AMD GPU14 Tech Day Event - Sept 25'th

[Andrew Lauritzen]

This isn't an "audiophile" thing, its an audio immersion thing. Techniques are available to bring more realistic sound within (gaming) environments, but they generally suck up CPU horsepower, to the point where game devs don't do it because it pushes the bar too high.

Meh, you have a long way to go to convince me that both A) these techniques are significantly and noticeably better and B) that they can't be made to work just fine on CPUs. The slides even quoted 20% of what I imagine is a single CPU core for a convolution reverb (which is arguably not even the best bang for the buck algorithm to use, but alas)... that's completely within the realm of reasonable. Modern CPUs can handle *tons* of channels and effects if half decently optimized... just check out some modern DAWs if you don't believe me. So yeah, extremely not convinced and sad that I wasted an hour watching that part of the presentation.

The Mantle mention is definitely more interesting, but I think we really need more information before getting excited. In particular, how does it interact with WDDM? Do things remain as robust and orthogonal as they would in a standard DX context? i.e. does it interact properly with the compositor (can I run it in a window?), memory subsystem, TDR, pre-emption, etc? I imagine the solution here must involve stuff like carving out big regions of memory that you just claim are in-use to WDDM, but that can actively interfere with the OS's ability to allocate resources in multi-context situations. I'm guessing the assumption is that this is primarily going to be for "full screen gaming", but if those limitations go too deep, I'm going to be much less interested. Robustness/stability is also a concern, but only time will tell on that one

The SteamOS angle is interesting, but has political problems in the Frostbite case due to EAs (poor) treatment of the Steam relationship in the past. Hopefully either other games can make use of it or EA relents on their need to control everything.

To those of you who think this is going to be just a different API that other vendors can support, I think that's extremely unlikely. They made that fact that it was "very low level" and specific to GCN quite clear. The point is one version of Mantle may only support one or two generations of AMD GPUs, and even forward compatibility is not entirely certain. Yeah, it's entirely possible that when you upgrade to your R3xx or R4xx GPU BF4 will get slower due to having to fall back on the DX11 path. Perhaps they are going to try and maintain better forward compatibility, but that's far from given... compatibility and low-level performance are often at odds and anything they do there sacrifices some of the efficiency they are trying to gain. (Or worse, it handcuffs future hardware architecture.)

OpenGL is fully out of the domain of Windows and wholly in the hands of the IHV's, Mantle will be no different in this respect. Memory re-write was already implemented many releases ago.

Uhh, I'm not sure that's entirely fair. In particular, GPU memory is still owned by WDDM. If you accept that, I'm not convinced you can do a lot better than GL/DX do today. If you don't and start playing games with lying to WDDM and using huge chunks of memory - which I imagine is what Mantle does in practice - then more things are possible, but there's also potential OS issues.

I think the point people dance around a bit too much is that properly supporting multi-application, multi-context situations with foreground/background applications over a wide range of harder is a legitimately harder problem than what consoles have. Thus while I'm the first to say there are inefficiencies still in the software stack (see my interview which was linked earlier), let's not pretend there's a free lunch to be had here with zero compromises. It's really just more of a question of where Mantle lies in the space of abstraction/compatibility/OS features vs. low-level/performance.

I also thought it was interesting to note that Eyefinity and Crossfire were listed explicitly as part of the DX11 release of BF4, not the Mantle update. Perhaps I'm reading too much into it, but I wouldn't be surprised if something like Mantle makes the games that these features play in the driver more difficult/impossible to do and thus games themselves would have to explicitly support them. That wouldn't really be a bad thing IMHO as I've made my feelings on multi-GPU abundantly clear on these forums, but yet another consideration.

Anyways, fun times ahead - looking forward to more concrete info! Cautiously optimistic.

 

[Love_In_Rio]

TrueAudio is an audio processor, not the delivery mechanism - if you route your sound off the motherboard audio jacks you'll still do that with TrueAudio.

So Dave, with TrueAudio ( an the Astoundsound genaudio plugin ) we could get easily Binaural like sound with stereo headphones in all the games without much developer effort?.
Something like this: http://www.youtube.com/watch?v=nKnhcsRTNME ?,
Seriously?.

