventually performance.

    Hell 3Dfx didn't even have true 16-bit color when Nvidia already had 32-bit, but in a bout of supreme arrogance 3Dfx basically went "16-bit is good enough for everybody". Well Q3A's alpha-blending certainly proved them wrong.
  • Creig - Wednesday, February 26, 2014 - link

    And you, Wreckage, may want to google "TROLL" and "OBSESSIVE BEHAVIOR". Reply
  • nathanddrews - Wednesday, February 26, 2014 - link

    I've generally got nothing against DX or OGL, but we're overdue for a shake-up of graphics APIs. Mantle has made clear that both bleeding edge (minimum frame rates) and mainstream (minimum and maximum frame rates) can benefit from more efficient API/driver combinations.

    P.S. The lockdown of DX features to only the newest versions of Windows also needs to go away. Also, I wouldn't be mad if OGL gained wider acceptance.
  • Wreckage - Wednesday, February 26, 2014 - link

    So Mantle is dead before it ever truly lived... OpenGL has had low level support for a while and DX looks to be catching up. Good news for everyone...except maybe AMD who have spent a lot of money on Mantle. Reply
  • The_Assimilator - Wednesday, February 26, 2014 - link

    And Dice and all the other chump game companies that have wasted time and money on adding Mantle support to their engines. Reply
  • DanNeely - Wednesday, February 26, 2014 - link

    Was the actual work done and paid for by Dice, etc; or did AMD send developer support people out to do the heavy lifting so that they'd have a big victory to point at when it launched? Reply
  • DigitalFreak - Wednesday, February 26, 2014 - link

    This. Reply
  • chizow - Wednesday, February 26, 2014 - link

    A bit of both, AMD said they spent 2 years developing Mantle and sent engineering resources on-site to help implement it (Nvidia does this too with PhysX and GameWorks). AMD went into unprecedented territory however, when it reportedly paid EA/DICE $5-8 million for exclusive BF4 rights and Mantle support/implementation. It should be no surprise Johan Andersson (formely DICE, now just Frostbite Team) has been a very vocal supporter of Mantle with him and his team providing direct development resources to support Mantle. He said it took about 2 months of man hours as of November, but that was before the 2 months of delays before launch and overtime after Mantle launched to simply fix the errors/bugs introduced by Mantle.

    I'm sure they envisioned it as a great differentiating factor for EA and DICE for all the EA games reportedly using their FB3 engine, but now it is looking more and more like a deal EA/DICE wish they could get out of.
  • inighthawki - Wednesday, February 26, 2014 - link

    OpenGL does not have "low level support." The closest it comes is the bindless resource extension from NVidia. Otherwise compared to say, Mantle, OpenGL is high level. Reply
  • chizow - Wednesday, February 26, 2014 - link

    AMD clarified Mantle recently too btw, it's not really low level either, it's just translated HLSL with lower level abstractions/extensions for specific hardware. Same thing can be done with the existing HLSL (DX and OpenGL) with more extensions and less abstraction for target hardware. OpenGL has already gone this route, as has MS with their DX11.x implementation for XB1. Expect more of this on the PC soon from what it sounds like. Just a matter of when, and if it is going to require a new DX designation or just DX11.x Reply
  • przemo_li - Thursday, February 27, 2014 - link

    Try composing command queues, and executing them on specific execution units on the GPU with OpenGL. Reply
  • chizow - Thursday, February 27, 2014 - link

    Already exists in DX, would just be up to the OpenGL ARB to implement additional lower level extensions for vendor specific hardware, same as anything else. Nvidia has already implemented numerous OpenGL extensions ahead of DX, like bindless textures for their own hardware, so if this is where the industry is going there's going to be more responsibility placed on the IHVs to support their hardware. Reply
  • chizow - Thursday, February 27, 2014 - link

    Oh, forgot, another good example of Nvidia implementing features in OpenGL ahead of DirectX would be tiled resources with their work using CUDA to decompress and address Rage's megatextures. Tiled Resources was just implemented in DX with DX 11.2. Reply
  • SleepyFE - Wednesday, February 26, 2014 - link

    Now that i got that out of the way: It was worth it. Even if Mantle was just a kick in the gluteus maximus for everyone to wake up.
  • saneblane - Wednesday, February 26, 2014 - link

    It's amazing how people are bashing Mantle but they assume that DX and GL are going to be different. News flash Numbnuts, these low level features you all are talking about are not going to be supported on your cards now, you're going to have to get New Cards to support them fully. These API are new, they have the name DX and GL but they are essentially new API and will have to break compatibility with cards today. You're cards are not going to "just" work Lmao. Reply
  • Margalus - Wednesday, February 26, 2014 - link

    maybe not, but at least we will have multiple choices of gpu vendors, whereas mantle only works with a few select amd gpu's Reply
  • chizow - Wednesday, February 26, 2014 - link

    lol saneblane, rough news cycle I know but try and keep it together.

