Ubisoft: AMD's Mantle API is a double-edged sword

[AnandThenMan]

Ass Creed = TWIMTBP right?

Say no more about this article.

Given that this Luc Poirier is saying anything at all positive about Mantle is telling IMO.

 

[3DVagabond]

All company make bogus claims to make hype but at the end both will be on neutral API called Dx.

Neutral as long as you are developing for Windows.

I have no idea why people want to continue to be anchored to Windows and M$ leveraging it against us for their own benefit.

 

[chimaxi83]

Oh hey, another AMD bash thread right before a launch. News at 11.

 

[3DVagabond]

Oh hey, another AMD bash thread right before a launch. News at 11.

Say it ain't so.

 

[sushiwarrior]

Well, devs who wanted them to make the API and also who are paid by AMD. Ubisoft has a deal with Nvidia atm, so its natural for them to show Mantle in a different light then what DICE says.

Doesn't really seem like a negative statement tbh.

EA may have paid AMD to have an exclusivity clause with mantle for the first few months... Kind of vice versa. Maybe there's a lot more titles with it, we won't know until the clause is over.

http://fudzilla.com/home/item/32771-the-real-reason-behind-apparent-lack-of-mantle-devs

 

[AnandThenMan]

No dev worth their salt is going to leave performance on the table if it really is as significant as Mantle could very well deliver. That's why I believe we are seeing statements such as this, and others saying "we will have to support it". At worst it will light a fire under Microsoft, and force them to stop with the nonsense of artificially limited new versions of DX to their latest OS to try and force adoption.

 

[tviceman]

And devs will continue to vastly underutilized pc hardware for years to come. Happy days

You do realize that AMD constitutes the VAST minority of hardware users among PC gamers right? It's kind of oxy-moronic for developers to target less-used and generally inferior hardware for "optimizations."

Just saying.

 

[SolMiester]

Hardware users or dGPU user?

 

[Silverforce11]

You do realize that AMD constitutes the VAST minority of hardware users among PC gamers right? It's kind of oxy-moronic for developers to target less-used and generally inferior hardware for "optimizations."

Just saying.

VAST minority? If you count only discrete, its what, around 1/3? If you count APUs, its bigger than NV. Gamers actually build AMD APU rigs to play games. They dont do it with Intel specifically.

 

[Dribble]

Do you as a dev split your time to develop 2 renderers - 1 DX for 90% of your market and a second slightly faster mantle one for 10% of your market, or do you just spend all of your time doing a better job of one renderer. Does having a mantle renderer sell more games then just giving everyone the DX one? These are the same devs that can't even be bothered to do a proper PC ui in their console ports - what do you think they will do?

The only way mantle makes it into PC games is if AMD pays for it like they did with BF4.

 

[flopper]

Do you as a dev split your time to develop 2 renderers - 1 DX for 90% of your market and a second slightly faster mantle one for 10% of your market, or do you just spend all of your time doing a better job of one renderer. Does having a mantle renderer sell more games then just giving everyone the DX one? These are the same devs that can't even be bothered to do a proper PC ui in their console ports - what do you think they will do?

The only way mantle makes it into PC games is if AMD pays for it like they did with BF4.

if Mantle brings with frostbite a change in the gaming landscape yes they will support it.
people buy the better deal they percive and if mantle adds that your off running or left behind.

 

[SiliconWars]

Do you as a dev split your time to develop 2 renderers - 1 DX for 90% of your market and a second slightly faster mantle one for 10% of your market, or do you just spend all of your time doing a better job of one renderer. Does having a mantle renderer sell more games then just giving everyone the DX one? These are the same devs that can't even be bothered to do a proper PC ui in their console ports - what do you think they will do?

The only way mantle makes it into PC games is if AMD pays for it like they did with BF4.

Or do you make a few small alterations to your console API code and launch the PC game at the same time? You know, instead of us waiting 6 months for GTA V on PC, why not have it now instead?

People just aren't paying attention. Mantle isn't ALL about the performance gain, it's of far more value to the devs as a time-to-market and ease-of-portability tool. We may see AMD Mantle versions of PC games to start with then the DirectX versions coming months later. When that starts to happen it's all over for Nvidia, they'll be irrelevant in a matter of months as the high end gamers are forced to move to AMD just to play games much earlier.

So what if AMD has to pay some of them to start with? Do you think they won't do it? If you understand what this means at the base level then AMD *must* do everything in their power, no matter the cost, to ensure that console games are instantly ported to PC and available on GCN. This is the killer feature of Mantle, not the performance optimisations.