 

[Love_In_Rio]

Meh, you have a long way to go to convince me that both A) these techniques are significantly and noticeably better and B) that they can't be made to work just fine on CPUs. The slides even quoted 20% of what I imagine is a single CPU core for a convolution reverb (which is arguably not even the best bang for the buck algorithm to use, but alas)... that's completely within the realm of reasonable. Modern CPUs can handle *tons* of channels and effects if half decently optimized... just check out some modern DAWs if you don't believe me. So yeah, extremely not convinced and sad that I wasted an hour watching that part of the presentation.

Well, and however till now nobody used cpu for simulation of 3D sound in a mainstream way. This is a start and if it forces things to happen it will be very good. Astounding sound seems that will be in XB1 and PS4, there will have to work in the CPU, and if console developers use it then TrueSound already will have been a success.

 

[Tim Murray]

if it's still going through dxgk, then Mantle seems like a big waste of time--the kernel thunking overhead and allocation lists will still kill perf. guessing they're working around it somehow, but carving out memory and suballocating was (last I checked, which was admittedly 15 months ago) not a thing you could do, nor was reserving some amount of memory at boot for non-WDDM purposes.

the more I think about it, the more I wonder if this is trickery with IOMMU secret sauce specific to AMD platforms. I can think of ways to make Mantle work in that case without blatantly circumventing WDDM protections.

it still makes me sad that there has been no progress whatsoever on better GPU driver models since WDDM. also, I find it INFINITELY hilarious that the massive improvement we're talking about basically comes down to "yeah let's go back to pretty much what we had in XP, just without that pesky abstraction layer on top."

 

[lanek]

Mantle...
Do we need another proprietary API ?
Glide all over again ?
They have so much free engineering resource at AMD for that ?
Looks like their drivers don't need any substantial improvement... frame pacing... UHD xfire issue...

Dont put all in the same bag ... Driver on a side, and gaming developement on the other. This is 2 different things.

UHD xfire is on the roadmap for be fixed ( dont ask me why 3-4 reviewers have got a Asus P8231 4K some days ago and use driver for write: AMD told us the fix for UHD is not yet in this driver. )

FramePacing have been fixed for standard crossfire ( even if maybe 1-2 games need yet some work ).. FramePacing dont touch single GPU ( i refer to a certain article about this conference on a certain website who was suddenly put framepacing as it is a problem for single gpu ). And lets remember framepacing is only usefull for peoples who dont use vertical syncro and like screen tearing.

 

[Brad Grenz]

All I know is that shit will get crazy this December if the Mantle version of BF4 makes AMD's $300 cards perform like nVidia's $600 cards in that game.

I'm really curious to hear what relationship and/or similarities Mantle has to the PS4's GNM API.

 

[MarkoIt]

All I know is that shit will get crazy this December if the Mantle version of BF4 makes AMD's $300 cards perform like nVidia's $600 cards in that game.

I'm really curious to hear what relationship and/or similarities Mantle has to the PS4's GNM API.

They said also linear scaling with the number of cores, which means better numbers for AMD CPUs as well.

 

[MarkoIt]

Hey Dave, Mantle will run also on Linux right?

 

[jimbo75]

Hey Dave, Mantle will run also on Linux right?

Coming soon I'd guess. This is the "big announcement" for Linux devs I assume, as anything that gets around DirectX is automatically a huge bonus for Linux gaming. All Frostbite engine games should eventually be running like lightning on Linux. AMD will be banking on the rest soon following suit.

 

[NThibieroz]

Emerging from lurking mode for a second to make a few points about Mantle.

Mantle is the direct result of a number of AAA game developers asking us for something like this because of several constraints they're experiencing with existing PC graphics APIs. Those developers want to get closer to the metal and able to achieve more with the GPU(s) on their PC without the API getting in the way of this effort. "More with more" is the term used by Raja and it couldn't be more spot on.
AMD listened to developer feedback and entered this venture with Johan Andersson leading the charge. Johan has a passion for writing efficient code that can allow him to accomplish his technical vision and Mantle exists to serve this purpose.

Mantle is not for every developer. It is a low-level graphics API designed to drive the GPU in the most efficient manner. This level of access requires a bit more development effort than existing APIs to reap the rewards it provides.

AMD remains committed to support DirectX and OpenGL and to keep optimizing our drivers for those APIs. Mantle is an option that some developers may choose, but it does not affect those that wish to keep using existing APIs.

