AMD GPU14 Tech Day Event - Sept 25'th

[KKRT]

Instead of working with Nvidia, work with MS and Nvidia for DX12.
Trouble is the reason there is even a DX is because vendors have their own interests, that's why I think it would be good to have MS involved.

I got the impression from the presentation that it's not on consoles, they said consoles have light API's which allows them to get more performance, and that's what Mantle will bring to the PC space.
Not saying it couldn't happen though, but I don't see MS being very eager to go that route. They'll try to streamline DX for XB1, and expose the additional functionality.
Sony have their own APS also but may be more open to it I guess....

Mantle supports HLSL, and some others I believe, to help with the porting process.

I really liked the True Audio stuff. Finally audio is being addressed, and it does make a lot of difference in games, it's just that most people haven't had the chance to experience it

I would prefer to Nvidia work on Mantle than DX 12, because we wouldnt need to change operating system again.
Mantle will enable for full access to GCN and because consoles have GCN too, all optimization on them will automatically benefit to games based on Mantle.

Yeah, True Audio was actually great, i would love to see more devs supporting it though.

 

[Rangers]

Yeah so umm, Mantle seems like it could be a gamechanger. Really huge.

 

[Jay]

I would prefer to Nvidia work on Mantle than DX 12, because we wouldnt need to change operating system again.
Mantle will enable for full access to GCN and because consoles have GCN too, all optimization on them will automatically benefit to games based on Mantle.

Have to admit I do not know that much about draw calls etc, apart from it has always been an issue, probably lot worse than it is today. If that's part of the problem in DX, get together and try sort it. Only 2 GPU vendors to deal with now.

Well hopefully they can both get together like you say, but as I said vendors have own interests and architectures.
Optimisations done on consoles can probably already be brought across to the PC, especially now.
What could be brought over would be greater if the consoles have Mantle, but we don't know that yet (I believe)

If Nvidia and consoles adopt it then it could very well be a seismic event.

 

[Nemo]

Wow, I will buy 290X for Mantle

 

[silent_guy]

So AMD organized a broken live stream to announce that at some point in the future there will be a faster GPU. Do I have this right? Did they hire the PR firm that handled the first XBO announcement? What's the point of doing this?

 

[itsmydamnation]

my paraphrasing of the mantel/repi section of the stream

15 games in dev with forstbite ( will all be mantel?)

work with amd for both CPU and GPU on BF4
traditionally on PC CPU easy to dev for but GPU hard, memory abstracted away from GPU, GPU reliant on CPU
API limited and have no ability to predict desicsion that the game engine can

mantel
game engine takes over most API functions
game engine has explicit control over memory and hardware components

CPU
low overheads for rendering/loading and streaming
parellel rendering, linnear scalling

GPU
full access to GPU hardware functions
lots of low level optimizations made possible
Low latency multigpu rendering


hint at other frostbite game support

 

[DieH@rd]

Here is the announcement of Mantle in Frostbite 15min presentation, starts @ 1 minute mark

 

[DavidGraham]

Well naturally only DICE is working with AMD right now, other studios haven't announced anything just yet , so we will have to wait on that front.

Still we don't know much about performance or architectural enhancement , or even the release date for that matter.

The fact that AMD didn't really stress the point of having the fastest GPU doesn't bode well for them beating NVIDIA this round, we will see.

However I fear judging by the Fire Strike slide , things are looking like this :

R9-290X = 7970Ghz @ 1300 = GTX780
R9-280X = 7970 = GTX 770 - GTX 680
R9-270X = 7950 = GTX670 - GTX760
R7-260X = 7850 = GTX 650Ti Boost


Feel free to correct me ..

 

[Alexko]

So AMD organized a broken live stream to announce that at some point in the future there will be a faster GPU. Do I have this right? Did they hire the PR firm that handled the first XBO announcement? What's the point of doing this?

At the end, one of the AMD guys said something along the lines of "We could have talked about hardware for 2 hours, but we thought it would be better to talk about software, collaboration with developers, AMD's commitment to gaming, etc."

So I guess that was the point. And it was driven home, but ever so boringly.

 

[itsmydamnation]

R9-290X = 7970Ghz @ 1300 = GTX780


Feel free to correct me ..

didn't they say 4 tri a clock? so no.....

 

[psurge]

At the end, one of the AMD guys said something along the lines of "We could have talked about hardware for 2 hours, but we thought it would be better to talk about software, collaboration with developers, AMD's commitment to gaming, etc."

So I guess that was the point. And it was driven home, but ever so boringly.

Can we expect to see any more information on the HW out of this event?

 

[silent_guy]

At the end, one of the AMD guys said something along the lines of "We could have talked about hardware for 2 hours, but we thought it would be better to talk about software, collaboration with developers, AMD's commitment to gaming, etc." So I guess that was the point. And it was driven home, but ever so boringly.