 

[MTDEW]

Or do you make a few small alterations to your console API code and launch the PC game at the same time? You know, instead of us waiting 6 months for GTA V on PC, why not have it now instead?

People just aren't paying attention. Mantle isn't ALL about the performance gain, it's of far more value to the devs as a time-to-market and ease-of-portability tool. We may see AMD Mantle versions of PC games to start with then the DirectX versions coming months later. When that starts to happen it's all over for Nvidia, they'll be irrelevant in a matter of months as the high end gamers are forced to move to AMD just to play games much earlier.

So what if AMD has to pay some of them to start with? Do you think they won't do it? If you understand what this means at the base level then AMD *must* do everything in their power, no matter the cost, to ensure that console games are instantly ported to PC and available on GCN. This is the killer feature of Mantle, not the performance optimisations.

I can't see "Mantle only" versions of games being released earlier on pc with DX versions to follow later.
Even AMD paying for BF4 Mantle support isn't resulting in an early "Mantle only" version.
Game devs simply cannot afford to ignore such a large portion of the pc gaming market when porting games to pc by ignoring DirectX at games launch.

 

[NTMBK]

NVidia's pet developer doubts Mantle, AMD's pet developer champions Mantle. Did anyone expect anything different?

 

[sandorski]

Speculation is fun and I can't bite my tongue anymore...so here are my speculations about Mantle..yay

1) Offtopic from this thread, but part of another recent thread re: Mantle on XBox1 - Microsoft itself won't be supporting Mantle, but it will be supported by AMD. Only exception would be if MS decided to forbid it being used, which seems unlikely IMO.

2) Performance - Unknown, but likely very significant for some particular tasks. Possibly no difference for other tasks.

3) Implementation/Usage - I suspect it will be used more and more as the XBox 1 and PS4 get longer in the tooth and as games get more graphically demanding. Given the expected disadvantages of the XBox1, I suspect it will be utilized earlier on the XBox1 than on PS4.

4) Frostbite, BF4, and AMD - AMD may have Paid $ to have it showcased. However, I suspect the real motivation involved just needing a partner to co-develop it. A way to work through the Bugs and/or have direct feedback from those who will be Coding/Implementing it.

Time will tell what it is, what affect it has, who uses it, etc.

 

[itsmydamnation]

what the mantle doom brings completely ignore is that the very vast majority of development costs for games like AC is asset/environment creation, engine licensing etc.

if you use a middle-ware engine someone else is taking the majority mantel development effort. If your building your own you already have written and tuned 3-4 renders and another one is only a incremental effort/cost in terms of programming. That then becomes a tiny slither of total project cost. When you then factor in advertising and other promotion costs your mantle cost isn't even going to register.

if there performance benefits are there,
if you can get code reuse from the consoles ( coming up with the logic is the hard bit, rewritting for another language/syntax is simple)

then cost isn't even going to matter.

 

[MTDEW]

NVidia's pet developer doubts Mantle, AMD's pet developer champions Mantle. Did anyone expect anything different?

NOPE!
I think Mantle will be used as nothing more than to give AMD a performance boost in supported titles.
I'd be shocked to see any titles being Mantle exclusive even for a limited time.
AMD would need a lot more cash to make a AAA title Mantle exclusive for even a few months after a games release on pc.

 

[SiliconWars]

NOPE!
I think Mantle will be used as nothing more than to give AMD a performance boost in supported titles.
I'd be shocked to see any titles being Mantle exclusive even for a limited time.
AMD would need a lot more cash to make a AAA title Mantle exclusive for even a few months after a games release on pc.

This totally depends on how easy it is. If the console API's are very close to Mantle that it's a breeze to port games to the PC, then why not do it? Even if it's only for 40% of the PC gaming market originally - that's still a large market. As time goes by more people will switch to AMD and more benefit will come from it for the devs. Just about everybody except Microsoft wants to see the end of DirectX and this is the easiest way to bring it about. The devs want this for many reasons - if DirectX is simply replaced with Mantle then that's a huge win for them - as Nvidia becomes irrelevant there would no more need to waste resources on DirectX ports at all. It's really bad from an Nvidia point of view but do you think games devs care? EA sure don't, Ubisoft seem to be paying it some very close attention and they are supposed to be Nvidia's #1. They will go with what makes them most money, there's no room for sentiment here.

On the cash point, consider what AMD has to gain and lose. They clearly don't mind throwing cash at devs for exlusive features, so why wouldn't they continue to do it with Mantle? Any game that is optimised for Mantle can basically be thrown on the "forget" list in terms of driver updates or any other optimisations, saving them money in the long run.

