"Ryan, Why don't you talk about how great nVIDIA is and how much AMD sucks? I mean I realize that when AMD had better power consumption, I cheered on the Geforce GTX 480, because I'm dishonest like that, but now that nVIDIA has better power/heat levels I have decided that these things matter"
When can we expect a full article/review on Mantle? So far we're seen preview for Thief and BF4 but not full detailed 20 pages or so review on the thing. I use it on my 7970 GHz and it seems fast and stable but it could also just be my imagination tricking me with a "placebo" effect.
Definitely should be some slower CPUs in the testing. Pentium G-series, maybe a C2D/Q. I've seen claims of 50-to-300% gains on forums (with numbers to back it up). for max, average, and minimum frame rates. Even AMD's press coverage made it known that the fastest CPUs wouldn't gain much (maybe 5%), so it would seem fitting to run tests using low-to-mid range hardware.
Definitely. On my 6core phenom2 & 290X, I doubled my minimum framerate, and nearly doubled the average. Why would the only graph be with a high-end overclocked intel cpu that shows minimal gains?
I agree. Using a very high end CPU and just cutting cores and clock rate doesn't give us much hint of the point behind Mantle.
Ryan, you wrote the Mantle portion like you expected everyone to have a high end CPU. I know you tested for lower CPU performance, but that's the angle of the article. See below...
"The gains at the high end aren’t worth writing home about, but since we need the CPU to churn out a fairly high framerate regardless, there’s a much greater opportunity to benefit from Mantle on lower end Intel CPUs and AMD’s CPUs/APUs."
So you're thinking is "does Mantle help, and how much?". That question doesn't cover the whole picture. Mantle may lead to purchasing decisions. Someone could buy a lesser CPU and still get close to higher end CPU performance because of Mantle. What about that? I agree that including some lesser CPUs in the testing would really be needed for this as well.
Of course it should, todays games are made for todays HW and API, means mantle cannot bring performance improvement if game itself doesn't need more. Only way is to use much slower IPC and more cores, still most improvements will see when optimized games will be made from ground up, like star swarm demo, I have 300% FPS there in RTS view. PS4/XBOX uses 8core jaguar architecture, same is used in ultrathins and tablets! CPU with few times higher TDP/ IPC and even less cores, cannot be even remotely comparable to that. But it proofs that today's CPU have much more performance than we really need since all todays games cannot optimally use more than 1core anyway.
Note "Shipped" as in, designed fully with Mantle in mind, not post-patched to toss in one or two effects. Granted, Thief utilizes the GPU speedup quite nicely.
I can understand your dissatisfaction with the Thief reverberation, I agree that it does sound unnaturally loud. However, I disagree with you in that I also think the Tuscany demo sounds unrefined.
I would expect in 2014 a dedicated audio processor could account for the noise blocking effects of walls and the noise transmitting ability of open windows. The fireplace is unnatural - it's roaring loud inside the building but as soon as you leave the building that is full of windows it can't be heard. And then even when you go near a window at around 0:34 the fireplace is still silent.
In my opinion, all this demo did was brought back feature parity of games from 1998. I would have expected more from a company showcasing audio 16 years later. Are my expectations set too high?
To really push audio simulation forward needs more than just a bit of convolution and managing not to screw up the spatialisation with a HRTf slapped on top. This needs a lot of different techniques to come together at once, something that isn't too easy to meld into your GPU (particularly as audio processing requires a few streams of audio processed intensively, rather than the lots of simple parallel operations GPUs are good at in general), meaning three's not much incentive for Trueaudio to implement something like GSound (https://www.youtube.com/watch?v=buU8gPG2cHI or other environmental audio simulation engines.
Great article, and really digging informative comments such as this one. I am old enough to remember the first push for environmental audio by companies such as Creative, and it's something you don't really think about but seems (at least from my perspective) to have fallen by the wayside. It's a little depressing we haven't really experienced much improvement in the audio space in so long. That GSound demo video you linked was impressive, and gives me hope that somewhere sometime a game engine or hardware company will integrate audio raycasting.
Actually, in 1998, there were better audio solutions that TrueAudio. A3D! AMD should have bought the patents from Creative and brought A3D into the 21st century. http://en.wikipedia.org/wiki/A3D
The Aureal vs. Creative story is a good example of why Creative is where it is today, instead of being a major player. I remember my first Creative Multimedia Kit, with a 2X CD-Rom, a SoundBlaster 16-bit sound card and speaks (if I remember correctly).
Ahh yes A3D was ear-candy.. Bastards at creative killed it... at the time Creative's implementations where much worse ( sometime after A3D was murdered, I purchased a creative card) I have never purchase any creative card again, ever.. on board sound for me, whatever crap it is.
When EAX worked, it worked well. I still break it out for playing Halo PC using Alchemy. The advantage it brings in hearing approaching footsteps and chambering rounds is astonishing. I can navigate some maps almost entirely by sound. Of course, there's always the hilarity that ensues when the physics freaks out with collision detection and a warthog starts bouncing half in a wall... the audio just farts relentlessly.
