AMD Mantle API [updating]

With Microsoft hellbent on spreading Win10 as wide as possible, going as far as giving it for free to existing Win7/8 users. I think the faithfulness to Win7 will be a moot point.

 

That would lead to an open standard API with low-level access and the ability to provide IHV-specific extensions. I'm going with the idea that Mantle helped point the APIs in a common direction, but what's left after they go there?
It was the lack of being able to add specific optimizations to DX12 that lead AMD to state that there was still enough to differentiate Mantle going forward.
Is there something else left when
OpenGL would be a standard, it would have buy-in from other vendors, a superset of major engines tied to Mantle, it would be open, it seems to be promising a break from the legacy cruft AMD campaigned against, and it would be present beyond legacy Windows PCs.
I'm asking what is still on the list of promises for Mantle that glNext cannot check off.

Does DX12 only get AMD cards 90% of the benefit of Mantle?
With IHV extensions, what doesn't a low-level OpenGL provide?

I have no reason to doubt the number. I'm asking what do they get by staying on the Mantle train when it seems the goal of convincing the industry on where it needs to go was accomplished.

I remember how DICE very unsubtly hinted at the desire for a multiplatform API. That wish came up more than once--at the Mantle announcement and in that three-way chat between Andersson, Carmack, and Sweeney. The steps that need to be taken to make that happen with Mantle have not happened, and EA is represented at GDC with glNext, (by guess who).

This is the one engine I did not see listed in the GDC schedule for glNext. If Crytek's troubles are settled, I would be surprised if it didn't support the next OpenGL given its multi-platform engine.
(edit: spelling of glNext fixed in the above)

Oxide is at the glNext event as well, as are Epic and Unity.

To do more than just ask vague questions, I am going to put forward a hypothetical:
glNext is generally aligned with what DX12 which is generally aligned with Mantle.
glNext provides greater opportunities for specificity, and it does so many of the things DICE, and Oxide, and Unity, and Epic, and Crytek want. It provides the disclosure Intel called AMD on not providing.
Mantle has slowed in delivering the things that the above wanted, and one of the notable executive voices in favor of the stronger software development focus was John Byrne. He's "pursuing other opportunities" now.
Mantle does provide the ability to get low-level DX12-style access on existing Windows PCs, which is the only thing AMD has made a clear drive for. This is good for a smoother transition, and might be helpful for a non-Microsoft API that wants the broadest deployment possible.
Mantle might be better for those 70 developers and for AMD if it is folded into the foundation for either DX12 or the open low-level API it has struggled to become.

 

One also needs to also ask , what will apple do regarding glNext, seeing as it's going fwd with "metal" ...

"Apple Ditching OpenGL?"

Unless something's happened recently that im unaware of, there's still no known confirmation what Apple is going to do regarding glNext and it's ecosystem?!

 

Apple has probably figuratively dropped more money between the cushions of its figurative money-couch that represents a quite real hundreds of billions of dollars in cash than AMD would likely afford to spend in a decade. It at least can afford to go its own way, and may conceivably find a win in having its own API because it feeds into its own vertically-integrated business.

AMD does not derive significant direct benefit from Mantle, so that initiative's real worth is bound up in how much utility everyone else gets from it. This goes back to my question of what is the sliver in the Venn diagram for Mantle that is not gobbled up by either DX12 or glNext.

edit:
Even with Apple's massive resources, does that blog's thesis hold true now that we know that the rigid model it rebelled against is going away?

 

... maybe if Google decided to take silicon and devices seriously and bought AMD .. and then it too would need it's own low-level graphics api, like apple/ms, that it could control the evolution of ...

... sadly I could actually see Google doing this ...

 

yea , it would be very foolish not to take the windows 10 upgrade from Microsoft unless the free upgrade is something silly like a year free of windows 10 or something . I suspect that a whole lot of the win 7 and 8 user base will be windows 10 in the first few months of release. I would expect only people that don't really care like someone with a 5 year old laptop that checks their emails wont do it.

I bet for windows 8 at least it will show up as a tile to do the free upgrade just like 8.1 did

 

I believe in the contract for the xbox 360 ms shoved in there that if someone offers to buy amd then ms would have a chance to buy them

 

close ... it was for NVIDIA

 

really ? wow I missed the mark on that and I had a 50/50 chance of being right !

 

why would that need to be in a contract ?

 

It's always a risk for a big company to be dependent on a small company (which Nvidia was in the early 2000s) for a crucial product. If they get bought, they may not necessarily be able to fulfill their promises. (See Motorola and their plan to use a fingerprint sensor plans before Apple bought the sensor company.)

 

Here's a preview comparison between DX12 and Mantle.
http://www.anandtech.com/show/8962/the-directx-12-performance-preview-amd-nvidia-star-swarm/6

Is there a case for pouring resources into a 6% performance gain, best case? How about the markedly inferior batch submission latency?
Or you could develop for DX12 and also give gamers with Nvidia cards good performance, especially GTX 980 owners in earlier pages.

 

Supposedly, Mantle exposes low-level, GCN-specific functions that DX12 does not. But I don't think they've ever been used so far.

 

Yep, looks like it's time to kill Mantle.

 

AMD has not demonstrated them, whatever they are. The worthiness of Mantle in that regard would likely be found in the Mantle vs. glNext comparison that should come out in due time.

It's notable that the DX12 path is so much faster at submitting batches. One would think the DX12 would be less mature.
Starswarm's batch-happy nature is very much an optimal case for Mantle versus DX11.

Nvidia's submit latency is also measurably better than AMD. This may even out over time, but I would find it ironic if some of the work that went into making DX11 not horrendous fed forward into Nvidia's DX12 implementation.

 

but if they were for sale then of course ms could offer to buy them. Ms can offer to buy any company they dont need a clause written into a contract.

Would nv want to give amd users good performance if the situation was reversed.

 

NV/AMD aren't the ones developing games. It's a developer decision which API's to support and most developers will choose to support as wide a range of vendors at different performance points as possible.

 

Nvidia might not care to help its competitor, as I'd expect AMD might not.
The devs partaking in Mantle have an interest in this, however. Maybe doing right by them matters more than screwing the competition. At least, I'd like an updated headcount in a few weeks.

 

Without such a clause, company owners/shareholders could easily decide not to sell to Microsoft for whatever reason.

 

The clause as Eastman quoted it doesnt state "that if someone offers to buy nvidia then they have to sell to ms if ms is prepared to buy"