If it's easier to do a quick Mantle port (it is), and AMD is giving them incentives to do so (they will), what dev is going to say no?

 

[Dribble]

if Mantle brings with frostbite a change in the gaming landscape yes they will support it.
people buy the better deal they percive and if mantle adds that your off running or left behind.

People buy games because they want the game, not because it runs a little faster/slower on their graphics card. Mantle will not sell PC games it will sell radeon graphics cards which is of no benefit to the game dev.

 

[SiliconWars]

People buy games because they want the game, not because it runs a little faster/slower on their graphics card. Mantle will not sell PC games it will sell radeon graphics cards which is of no benefit to the game dev.

Spending a few days porting to Mantle or a few months porting to DirectX, which is of most benefit to the game dev?

 

[ShintaiDK]

Spending a few days porting to Mantle or a few months porting to DirectX, which is of most benefit to the game dev?

More of the nonsense hype and factless comments.

Xbox One ports are close to copy/paste to PCs. Windows and DirectX all around.

 

[SiliconWars]

Xbox One ports are close to copy/paste to PCs. Windows and DirectX all around.

You missed an important part. Mantle.

 

[zlatan]

Xbox One ports are close to copy/paste to PCs. Windows and DirectX all around.

Impossible to do this with DX and even with Mantle. The Xbox One will use a very special low-level DirectX API, it will be a superset of DX11.2 functionality.
On the PC with DirectX/WDDM or OpenGL you won't get explicit control over the memory. This is a huge difference. Mantle is a solution for this, but a console code can't be copied.

The only thing that the developers can copy are the HLSL codes. But this is not recommended, because in DX11.x you don't have unlimited resources/samplers, and this API can't read/write all UAV formats. On the console APIs and with Mantle you don't have these limitations.

 

[JoeRambo]

Those who dream that Mantle can give 3x the performance in real world game are clearly clueless. ( in fact performance could improve 10x, but only if all your App is doing is draw calls ).

Remember, hardware is not getting any faster, same Gflops, same fillrate, same geometry perf...

What is changing is overhead in accessing that performance, but it is not like AMD/NV are leaving 67% of performance on the table and blowing rest in DX API.

So lets skim over an one thing where stuff like Mantle can help - amount of draw calls that dev can use to draw their stuff on screen. Simplified little background - to draw lets say a box, you need to prepare stuff like vertices and issue draw calls - literally call a Draw() with a set of geometry. Cool? But not simple, you have the consider the following:
1) Hardware is not sleeping, you need to keep it fed with things to render and issue stuff fast enough to keep it busy with work ( but hardware is bound by well - hardware resources).
2) Due to DirectX overhead, there is massive CPU overhead for each Draw call, we are talking about thousands to maybe teen thousands here depending on CPU budget.
3) In DX9 days it was completely ridiculous, i think draw call budgets were several hundred max per frame.


So devs need resort to different techniques, most common is to limit Draw calls by batching stuff, like submitting a ton of geometry. Clearly this approach is not ideal, since not everything is batchable (as in not everything is a grass blade that can be "instanced", richer scene requires richer geometry) due to game demands AND it can have noticeable overhead of its own ( additional buffering etc).

2nd approach available since DX11 is to use concurrent command lists, instead of limiting your dx work to one thread, you can do it from multiple ones. The problem is of course drivers, only NV has got it right at the moment (since it increases driver complexity). Notice that CPU usage is still there, and probably even increased, since driver is doing complex sinchronization, enqueing and buffering behind scenes, but you are no longer bottlenecked by the amount of draw calls you can issue on one CPU. = smoother run, higher min FPS on underpowered hardware (ironically AMDs lack of concurrent command lists is hurting their disastrous single threaded performance CPUs the most).

Mantle can supposedly bypass this CPU overhead for a lot of things like draw calls, resource setup etc, so you can cram console levels of draw calls without being bottlenecked by CPU and reap other benefits already outlined by AMD and other proponents. But it won't give 3x performance like someone in this thread claimed, i won't speculate how much it can give, but my guess it will be PR blitz like stuff - a bit better performance and better image quality to avoid apples to apples comparisons.

 

[GaiaHunter]

People buy games because they want the game, not because it runs a little faster/slower on their graphics card. Mantle will not sell PC games it will sell radeon graphics cards which is of no benefit to the game dev.

Mantle allows the developers to use code they already written for the consoles though.