Come to think of it, A3D had a demo that was nearly identical to this Tuscany demo, I remember playing with it and thinking it was neat.
Hah! I was thinking of this too when I watched the Tuscany video. I remember playing with this on my Aureal Vortex 2 SQ2500 Superquad back in 2000. Which reminds me... what is it with old sound cards and ridiculously long product names? My brother had an Aztech Labs Sound Galaxy Washington 16A soundcard as well... how did they fit it all on a box?
Wow! I was floored by that. I felt like I was immersed in one of those adventure rides at a Theme Park. Too bad no games use anything close to this today. That was way better than the demos from my SB EAX cards. And blows away the Tuscany demo too.
I'll quote wiki: 'A3D uses a subset of the actual in-game 3D world data to accurately model the location of both direct (A3Dspace) and reflected (A3Dverb) sound streams (A3D 2.0 can perform up to 60 first-order reflections). EAX 1.0, the competing technology at the time promoted by Creative Labs, simulated the environment with an adjustable reverb—it didn't calculate any actual reflections off the 3D surfaces.' Well EAX (even the newest) never got this killer feature. That's why A3D worked so well and consistent.
Half Life was *the* killer app for A3D, maybe AMD can bring TrueAudio alive with HL3 ;) Way back with HL1 the framerate did take a hit from enabling A3D but it was absolutely worth it. Playing beta CS with headphones and A3D was almost like cheating, you could easily track players through walls. I think I have my vortex 2 stashed somewhere in my closet..
A3D had a horrible CPU utilization rate. I remember having my frame rates halved and I was using either my Voodoo 2 12MB SLI setup or my Voodoo 3 3000 16MB back then.
While EAX doesn't do all the calculations, it adds custom DSP to in-game elements and each virtual environment to better reflect (no pun intended) the audio heard by the listener without impacting system performance. Long before ambient occlusion, there was aural occlusion, which EAX brought to the table after they consumed A3D.
Probably the best we can hope for with AMD TrueAudio is competition from Creative in the form of A3D's resurrection... but I'd like to think that A3D (like EAX) can be adequately surpassed by modern technology.
A3D was unbeatable in true HRTF processing when one was wearing headphones. I had Aureal Vortex 2 card and in Quake 3 Arena (before the patch which removed A3D sound) and in Half-Life and also Deus Ex it was like having third eye. A3D did occlusion and reverb processing of the sound sources on the scene geometry. I have never heard anything close enough since then in any game.
As someone who also implemented A3D support in some graphics app, I can confirm that API was quite similar to OpenGL philosophy. The scene could be passed to OGL and A3D in very similar way (back then i.e. in 2001). There was however a drawback in CPU utilization (15-30% at that time), which today with current CPU power and resources could probably translate into negligable numbers.
So I do not think we need specialized HW for this, just standardized SW API would be enough and with current CPU, even bringing in old A3D API would definitely move the 3D sound forward.
The reason A3D worked so well was the 3D model of the scene they had to pass to the card. The penalty paid was mostly for the transfer of those models into the sound card memory, which happened over PCI bus (slow). As scene complexity grows, so would this cost. You also would need additional memory on the sound card.
With by modern standards ancient GPU and CPU hardware and a vortex2 card enabling A3D did not impact games noticeably, except for having the great sound. You can talk theory all day long but the fact remains it's incredible that so long ago you had a top notch soundcard and soundsystem and after all the years they can not catch up to it. And then dare to claim it's because of hardware limitations? That's insane and you must be delusional to convince yourself of that.
AMD already implements a D3D API, why not simply take the geometry passed to GPU and use it also for sound rendering at the same time? The only problem is that you don't get the "textures" of the geometry as all you get is graphical textures which do not express properties of the material with respect to sound waves. That could be solved with some very simple extension to D3D (or Mantle) which would allow one to pass sound properties of the material together with the graphical texture.
If they keep working on the project I guess that in another few years AMD will reach that level shtldr, although no doubt it will only work with the then current model graphics card I bet. For some reason many developments seem to deliberately go in small steps when they could easily leap since the concept is a repeat of the past and rather obvious.
A3D was not that great. I still have my Aureal Vortex cards. They really were not that great at all. In fact my EAX cards, at the time, produced better positional audio (SB Live!). The problem with creative, back then, were bad drivers and compatibility issues (especially with VIA chipsets).
I agree with you that Tuscany demo is by no means a leap forward since 16 years ago it is a leap forward in what is found today. I could easily pick out exact locations of sounds locations and distance. I've tried this on games like Guild Wars 2 or Skyrim and though I can kinda get a since of directions it is not nearly as clear as this demo.
Fireplaces will resonate within a building but not outside. What you demand is unrealistic. Sorry bud but you just don't know what you're talking about. There are high frequency sounds and low frequency sounds. High frequency sounds carry themselves over longer distances. Just an FYI.
joying the fray -> joining but more on that latter -> later
also lots of places could have used some more commas: launch title for Mantle Thief -> Mantle, Thief along with Thief AMD -> Thief, AMD unlike Battlefield 4 Thief -> Battlefield 4, Thief
Sites should make separate closed comment section for such help tynopik, and then when they spot a reliable person assisting like you, give them access to it. That way they have free proofreading and improve their copy, to everybody's benefit.
