AMD Mantle Details from APU13

[inf64]

Images taken from XS topic on Mantle:

Some optimizations related to multicore CPU dispatching

Hardware.fr article covering newest details

Batch definition

Some perf. numbers:

James Prior ‏@cavemanjim 11h "With Mantle, an @AMDFX 8350 underclocked to 2GHZ, the game is still GPU bound on a R9 290X - no loss in frame rate" #APU13 #Mantle #Oxide

Ryan Shrout ‏@ryanshrout 11h API overhead reduced by at least 10x. In some cases StarSwarm is 3x faster with Mantle.

James Prior ‏@cavemanjim 11h "3000 units on screen right now, an order of magnitude more effects - 25000 discrete objects simluating and moving" #APU13 #Mantle #Oxide

With Mantle the CPU becomes ever so less important. No wonder AMD won't bother with new FX in 2014, 4Ghz QC SR APU should be doing fine in Mantle supported games in 2014 and onwards (and will probably be faster than 8350 in older games too).

 

[Kallogan]

AMD will rule the world

 

[LogOver]

[offtopic]

AMD will rule the world

Yep... will rule the world using power point presentations. Seriously, what is the point of such meaningless posts?
[/offtopic]

 

[Ed1]

We have to see how this plays out on desktop .

I can see this helping a lot in the mobile markets where cpu performance is linked to battery life .

But no matter how you shake it , "It" is more work for Dev's . Lets say it is big 2-3x performance boost, even in desktop dGPU's . Then Nvidia and Intel would have to do the same and now you got your code base 4x more work/versions .
And those that say Mantel could be used for Nvidia , well maybe but there GPU works different so its not probably going to help on Dev side of things .

To bad they didn't give real numbers with BF4 , as is Dx11 verse Mantle on there hardware 280-290x .

The good thing that might come out with this is it might force MS to update with some new features in Dx 11-12 etc ,

 

[inf64]

BF4 Mantle version should be released in December I believe.

 

[BallaTheFeared]

With Mantle the CPU becomes ever so less important.

I think this is the wrong way to look at it, if anything it makes the CPU even more important, because the limitation of what can be achieved is still directly tied to the CPU, as it is today. This just opens up considerably more possibilities, however there is still a finite limitation with todays cores. As an example they used an Intel x6 to show even more calls are possible than with the i7 mainstream or 8350.

This isn't meant to design and runs todays games faster, it's meant to increase what is possible; as such the idea that AMD could cripple their x86 market makes little to no sense, as they still want to make money. What this does is bring up their more core strategy to be more inline with the performance of Intel's mainstream.

Look at it like this, with todays hardware you can run 8,000 draw calls in DX. With the best mainstream on mantle that shoots to 50k, or with an x6 80k. Now do you want a game that does 8k or 80k? Why not 120k with an x8 Haswell? See where I'm going here? Games are driven by ever more complex and feature rich technology, Mantle won't drive less for more, it would enable more with more.

 

[SiliconWars]

The important thing is the Oxide guys saying that an 8350 FX is basically equal to an i7 4770K. This will allow AMD to get away with another year or two of Piledriver. You can see how HSA and everything is all coming together for AMD now. I must admit that even 6 months ago I had serious doubts about all of it, but these past couple of months have been incredible and for once AMD appears to be executing to perfection.

Mantle must have both Nvidia and Intel (to a much lesser degree) terrified. And everybody else tbh, Qualcomm, Apple too are so far behind on the software front they are basically nowhere. I just wonder if AMD needs a strategic alliance with one of them to make this happen. Right now they must be a very enticing takeover target.

 

[BallaTheFeared]

Mantle must have both Nvidia and Intel (to a much lesser degree) terrified

Quite the opposite actually, AMD has to be terrified they won't be able to get Intel and Nvidia on board with it.

Without both it's just another Glide without the market share Glide had.

 

[Phynaz]

There's already a large Mantle thread. You have posted in it.

 

[inf64]

Yes it's in the different section (Videocards). We have some CPU specific information here regarding Mantle.

 

[Meekers]

Quite the opposite actually, AMD has to be terrified they won't be able to get Intel and Nvidia on board with it.

Without both it's just another Glide without the market share Glide had.

Why would they be terrified of it? It is not even out yet and they have already landed most of the big players. With the Oxide guys claiming it only took 2 man hour months to implement Mantle it is likely we will continue to see wide spread Mantle adoption even without Nvidia and Intel getting on board.