    It will be different because Microsoft and Khronos bring credibility to the whole deal, along with the fact DX and OpenGL will support 100% of graphics hardware on the market. Unlike Mantle which only supported AMD, which only has 35% of the discrete market and 18% of the total market. With that little marketshare and AMD's driver problems and financial woes, it was easy to see why Mantle was destined to fail.
  • SleepyFE - Thursday, February 27, 2014 - link

    Looks like you weren't keeping up with Mantle news. AMD said there is no reason it could not work on other graphics cards like Nvidia's. The question is, will Nvidia provide support with their hardware once the api is out (right now it's still beta so complaning about bugs and limited support means you don't know what a public testing period is = beta). Reply
  • chizow - Thursday, February 27, 2014 - link

    You mean AMD said anyone could write their own low level API for their own hardware and call it Mantle? LOL yeah of course, for Nvidia it's called NVAPI, but why would they bother calling it Mantle when it's already called NVAPI?

    Don't drink the Kool-Aid, Mantle is proprietary, if it was open as AMD claimed we'd have source code or at the very least, an SDK and it wouldn't be tied to GCN as it was always designed to be.
  • Friendly0Fire - Wednesday, February 26, 2014 - link

    If the DX/GL updates require new cards (which isn't as clear as you make it up; whereas Mantle was an AMD-driven push to ultimately sell more of their GCN cards, DirectX and OpenGL are controlled by independent third parties who have zero interest in pushing new card sales), at least any new card would support it. With Mantle, only the AMD cards would have ever worked (fat chance of NVIDIA supporting that, let alone Intel). Reply
  • SleepyFE - Thursday, February 27, 2014 - link

    Right. When Mantle comes out it's on Nvidia to support it and they probably won't.

    As for not needing new cards: If you want hardware support you need new hardware. Software support solves nothing since they are trying to put less software abstraction between the card and the game.
  • jwcalla - Thursday, February 27, 2014 - link

    The OpenGL presentation is most likely this one:

    No need to wait until GDC.
  • beginner99 - Thursday, February 27, 2014 - link

    I don't see why this should male mantle obsolete. This new D3D will soonest be available with Windows 9. Considering how long it took to go from dx9 to dx10, that will be years till it is wide-spread. Mantle is available now and guess what it works in Win 7 and 8. No need to buy a new CPU and OS now if you get an AMD GPU. Reply
  • tuxRoller - Thursday, February 27, 2014 - link

    The consensus at Phoronix seems to be that it will be quite similar to an Nvidia talk given during Steam Dev Days.
  • dwade123 - Friday, February 28, 2014 - link

    RIP Mantle. You will not be missed. Reply
  • Kamus - Monday, March 03, 2014 - link

    It's hilarious to see the amount of trash talk towards mantle here.
    These are the facts:
    Mantle works as advertised, and it's available now.
    I'm sure in a year or two the new APIs will start to become relevant... but for now, Mantle is the only way to play BF4 with 64 players with out the frame rate being inconsistent.

    Yes, you can argue that BF4 is just one game, but for those of us that do play that one game a lot online, this has been a very welcome and noticeable performance boost.

    So all you haters out there, keep on hating, meanwhile those of us that are actually taking advantage of this free performance boost on BF4, have nothing to complain about.
  • remc86007 - Monday, March 03, 2014 - link

    I do agree with you, but I thought I'd mention that my two 660tis and 4770k never drop below 60fps even in the craziest of 64 player battles. Reply
  • Kamus - Monday, March 03, 2014 - link

    Forgot to add...

    To those people claiming that mantle will fail because no one will support it you are missing a huge little fact:

    Most "AAA" games (you know, the ones with huge budgets) are for consoles... and guess which consoles are using GCN?
    This is why mantle can be relevant for the foreseeable future.

Log in

Don't have an account? Sign up now