No, it really wouldn't make sense since AMD does have the extra CPU cores, in DX games, that generally remain idle. AMD do have the extra CPU resources available for audio.
Problem is Direct3D and how much it sucks donkey nuts.
Out of all the games that could potentially benefit from this, Thi4f is absolutely top of the list but has pretty much already flopped for a variety of reasons, ironically, one of which is a pretty crappy sound engine that still can't match a 1998 Looking Glass Studio game... Despite its epic franchise's history, Thi4f has an awful sound engine. Even the original game managed to more accurately replicate the effect of attenuation over different surfaces, etc. "Thi4f" somehow manages to screw it all up so that solid walls can make no difference in volume / frequency to a sound. It's also bugged in that one conversation 30ft away can jump in volume to match one 2ft away. All surfaces except glass & water now sound the same, etc. It's nothing like the stone vs wood vs tile vs moss vs grass vs metal, etc of the original. They can't even get the default music vs speech balance right with people complaining of not being able to overhear conversations in many sections.
TrueAudio is promising, but it won't fix flawed psycho-acoustic programming in the core game which is the #1 key issue of "what sounds real". More audio channels won't make a game sound better unless devs spend a lot more time & money on realistic subtle sounds and learning to position & mix them well rather than just "slap more reverb on more surround sound explosions". 96-192khz sampling rates are pointless for everyone not sucked into audiophile snake-oil, etc.
I don't doubt that TrueAudio may provide some great effects in demo's, but really game devs aren't going to put in a lot of effort to perfect it since audio typically takes a back seat to the "Oooh! Shiny!" GFX obsessed brigade. After all, we're talking about the same devs who churn out games as quickly as possible and release them in a half bugged state and use "Day One Preorder" gamers as glorified beta testers just to save a few bucks on testing...
Many new technologies sadly end up unrealistic because the devs deliberately exaggerate them as a "showcase". With GFX we've had unrealistic bloom, DOF, etc, and with audio, usually they put in too much echo, and have limited ambient sounds for budgetary reasons. Think of real life - you often hear over 200 ambient sounds in the space of an hour. Now think games - you're lucky if you get 20 throughout the game. and the few you get are often overly loud or "crammed" into too small an area. The issue isn't really "GPU offloading", it's the psychology of developers.
And just to add - if anyone's in any doubt over the attitude of developers making things "more real at any cost", with Thief 4, they took the 11 light gem visibility levels of the original, and dumbed it down to just 3 levels : "pitch black", "as visible as wearing a fluorescent yellow jacket in broad daylight", and "half-way between them"... It ain't the engine's that need over-hauling, it's the pervasive dumb-it-down-for-the-casual-audience attitude...
I would disagree with your assessment: game devs already have to put A LOT of effort into sound, but ultimately it all comes off as cheap and useless crap because the audio engines themselves are crap.
For a classic reverb sample: firing a gun in open grassland would normally generate little to no reverb while the same gun going off in a super-sized cathedral would generate tons of it. The difference between a good sound engine and a bad one is that a bad sound engine gives you option to use reverb (and games today do when they think you're in a cathedral) and a good sound engine will take the sound source and run it against scene geometry. That way you'd still get lots of reverb for gun going off inside the cathedral even if you are standing a substantial distance away from the entrance.
The first problem with the above "good" audio engine is that each piece of geometry now also has to have all the sound properties (reflection / absorbtion / scattering / whatever) just like it now only has light properties. The second problem is that you need to push the geometry to audio processing unit as well as to graphics processing unit. What TrueAudio could offer is elimination of this dual transmission since this is supposed to be processed on the gfx card simultaneously.
Nice article, I understand there is no time yet for a very big article because there are only 2 games supporting Mantle but good to see it making quite a difference in some situations. About TrueAudio, I agree there are very few games with outstanding audio, I hope this will improve. Ryan, have you ever played the original Bioshock? I think that game actually has the best audio of all games I've ever played, from a technical standpoint. I played it with a Creative Audigy 1 soundcard and 5.1 setup and the combination is absolutely astonishing. Maybe with headphones it will sound equally impressive.
Damn sad news for sound. Thanx for the subjective impression. Thats what matters. Hopefully in the future ps4 sales can drive devs to use this tech better. How many play consoles with headphones?
The irony is they were so afraid that you wouldn't notice the reverb effects, they amped them up too high and you definitely noticed them...
...too much.
But listening and perception of sound is so subjective that they'll never get it right for everyone. I wonder if they'll get it right for the majority.
That said, the feature is nice to have. I'd like to think nVidia and Intel will follow suit. Matching what consoles have done with regards to audio processing would be nice for the ports that are coming.
After reading this article and the comments I'm convinced another piece of hardware isn't the solution. CPUs are already unbelievably powerful DSPs. If they nullified the need for dedicated audio hardware 15 years ago think about where they are now. Good audio engines can do these fancy things like terrain mapping for reverb iff the developer can tag the "type" of terrain each surface is. We're all sitting here with hefty CPUs idling the majority of the time and GPUs that are maxing themselves out. Why not implement an exotic audio engine licensed from some company whose only purpose to make exotic audio engines and forget the hardware and all the headaches that come with it? A DSP won't be a fourth as powerful as a 4670's idle time during gameplay and for those still on Q6600s there's always the "medium quality" option.