 

[Arachnotronic]

Why would they be terrified of it? It is not even out yet and they have already landed most of the big players. With the Oxide guys claiming it only took 2 man hour months to implement Mantle it is likely we will continue to see wide spread Mantle adoption even without Nvidia and Intel getting on board.

Yeah, those big players back in the day supported Glide, too ;-)

 

[AtenRa]

Mark my words(and quote them if you like), Intel will pay the same coin in Graphics/GPU Compute etc that AMD payed for the luck of CPU optimized software in x86.
They will throw double the transistor count to only get less performance than AMDs Software Optimized hardware.

This will become even more evident with Kaveri, hUMA, HSA and Mantle for Gaming in 2014.

AVX is already dead with ZERO applications when there are hundreds of OpenCL applications and lots more are coming every year.
Every developer will write code for HSA hardware because the majority of Mobile APUs will be from the HSA partners (AMD, Samsung, Qualcomm, Texas Instruments etc) that will push developers in to that direction.

 

[ShintaiDK]

AVX is already dead with ZERO applications when there are hundreds of OpenCL applications and lots more are coming every year.

No need to make up lies.

 

[Homeles]

Mantle must have both Nvidia and Intel (to a much lesser degree) terrified.

Apparently you have missed the part where this is largely vendor agnostic. It wasn't even buried -- it's on the first slide.

 

[Arachnotronic]

Mantle must have both Nvidia and Intel (to a much lesser degree) terrified. And everybody else tbh, Qualcomm, Apple too are so far behind on the software front they are basically nowhere. I just wonder if AMD needs a strategic alliance with one of them to make this happen. Right now they must be a very enticing takeover target.

Then load up on AMD stock...

BTW, nobody will take over AMD. Apple poached all of the CPU and GPU talent that it needed from the AMD school, and Samsung got a bunch too. Nobody wants a debt laden firm that can hardly compete in its core markets.

 

[AtenRa]

No need to make up lies.

Post the links of 10-15 consumer AVX applications.

 

[Homeles]

Post the links of 10-15 consumer AVX applications.

All anyone would need to do to falsify your inane statement is post a link to a single application.

 

[AtenRa]

All anyone would need to do to falsify your inane statement is post a link to a single application.

You know very well that even if he posts 5 apps they are close to nothing compared to hundreds of OpenCl apps available today.
Everyone was touting about AVX and 256bit AVX before Haswell release, where are the apps today ??? Nobody even mentioned AVX again since then

 

[Homeles]

You know very well that even if he posts 5 apps they are close to nothing compared to hundreds of OpenCl apps available today.

That doesn't make your statement any less asinine than it is.

Everyone was touting about AVX and 256bit AVX before Haswell release, where are the apps today ??? Nobody even mentioned AVX again since then

Who's this "everyone" you speak of? I recall only a single user on this forum who was claiming that AVX2 was going to be a huge deal.

 

[AtenRa]

I will not use the F word, the Intel supporters

 

[SiliconWars]

Then load up on AMD stock...

BTW, nobody will take over AMD. Apple poached all of the CPU and GPU talent that it needed from the AMD school, and Samsung got a bunch too. Nobody wants a debt laden firm that can hardly compete in its core markets.

And yet they didn't appear to get the guys who thought up Mantle. Makes you wonder what they got. And last I knew AMD was poaching Apple guys back.

As for not being able to compete in core markets, the 290X says hi. Also this

8350 comparable to i7 4770K. The only real problem with Piledriver is the lack of proper multi-threaded software. Which you know, is a problem that the guys that Apple failed to steal away are changing the industry with, with Mantle.

 

[inf64]

 

[SiliconWars]

Yep that's a good one, and while some might think it's suicide it's really not.

The PC market is dying out and AMD lost the high end anyway. If they can now get by with Piledriver and 4C Kaveri's that are at least equal to the competition (I would imagine that under Mantle a 4C Kaveri would destroy any i3 however), AMD can spend the money on developer relations instead. They already have the hardware to make use of it so why not? Paying devs to adopt Mantle is the cheaper, easier option instead of trying to catch Intel on hardware alone.

We all know how big it can be and how great it could be for gaming at all levels. Just sad that some people would rather see it fail.

 

[Homeles]

We all know how big it can be and how great it could be for gaming at all levels. Just sad that some people would rather see it fail.

There's nothing sad about rejecting a proprietary standard. The initial information from AMD had led us to believe that Mantle was AMD-exclusive, which would have been an extremely hypocritical move for AMD to make.

The only thing that's "sad" here is AMD's failure to communicate such a critical point.