Some technical concerns are mentioned on this thread. The Mantle API has been in design and development for more than 2 years and rest assured we've had some time to think about the best way to address those (and loads of others).

More technical details about Mantle will be revealed in due course. Johan is a keynote speaker at the AMD Developer Summit (11-14 November, San Jose).

Nick - Gaming Engineering Manager, AMD

 

[Bouncing Zabaglione Bros.]

TrueAudio is an audio processor, not the delivery mechanism - if you route your sound off the motherboard audio jacks you'll still do that with TrueAudio.

That should have been made clearer. I've seen a lot of people saying "great, now I don't need to have a soundcard". They didn't understand that TrueSound is an in-game soundscape system, not a soundcard-on-a-gpu.

 

[Brad Grenz]

Thanks for commenting. I do think it will be interesting to watch the adoption among big engine providers. Frostbite is already going to be used in a ton of EA games, but if Unity and/or Unreal offered Mantle support that would be a rather massive shift.

 

[Bouncing Zabaglione Bros.]

I think Manltle is a good move for AMD. In the same way that Nvidia uses PhysX to say to their customers "you can get more with us because we have PhysX", now AMD has gone up a league by doing that with Mantle, and for more than just physics effects.

Yes Mantle is (for now) non standard and it will fragment the market somewhat. My first thought was "does that mean the Nvidia won't be able to run Mantle games, or they will have to run it much slower and with less effects?" That will be good for AMD, and for AMD's customers, and Nvidia can hardly complain that what is good for the goose is now good for the gander.

Ultimately, Mantle is what the devs have been asking for. We've seen for years various developers complaining that the PC APIs take up too much performance, that they want something closer to the metal. It's difficult to argue about it when people are being given what they wanted. And for that reason, I'm in.

 

[KKRT]

Thats fantastic I was worried it was only for people with an av receiver with a hdmi input
does it have to be via mboard jacks ? can it be via soundcard jacks ?
now can we get a guarantee amd will support it and not drop it after a generation or two aka tru-form

http://www.youtube.com/watch?v=nKnhcsRTNME
It works the same way as audio in this video, the only difference is that they calculate it real-time in games. I hope that Nvidia will support it too, we need this tech in every game possible, especially in future Elder Scroll title.

 

[jimbo75]

I suppose the real question is, can Mantle and DirectX be used simultaneously, so that devs could use calls to Mantle for some more extreme effects or when much more performance is required?

If so, this is a heavy, heavy blow for AMD. If it's a choice of either Mantle or DirectX it's going to be much more difficult for them to convince more devs to go with it.

 

[liolio]

Overall and whereas the pace was suboptimal, I think that reactions are completely unfair to what AMD is trying to do.
They spent a lot of time showcasing the strong relations they built with various studios. Whereas people were hoping for more hardware specs to feed the always on going specs war, I think it is a good thing that AMD gave some air time to the people actually working with the actual hardware be it games developers or people working on middleware solution. They are also the guys that make hardware relevant as without content... They don't have many venue to speak about their work or usually it is pretty obscure presentations. Now this was an AMD public event and an attempt to put the spotlight on those people. I think it was a good thing to do, showcase the people you work with.

Now the pace was indeed not OK, it was there first time and it showed. It was late, even the audience seems lethargic, I guess a side effect of having to wait for one hour with not much to do and possibly a side effects of jetlag (not sure about time zone in Hawai).

The reaction are tedious and I think it is sad, especially on a tech forum, though as I said the whole thing needs fine tuning. Now they might giving people who attended lots of infos about what is coming and it is sadly under NDA, it is weird as they won't be able to capitalize on the effort they made in the press. I guess they are constrains that are out of AMD control (some consoles related nda, or what not).

Anyway I think that the picture is not as bad as people are picturing it.
For GPU we know that:
PRT tiers 2 is supported by the new GPU.
They optimized power consumption a lot, claiming that the arch can go as low as 1 Watts.
It is still a flavor of GCN and imo it is great, they need some stability on the hardware if they want the software to keep up. GCN is pretty great already, they are fine tuning it.