All that said I don't know of any exotic audio engines or if they even exist, but they should and AAA games should be using them. Idk if it can ever be abstracted so it's, as always, up to the devs. The big deal about trueaudio is that it draws attention and maybe the devs will start paying more attention/money to the engines and implementations.
I've never seen a Creative product do these kinds of 360-degree surround effects in normal stereo. Just set up 5.1 or 7.1 and then plug in a pair of stereo phones--it's really, really poor...;) Most of the sound is just lost. This is so cool because it happens with normal stereo phones, imo.
Sound in individual games is just like graphics--all games are not equal. TrueAudio support is just ramping up. The Tuscany demo and the one I linked above reveal what's possible--just *try* doing those on a normal cpu...;) The reverb is small pototoes--its getting a 360-degree sound stage from stereo that's the really impressive attribute.
"I for one find 24fps material jerky, blurry, and blatantly unnatural."
that's totally the opposite reaction of filmmakers since the advent of the film camera, including filmmakers with giant statures like kubrick to name just one. and not just filmmakers, but cinematographers who've worked for decades (perhaps their entire lives) looking thru that tiny viewfinder and seeing 24 fps. and my bet is that it didn't look unnatural to them. or jerky.
jerky, unnatural and those other adjectives you use did not crop up until video games. just to make it clear her,e buddy!
24fps is actually the most optimal fps as far as realism goes. For games, more is desirable because you want faster responses from the game. And also because of refresh rate crap.
24fps is as close to how our eyes see as is feasible today.
That Tuscany villa demo has the same problems as most games these days - the underlying recordings aren't good. For instance, that church bell sounds terrible. There's only so much you can do with dsps and processing if what you're starting with isn't good.
Would be nice to have some benchmarks with the Kaveri APU. I think Mantle would be very useful in that segment. Low details, good kaveri APU and Mantle should show good gains in my opinion. Why not trying it out Ryan?
The reason TrueAudio in Thief sounds unnatural isn't because it's "overdone", it's because it's not working properly. In the example clip (and other youtube videos I've watched) you hear a doubling effect which isn't supposed to be there. When you hit the wall a sound is generated which bounces off the wall with a slight delay, so you hear the same sound twice. In real life this also happens, but it's so fast that you can't hear the reflection when you're right next to the sound source. So what you're hearing is an obvious problem with the implementation of TrueAudio since these kinds of delays aren't supposed to be part of any realistic simulation of sound space.
You claim that the software implementation of Convolution is broken, but it isn't. I tried it myself just now and it's working just fine, but it's subtle. Too subtle to be realistic, and certainly for the average player to notice. There isn't too much reverb with the current implementation, in fact there isn't nearly enough. The only thing that jumps out at you with the current implementation is the bugs.
"You claim that the software implementation of Convolution is broken, but it isn't. I tried it myself just now and it's working just fine, but it's subtle."
To be clear then, you're hearing a difference between Off and Software? I'm not hearing any difference in our test sample or on play-throughs, so I'm curious if you're hearing differently.
Though it should be noted that both Software and TrueAudio should be producing the same results. There is not supposed to be a difference between the two.
It's subtle, but definitely there. When I started the game I noticed the lightning sound in the main menu producing a reverb it normally doesn't. The game had defaulted to software as I have an Nvidia card. I had a save file in chapter 5 where you're in an insane asylum. Banging on the bars produced a distinct echo in the room, while turning convolution off removed any echo. Similarly when exploring the city I noticed echo effects on sounds like distant crows and dogs. It was hard to hear on NPC dialogue as they're unnaturally loud and make it very hard to focus on any reverb decay. Some areas don't appear to have reverb and sound like your example clip. Maybe it's being simulated, but it's to such a low degree in volume to make it inaudible.
Thank you, Ryan. The Tuscany demo is great. I initially put my headphones on with the left and right sides flip-flopped. That turns out to be a good negative test, since I immediately felt disoriented. I rectified my headset and fell in love with it. For years, we had to rely on on-screen visual cues in FPS to tell where bullets come from. I hope this technology will allow us to truly immerse ourselves in a firefight! Perhaps, Ryan, you can set up a scene like that for comparison?
As other here requested - a minimal numbers will be nice. Also, the distribution of frame delay time / rendering times. Mantle might also result in a smoother gameplay that way..
I did my own tests with 4,4 GHz 2500K and 280@1100/1500. At 1080 max settings with no SSAA I got 67.1 FPS with DX11, and 77.7 with Mantle, so there was definetely some benefit even for a high end system, but with high SSAA the difference was 49.9 vs. 54.7. I also tested eyefinity, but gains were marginal with ssaa off, and high SSAA broke the memory barrier. Personally I'll keep playing the game at 6048x1080 with DX11 CrossFire and no SSAA since the game does have occasional issues with SSAA even at low when using Eyefinity. Hopefully they'll get the support for CF in Mantle before I get tired of the game. I wouldn't mind using high SSAA, but then again it doesn't seem to fix all the aliasing.