Audio and APIs are the kind of stuff that you'd expect as filler for those poor journalists who want to write about more than just benchmark bar graphs. Not necessarily uninteresting, but filler material without a doubt.

They completely misjudged their audience and what it wanted to hear.

I only saw 2 minutes of the presentation with that, no doubt very smart, but mind-numbingly boring Indian guy and then dialed out. Looks like I didn't miss anything.

 

[DavidGraham]

So it seems the reason AMD went with Mantle is -as stated in the press conference- that DX and OGL pose great challenges on utilizing all CPU cores ,not to mention the draw call deficit and the reduced memory control due to software overhead. Also the limited CPU/GPU cooperation that is key for AMD's APU business .

So I guess their solution sounds logical in the light of these things. an option to use a lower level API as a performance booster is nothing particularly bad, NVIDIA could do the same too , so long as this trend doesn't cross the boundaries of performance enhancement, then I presume all will be good.

 

[swaaye]

I could see NV rebooting Glide as their own proprietary API. There's nothing to cheer about with such things coming back. Ideally AMD, NV and Intel would all go OpenGL and develop it into the ultimate game API. But that involves being interested less in market dominance through API lock in.

 

[BRiT]

I could see NV rebooting Glide as their own proprietary API. There's nothing to cheer about with such things coming back.

WebGLide!

 

[karlotta]

hmmm, Cant wait till Nov. I see that Mantle is open?

Oh look!

http://www.beyond3d.com/content/articles/120

http://forum.beyond3d.com/showthread.php?t=61298

 

[hoom]

So far I see no reason that all the arguments against CUDA don't apply to Mantle.

Its not cool when Intel, MS or NV take advantage of a competitive advantage to lock out others, even more not cool for AMD to do it when its their turn after pushing open standards so long.

Now, if they have built Mantle to allow NV (& others) in & NV will be happily building their own implementation then unlikely as it is, that would be all good

DX has filled an important role for a long time in providing an industry standard API but MS has their own agenda: not compatible with multi-platform & tending to lock improvements to new OS versions.
OpenGL while neutral & multi-platform has its own issues.


Audio stuff, yeah cool but not really a game changer, I'm no audiophile.


No juicy hardware tech details or launch date for new card: AMD I am disappoint.

 

[Kaotik]

One thing odd, from Anand's liveblog:



And



According to this, it fits for Tahiti, but Pitcairn 7870 only gets ~4700 Fire Strike according to this http://www.hardwarezone.com.sg/feature-3dmark-fire-strike-desktop-gpu-performance-review

It's a 2 GB card which seemingly rules out Tahiti.

So yeah, an apparent discrepancy, I hope they have some turbocharged Pitcairn for 199 (you can get a 7950 on sale now though for close IIRC).

260 can't be pitcairn since it supports the new audio features

I thought it was some frostbite thing from Johan whatsisname's presentation - though I admit I only half listened to what he was saying, his english was pretty terrible. However, it looks to be a general AMD API instead.

In other words, 3dfx Glide anno 2013, essentially.


No idea. It would mean extra work for them. However if the performance benefit is sufficient, maybe at least the big gun developers would bother, I'm not sure I actually want it to succeed though, I hate proprietary shit, because when you get down to it, all it ever really is is vendor lock-in to prevent you from buying a competitor's product.

Glide was like that, physx is like that, and now mantle is like that too. And it's not anything PC users should applaud, even though we'd benefit from it in the short run, because limited choice limits competition, and if it's one thing that's an universal truth with us humans, it's that limited competition breeds corruption and lazyness. All we'd get from decreased competition is less development, less reliability, less performance and higher prices.


Yeah, that would be really interesting to know, but the presentation didn't really do a good job explaining that aspect. Maybe NDAs were involved, I dunno.

Mantle isn't like Glide or PhysX based on what I've read - NVIDIA is free to support it without any fees or such, unlike Glide (not allowed for others) or PhysX (CUDA requires fees)

 

[Grall]

who said mantel is vendor locked?

You can pretty much infer it first from AMD not saying it's either open or vendor-independent, and second, the fact the innermost sphere in their graphical earth representation is labeled GCN...so with this in mind, I think most people would assume it is proprietary.

Also, AMD could have structured their API closely around the functioning of their own hardware, making it more or less inefficient for a different vendor's architecture (IE: NV, Intel.)

Of course, this could be wrong, but again, AMD never said mantle would be open, so it doesn't look that way right now. Of course, there's A LOT they did not say in those two hours and fifteen fucking minutes, so that might not say a great deal anyway...

 

[karlotta]

...

Now, if they have built Mantle to allow NV (& others) in & NV will be happily building their own implementation then unlikely as it is, that would be all good

...

Its open..."We've been told at the GPU14 Tech Day event that the Mantle API is open, so theoretically Nvidia could purpose the technology in their GPUs."

Hope to see a sdk or some such in Nov! This is brilliant. Much better than the next H/W, tho next H/W is needed.