Their audio dsp is a really good idea. I'm eager to hear what is the return of the press that attended the show and have experienced the difference from them-selves. I don't think it is a game changer, but better is always nice.
In the gran scheme of things, it is great move, I expect the dsp to be more power efficient than either the CPU or GPU doing the job, it ain't that much of a win for high end set-up burning lot of power, though in power constrained environment (laptop and below) it will help, especially against INtel which could very well be 2 node ahead of the competition (at large) sometime in 2014.
This true audio could be a nice addition to AMD APU.
AMD is struggling on the CPU side of things, let hopes hope steamroller and excavator after him are significant improvements, in any case freeing CPU cycles will help AMD CPU especially if it both free resources and at the same time translates in a better experience.

Mantle, it is really interesting too bad we have to wait for two months before we get proper details about it

RaptR is also neat, matching or exceeding what Nvidia were doing.

Overall AMD gaming strategy seems to comes together well, even though the presentation was subpart.

As a side note somebody asked me why I spoke about SteamOS, so I'll answer. Indeed nobody spoke about it during the show, but I've feeling, gut feeling nothing more, that both on the software and hardware side, there is a growing will to break out of "constrained" (when not close) environments.
To some extend one could see msft (either with windows or xb1) or Sony as companies that stand between the hardware guys and the content creators and take a tad more than their fair share of the revenues they could not generate without the two former categories of company. It is not "usury" but it is not that far either.
So we don't know much about Steam OS now and the compromise Valve was willing to do to break free. We should knows soon enough, on top of it I don't expect a violent disruption, if anything I would expect a slow, or even sluggish, start, though even with mild support, just enough to keep the thing slowing rolling, and given a bit of time and better tech there is a potential for a snowballing effect. Anyway it is in the hand of Valve and how far they were willing to go in "open-ness".

 

[Arwin]

Thanks Nick, for commenting.

I think Mantle has huge potential. There's some really strong alignment of stars that could make this work really well - overlap with Playstation 4 development (which I hear supports low-level GPU access in a very similar manner), the prominence of the CryEngine's, Unreal's, Unity and other engines of this world that would seem likely to take advantage of Mantle, the recent SteamOS announcements, the resurgence of OpenGL thanks to the portable market - all things combined the impact could be quite large indeed.

Also, it should be very clear where this is coming from - there have been lots of talks by various high-profile developers that PC GPU efficiency through DirectX is a fraction of what it could be, and DICE in particular have made several comments that they were trying to work with Microsoft to get DirectX to a better place for developers in terms of performance. That they have now worked together with AMD and Mantle is the result is very significant.

 

[ChrisK]

how long has it been in development for?

It instantly reminded me of this interview from early 2011:

'It's funny,' says AMD's worldwide developer relations manager of its GPU division, Richard Huddy. 'We often have at least ten times as much horsepower as an Xbox 360 or a PS3 in a high-end graphics card, yet it's very clear that the games don't look ten times as good. To a significant extent, that's because, one way or another, for good reasons and bad - mostly good, DirectX is getting in the way.' Huddy says that one of the most common requests he gets from game developers is: 'Make the API go away.'

'I certainly hear this in my conversations with games developers,' he says, 'and I guess it was actually the primary appeal of Larrabee to developers – not the hardware, which was hot and slow and unimpressive, but the software – being able to have total control over the machine, which is what the very best games developers want. By giving you access to the hardware at the very low level, you give games developers a chance to innovate, and that's going to put pressure on Microsoft – no doubt at all.'

 

[liolio]

Wrapping things up, I had a chance to briefly try Xaviant’s Lichdom audio demo, which is already TrueAudio enabled. As someone who’s already a headphones-only gamer, this ended up being more impressive than any game/demo I’ve tried in the past. Xaviant has positional audio down very well – at least as good as Creative’s CMSS3D tech – and elevation effects were clearly better than anything I’ve heard previously. They’re also making heavy use of reverb, to the point where it’s being overdone for effect, but what’s there works very well.

Sounds nice (lame joke...). the paper is here

 

[rapso]

It instantly reminded me of this interview from early 2011:

they had been probably working on it back then already, that's why they tried to check what feedback they might get if they'd announce it, I guess.

it remind me also of

I think CUDA is doomed. Our industry doesn’t like proprietary standards. PhysX is an utter failure because it’s proprietary. Nobody wants it. You don’t want it, I don’t want it, gamers don’t want it. Analysts don’t want it.