One of the most aurally amazing games I've heard, using EAX 5, is Uru: Ages Beyond Myst. (You can try Uru Live for free.) While the game isn't the greatest, the audio is amazing -- walk through caves, and you'll hear the acoustics gradually change, and sound sources get properly obstructed/occluded.
My ideal audio would be that sort of detail, but with the wave-tracing sort of thing that Aureal hardware apparently did.
I hope the enhancements won't be restricted to just headphones. Even with 5.1 speakers, there's definitely room for improvement.
The quality of "g33ks" has gone down significantly over the years. Most people calling themselves g33ks (or commenting on tech articles) don't know anything about the topics they're covering.
I thoroughly enjoyed this article because it clearly pin pointed the shortfalls and the positive aspects of both Mantle and TrueAudio.
Also since 2004 (around then) I have opted to not take many comments seriously as nVIDIA has been hiring agencies to make fake posts disparaging competitor products (heck the NSA now does this as well).
THE REAL QUESTION IS "ARE MANTLE AND TRUEAUDIO A PART OF THE PS4 OR SHOULD I GO AHEAD AND CAST MINE INTO THE ABYSS NOW?" LUCKILY, AnandTech answered my question here: http://www.anandtech.com/show/7513/ps4-spec-update...
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
75 Comments
Back to Article
blanarahul - Tuesday, March 18, 2014 - link
Ryan, why don't use measure GPU power consumption like W1zzard from TechPowerUp!.Ryan Smith - Tuesday, March 18, 2014 - link
On this article or in general?In general we do a system wide measurement because we use a closed testbed, which means it's impractical to tap to measure the individual lines.
On this article in particular we were only doing a (relatively) short overview on performance. There wasn't time to look at much else.
TheElMoIsEviL - Thursday, March 20, 2014 - link
"Ryan, Why don't you talk about how great nVIDIA is and how much AMD sucks? I mean I realize that when AMD had better power consumption, I cheered on the Geforce GTX 480, because I'm dishonest like that, but now that nVIDIA has better power/heat levels I have decided that these things matter"Is how I interpreted your post blanarahul.
Le Québécois - Tuesday, March 18, 2014 - link
When can we expect a full article/review on Mantle? So far we're seen preview for Thief and BF4 but not full detailed 20 pages or so review on the thing. I use it on my 7970 GHz and it seems fast and stable but it could also just be my imagination tricking me with a "placebo" effect.nathanddrews - Tuesday, March 18, 2014 - link
Definitely should be some slower CPUs in the testing. Pentium G-series, maybe a C2D/Q. I've seen claims of 50-to-300% gains on forums (with numbers to back it up). for max, average, and minimum frame rates. Even AMD's press coverage made it known that the fastest CPUs wouldn't gain much (maybe 5%), so it would seem fitting to run tests using low-to-mid range hardware.rms - Tuesday, March 18, 2014 - link
Definitely. On my 6core phenom2 & 290X, I doubled my minimum framerate, and nearly doubled the average. Why would the only graph be with a high-end overclocked intel cpu that shows minimal gains?Ammaross - Tuesday, March 18, 2014 - link
I would think an AMD APU akin to the one(s) found in the XBone and PS4 would be a REQUIREMENT for any and all games testing (at least for PC-ports).mikato - Thursday, March 20, 2014 - link
I agree. Using a very high end CPU and just cutting cores and clock rate doesn't give us much hint of the point behind Mantle.Ryan, you wrote the Mantle portion like you expected everyone to have a high end CPU. I know you tested for lower CPU performance, but that's the angle of the article. See below...
"The gains at the high end aren’t worth writing home about, but since we need the CPU to churn out a fairly high framerate regardless, there’s a much greater opportunity to benefit from Mantle on lower end Intel CPUs and AMD’s CPUs/APUs."
So you're thinking is "does Mantle help, and how much?". That question doesn't cover the whole picture. Mantle may lead to purchasing decisions. Someone could buy a lesser CPU and still get close to higher end CPU performance because of Mantle. What about that? I agree that including some lesser CPUs in the testing would really be needed for this as well.
I also have to say, what a lovely sound demo!
savage.r - Friday, April 4, 2014 - link
Of course it should, todays games are made for todays HW and API, means mantle cannot bring performance improvement if game itself doesn't need more. Only way is to use much slower IPC and more cores, still most improvements will see when optimized games will be made from ground up, like star swarm demo, I have 300% FPS there in RTS view. PS4/XBOX uses 8core jaguar architecture, same is used in ultrathins and tablets! CPU with few times higher TDP/ IPC and even less cores, cannot be even remotely comparable to that. But it proofs that today's CPU have much more performance than we really need since all todays games cannot optimally use more than 1core anyway.Wreckage - Tuesday, March 18, 2014 - link
Mantle is still beta and has yet to be shipped with a game. Give it a few yearstcube - Tuesday, March 18, 2014 - link
There are 2 launched games with mantle: bf4 and thief... what are you talking about?Ammaross - Tuesday, March 18, 2014 - link
Note "Shipped" as in, designed fully with Mantle in mind, not post-patched to toss in one or two effects. Granted, Thief utilizes the GPU speedup quite nicely.TheElMoIsEviL - Thursday, March 20, 2014 - link
*facepalm* @ people who just don't get it.You don't "patch" a game for a new API. You write an entire separate path. Hence why it takes time and man hours.
