With only two separate APIs PCs become Consoles.

Across the comments sections and forums I have seen people praising AMDs Mantle.

I honestly think it is interesting in prospect, but am writing this because a specific user made a suggestion that I cannot ignore the possibility of actually happening: Nvidia and AMD each creating their own APIs AND direct X falling out of use.

Now the user mentioned that BF4 is still working on Nvidia cards.

The obvious reason BF4 is still working is BECAUSE they code to BOTH Direct X and Mantle. They are devoting extra resources to it. It's very obvious considering that Mantle won't release for BF4 until 3 MONTHS after the game itself releases.

The user also mentioned SteamOS, but I am going to quickly point out why steamOS is an entirely different situation.

SteamOS if fine as it replaces or accompanies an abstract OS with a simple one and it is FREE for everyone to adopt, while APIs, unlike steamOS can REQUIRE certain hardware just for a game to be playable.

So, why exactly am I so against it? Just look to the past.

Glide API.

It was a single API that allowed voodoo cards to essentially create a monopoly on the industry.

So, why did Glide fail?

It had better performance. But, this was only for a time. Voodoo cards stagnated and became a constant tick in a tick tok industry. They became unable to reinvent their architecture without loosing support of older games.

What happens when we have two different APIs designed for two specific architectures? The architectures stagnates to support older games. History repeats itself, just like glide and the voodoo cards that were pummeled by new architectures that were more open.

Voodoo cards completely flip flopped their advertising position from power to "quality".

API fragmentation turns PC hardware into console cycles.

There are situations in which APIs can exist without fragmentation, but it requires extra work on behalf of the developers.

Ultimately, it's fine to have an API as long as there is a middle ground that is coded for like direct X or OpenGL first, but as soon as that is gone (which is exactly what the user suggested), PCs become fragmented or generational JUST LIKE CONSOLES.

They who can give up [steam library] to obtain a little temporary [power], deserve neither [steam library] nor [power]. - Ben Franklin (sort of)

[LightofDarkness]

I imagine Mantle will see some decent support early on in a few big name titles, but will ultimately fizzle out like Nvidia's PhysX efforts.

Making PC development more complicated will only alienate developers. AMD only make up about 35% of the discrete GPU market, meaning developers will not prioritize Mantle over a Direct X renderer and run the risk of limiting their potential sales. Unless AMD suddenly rule the discrete GPU market, I wouldn't expect to see much support outside a few big name studios bolting on a Mantle renderer as an after-thought.

[ZoyosJD]

Indeed, I fully expect it to fizzle, but this blog was more so a warning for those who think it is just a "BRILLIANT" idea.

I have no doubt that there will be performance gains for a short time. And it will catch attention, particularly among the newcomers.

[Werix357]

pc's are fragmented by nature even with out mantle and the one size fits all approach of direct x and opengl in my opinion is a hindrance to progression.

[SnakeCQC]

Whats the alternative relying on microsofts dx? For me the ideal situation is mantle api gives massive performance increase to the point where nvidia supports it too in a big way and direct x fizzles out.
As a gamer who upgrades his pc every two years or so im never that happy with my performance for too long because of something ive complained about for some time; the lack of optimisation.
Games on my ps3 actually evolve on old hardware; the first bout of console games looks massively different to the twilight bout. Compare visuals from resistance 1 to the last of us or uncharted etc. As a pc gamer too i find this frustrating as i need to up my gpu every two years for a consistent 30-40% increase that will keep me gaming at 1080p with 60-80 fps and anhilated by demanding games like hitman absolution etc.

[LightofDarkness]

Well, we actually have no proof that Mantle is indeed much faster than DX. 9x more draw calls doesn't mean much in most scenarios. DX uses instancing to get around a lot of the issues presented by its slower draw calls. More to that, the issue with draw calls only comes into play between CPU and GPU when the game is utilising more than 4 cores of the CPU. Very few games are actually scaling over multiple cores efficiently at this point. And again, instancing resolves the problem in most cases, it's only when you want a large number of unique assets as opposed to "instanced" assets (instanced assets require only one call for all assets in the instance, whereas unique assets would require a separate call for each asset).

I reckon Mantle will net you somewhere between a 5% and 10% increase in performance over DX. There's MUCH more to GPU/API performance than draw calls.

[Kleptic]

A good blog post imo...

I still think its too early to get any real info about mantle...until that dev summit event thing in november, at least...

AMD's information on mantle so far isn't enough to go on, and it has a few contradictions...in some places its referred to as an all new API (with lower level capabilities), while others say its an additional layer to be used with existing API techniques (in which it doesn't replace directx, just works with it on specific hardware)...

We can all agree...no one wants another Glide situation...but so far, especially with AMD claiming that mantle will be an open API platform for any developer and/or hardware supplier...i don't think that'll be the case...its more an issue of its flexibility, with fundamental architectural stuff between the GCN hardware and mantle's ability to grow with it...just as you said, glide worked great until manufacturing capabilities grew to the degree that 3DFx needed to build an entirely new gpu from the ground up...the problem was that glide didn't support that...and it was over...

mantle has to fundamentally not limit future hardware...that is the only reason DirectX became standard to begin with...directx supports all kinds of different hardware...but doesn't really squeeze any of them...if mantle can get more out GCN cards, but not hold back future products in the process...it could be great...

but, i'm guessing it'll go the other direction...in which it just won't get much support...the only thing to prevent that is very clear evidence of its benefits to gamers...if BF4 does run significantly better via mantle, with the same hardware you already had (as in, the same GCN amd gpu you already had)...they're getting somewhere...but if not, its going downhill fast...

also, posts on battlelog showed mantle's release for BF4 coming in early to mid December...so its most likely more like 6 weeks, not 3 months...but we'll see...

[kryteris]

"What happens when we have two different APIs designed for two specific architectures? The architectures stagnates to support older games. History repeats itself, just like glide and the voodoo cards that were pummeled by new architectures that were more open."

I am not convinced that Voodoo's downfall was because of directx, and/or opengl. Or that the glide api somehow could not adapt to newer games. So, I respect your intent but I cannot really accept the data or lack of to back up your claims. Also the reality we live in today, with solid system architectures from radeon series, and geforce do not reflect past hardware trends.

[ZoyosJD]

It wouldn't matter what what high level API was there. What matters is that glide was too low level to support architectural changes that would eventually come along.

If you look just past the the part you quoted, you will notice I mention their marketing strategy shifting from power to quality. That is because they were being demolished from a specs perspective as Nvidia and ATI introduced new architectures (this is well documented).

Architectures may be "solid", but they are constantly changing. Just look how quickly we went from GCN 1.0 to 2.0 and we can expect then to change with every die shrink.