Whether a game ships with Mantle (or not) doesn't matter. What matters is how much time the developers spend optimizing the Mantle path (code).
TheElMoIsEviL - Thursday, March 20, 2014 - link
I'd like to see Crossfire tests with Mantle and a powerful CPU (say a 4930K or 3930K @ 4.5GHz or something).The Von Matrices - Tuesday, March 18, 2014 - link
I can understand your dissatisfaction with the Thief reverberation, I agree that it does sound unnaturally loud. However, I disagree with you in that I also think the Tuscany demo sounds unrefined.I would expect in 2014 a dedicated audio processor could account for the noise blocking effects of walls and the noise transmitting ability of open windows. The fireplace is unnatural - it's roaring loud inside the building but as soon as you leave the building that is full of windows it can't be heard. And then even when you go near a window at around 0:34 the fireplace is still silent.
In my opinion, all this demo did was brought back feature parity of games from 1998. I would have expected more from a company showcasing audio 16 years later. Are my expectations set too high?
edzieba - Tuesday, March 18, 2014 - link
To really push audio simulation forward needs more than just a bit of convolution and managing not to screw up the spatialisation with a HRTf slapped on top. This needs a lot of different techniques to come together at once, something that isn't too easy to meld into your GPU (particularly as audio processing requires a few streams of audio processed intensively, rather than the lots of simple parallel operations GPUs are good at in general), meaning three's not much incentive for Trueaudio to implement something like GSound (https://www.youtube.com/watch?v=buU8gPG2cHI or other environmental audio simulation engines.WagonWheelsRX8 - Tuesday, March 18, 2014 - link
Great article, and really digging informative comments such as this one. I am old enough to remember the first push for environmental audio by companies such as Creative, and it's something you don't really think about but seems (at least from my perspective) to have fallen by the wayside. It's a little depressing we haven't really experienced much improvement in the audio space in so long.That GSound demo video you linked was impressive, and gives me hope that somewhere sometime a game engine or hardware company will integrate audio raycasting.
milli - Tuesday, March 18, 2014 - link
Actually, in 1998, there were better audio solutions that TrueAudio. A3D! AMD should have bought the patents from Creative and brought A3D into the 21st century.http://en.wikipedia.org/wiki/A3D
banvetor - Tuesday, March 18, 2014 - link
The Aureal vs. Creative story is a good example of why Creative is where it is today, instead of being a major player. I remember my first Creative Multimedia Kit, with a 2X CD-Rom, a SoundBlaster 16-bit sound card and speaks (if I remember correctly).I guess they deserve what they've got...
Uvaman44 - Tuesday, March 18, 2014 - link
Ahh yes A3D was ear-candy.. Bastards at creative killed it... at the time Creative's implementations where much worse ( sometime after A3D was murdered, I purchased a creative card) I have never purchase any creative card again, ever.. on board sound for me, whatever crap it is.nathanddrews - Tuesday, March 18, 2014 - link
When EAX worked, it worked well. I still break it out for playing Halo PC using Alchemy. The advantage it brings in hearing approaching footsteps and chambering rounds is astonishing. I can navigate some maps almost entirely by sound. Of course, there's always the hilarity that ensues when the physics freaks out with collision detection and a warthog starts bouncing half in a wall... the audio just farts relentlessly.Come to think of it, A3D had a demo that was nearly identical to this Tuscany demo, I remember playing with it and thinking it was neat.
nathanddrews - Tuesday, March 18, 2014 - link
Ha! Found it!http://youtu.be/-oSlbyLAksM
ozzuneoj86 - Tuesday, March 18, 2014 - link
Hah! I was thinking of this too when I watched the Tuscany video. I remember playing with this on my Aureal Vortex 2 SQ2500 Superquad back in 2000. Which reminds me... what is it with old sound cards and ridiculously long product names? My brother had an Aztech Labs Sound Galaxy Washington 16A soundcard as well... how did they fit it all on a box?DroidTomTom - Tuesday, March 18, 2014 - link
Wow! I was floored by that. I felt like I was immersed in one of those adventure rides at a Theme Park. Too bad no games use anything close to this today. That was way better than the demos from my SB EAX cards. And blows away the Tuscany demo too.milli - Tuesday, March 18, 2014 - link
I'll quote wiki:'A3D uses a subset of the actual in-game 3D world data to accurately model the location of both direct (A3Dspace) and reflected (A3Dverb) sound streams (A3D 2.0 can perform up to 60 first-order reflections). EAX 1.0, the competing technology at the time promoted by Creative Labs, simulated the environment with an adjustable reverb—it didn't calculate any actual reflections off the 3D surfaces.'
Well EAX (even the newest) never got this killer feature. That's why A3D worked so well and consistent.
SilentSin - Tuesday, March 18, 2014 - link
Half Life was *the* killer app for A3D, maybe AMD can bring TrueAudio alive with HL3 ;) Way back with HL1 the framerate did take a hit from enabling A3D but it was absolutely worth it. Playing beta CS with headphones and A3D was almost like cheating, you could easily track players through walls. I think I have my vortex 2 stashed somewhere in my closet..TheElMoIsEviL - Thursday, March 20, 2014 - link
A3D had a horrible CPU utilization rate. I remember having my frame rates halved and I was using either my Voodoo 2 12MB SLI setup or my Voodoo 3 3000 16MB back then.nathanddrews - Tuesday, March 18, 2014 - link
While EAX doesn't do all the calculations, it adds custom DSP to in-game elements and each virtual environment to better reflect (no pun intended) the audio heard by the listener without impacting system performance. Long before ambient occlusion, there was aural occlusion, which EAX brought to the table after they consumed A3D.Probably the best we can hope for with AMD TrueAudio is competition from Creative in the form of A3D's resurrection... but I'd like to think that A3D (like EAX) can be adequately surpassed by modern technology.
risa2000 - Tuesday, March 18, 2014 - link
A3D was unbeatable in true HRTF processing when one was wearing headphones. I had Aureal Vortex 2 card and in Quake 3 Arena (before the patch which removed A3D sound) and in Half-Life and also Deus Ex it was like having third eye. A3D did occlusion and reverb processing of the sound sources on the scene geometry. I have never heard anything close enough since then in any game.As someone who also implemented A3D support in some graphics app, I can confirm that API was quite similar to OpenGL philosophy. The scene could be passed to OGL and A3D in very similar way (back then i.e. in 2001). There was however a drawback in CPU utilization (15-30% at that time), which today with current CPU power and resources could probably translate into negligable numbers.
So I do not think we need specialized HW for this, just standardized SW API would be enough and with current CPU, even bringing in old A3D API would definitely move the 3D sound forward.
frenchy_2001 - Tuesday, March 18, 2014 - link
The reason A3D worked so well was the 3D model of the scene they had to pass to the card.The penalty paid was mostly for the transfer of those models into the sound card memory, which happened over PCI bus (slow). As scene complexity grows, so would this cost. You also would need additional memory on the sound card.
edzieba - Wednesday, March 19, 2014 - link
You can probably cheat by using collision volumes rather than the actual geometry (or by grabbing the lowest available LoD).Wwhat - Wednesday, April 2, 2014 - link
With by modern standards ancient GPU and CPU hardware and a vortex2 card enabling A3D did not impact games noticeably, except for having the great sound. You can talk theory all day long but the fact remains it's incredible that so long ago you had a top notch soundcard and soundsystem and after all the years they can not catch up to it. And then dare to claim it's because of hardware limitations? That's insane and you must be delusional to convince yourself of that.shtldr - Wednesday, March 19, 2014 - link
AMD already implements a D3D API, why not simply take the geometry passed to GPU and use it also for sound rendering at the same time? The only problem is that you don't get the "textures" of the geometry as all you get is graphical textures which do not express properties of the material with respect to sound waves.That could be solved with some very simple extension to D3D (or Mantle) which would allow one to pass sound properties of the material together with the graphical texture.
Wwhat - Wednesday, April 2, 2014 - link
If they keep working on the project I guess that in another few years AMD will reach that level shtldr, although no doubt it will only work with the then current model graphics card I bet.For some reason many developments seem to deliberately go in small steps when they could easily leap since the concept is a repeat of the past and rather obvious.
TheElMoIsEviL - Thursday, March 20, 2014 - link
A3D was not that great. I still have my Aureal Vortex cards. They really were not that great at all. In fact my EAX cards, at the time, produced better positional audio (SB Live!). The problem with creative, back then, were bad drivers and compatibility issues (especially with VIA chipsets).Wwhat - Wednesday, April 2, 2014 - link
Uhm.. you are either a liar or an creative employee. And I dare say that because I know for a fact that is nonsense.bugnguts - Wednesday, March 19, 2014 - link
I agree with you that Tuscany demo is by no means a leap forward since 16 years ago it is a leap forward in what is found today. I could easily pick out exact locations of sounds locations and distance. I've tried this on games like Guild Wars 2 or Skyrim and though I can kinda get a since of directions it is not nearly as clear as this demo.TheElMoIsEviL - Thursday, March 20, 2014 - link
Fireplaces will resonate within a building but not outside. What you demand is unrealistic. Sorry bud but you just don't know what you're talking about. There are high frequency sounds and low frequency sounds. High frequency sounds carry themselves over longer distances. Just an FYI.futrtrubl - Saturday, March 22, 2014 - link
I think you mean low frequency sounds travel further. They also travel through obstacles better.mayankleoboy1 - Tuesday, March 18, 2014 - link
I see that in the benchmark, there is an option to use 32bit executable. Can you do a test of 32 bit Vs. 64 bit executable comparison ?tynopik - Tuesday, March 18, 2014 - link
joying the fray -> joiningbut more on that latter -> later
also lots of places could have used some more commas:
launch title for Mantle Thief -> Mantle, Thief
along with Thief AMD -> Thief, AMD
unlike Battlefield 4 Thief -> Battlefield 4, Thief
Wwhat - Wednesday, April 2, 2014 - link
Sites should make separate closed comment section for such help tynopik, and then when they spot a reliable person assisting like you, give them access to it.That way they have free proofreading and improve their copy, to everybody's benefit.
Askilbrei - Tuesday, March 18, 2014 - link
I would be interested in what this would do for a low budget AMD cpu paired with a medium budget GPU.LarsBars - Tuesday, March 18, 2014 - link
+1It makes sense that AMD would use all of these "offload from the CPU" tricks when its CPUs are currently toward the bottom end of the pack.
TheElMoIsEviL - Thursday, March 20, 2014 - link
No, it really wouldn't make sense since AMD does have the extra CPU cores, in DX games, that generally remain idle. AMD do have the extra CPU resources available for audio.Problem is Direct3D and how much it sucks donkey nuts.
bsim500 - Tuesday, March 18, 2014 - link
Out of all the games that could potentially benefit from this, Thi4f is absolutely top of the list but has pretty much already flopped for a variety of reasons, ironically, one of which is a pretty crappy sound engine that still can't match a 1998 Looking Glass Studio game... Despite its epic franchise's history, Thi4f has an awful sound engine. Even the original game managed to more accurately replicate the effect of attenuation over different surfaces, etc. "Thi4f" somehow manages to screw it all up so that solid walls can make no difference in volume / frequency to a sound. It's also bugged in that one conversation 30ft away can jump in volume to match one 2ft away. All surfaces except glass & water now sound the same, etc. It's nothing like the stone vs wood vs tile vs moss vs grass vs metal, etc of the original. They can't even get the default music vs speech balance right with people complaining of not being able to overhear conversations in many sections.TrueAudio is promising, but it won't fix flawed psycho-acoustic programming in the core game which is the #1 key issue of "what sounds real". More audio channels won't make a game sound better unless devs spend a lot more time & money on realistic subtle sounds and learning to position & mix them well rather than just "slap more reverb on more surround sound explosions". 96-192khz sampling rates are pointless for everyone not sucked into audiophile snake-oil, etc.
I don't doubt that TrueAudio may provide some great effects in demo's, but really game devs aren't going to put in a lot of effort to perfect it since audio typically takes a back seat to the "Oooh! Shiny!" GFX obsessed brigade. After all, we're talking about the same devs who churn out games as quickly as possible and release them in a half bugged state and use "Day One Preorder" gamers as glorified beta testers just to save a few bucks on testing...
Many new technologies sadly end up unrealistic because the devs deliberately exaggerate them as a "showcase". With GFX we've had unrealistic bloom, DOF, etc, and with audio, usually they put in too much echo, and have limited ambient sounds for budgetary reasons. Think of real life - you often hear over 200 ambient sounds in the space of an hour. Now think games - you're lucky if you get 20 throughout the game. and the few you get are often overly loud or "crammed" into too small an area. The issue isn't really "GPU offloading", it's the psychology of developers.
bsim500 - Tuesday, March 18, 2014 - link
And just to add - if anyone's in any doubt over the attitude of developers making things "more real at any cost", with Thief 4, they took the 11 light gem visibility levels of the original, and dumbed it down to just 3 levels : "pitch black", "as visible as wearing a fluorescent yellow jacket in broad daylight", and "half-way between them"... It ain't the engine's that need over-hauling, it's the pervasive dumb-it-down-for-the-casual-audience attitude...velis - Tuesday, March 18, 2014 - link
I would disagree with your assessment: game devs already have to put A LOT of effort into sound, but ultimately it all comes off as cheap and useless crap because the audio engines themselves are crap.For a classic reverb sample: firing a gun in open grassland would normally generate little to no reverb while the same gun going off in a super-sized cathedral would generate tons of it.
The difference between a good sound engine and a bad one is that a bad sound engine gives you option to use reverb (and games today do when they think you're in a cathedral) and a good sound engine will take the sound source and run it against scene geometry. That way you'd still get lots of reverb for gun going off inside the cathedral even if you are standing a substantial distance away from the entrance.
The first problem with the above "good" audio engine is that each piece of geometry now also has to have all the sound properties (reflection / absorbtion / scattering / whatever) just like it now only has light properties. The second problem is that you need to push the geometry to audio processing unit as well as to graphics processing unit. What TrueAudio could offer is elimination of this dual transmission since this is supposed to be processed on the gfx card simultaneously.
Panzerknacker - Tuesday, March 18, 2014 - link
Nice article, I understand there is no time yet for a very big article because there are only 2 games supporting Mantle but good to see it making quite a difference in some situations. About TrueAudio, I agree there are very few games with outstanding audio, I hope this will improve. Ryan, have you ever played the original Bioshock? I think that game actually has the best audio of all games I've ever played, from a technical standpoint. I played it with a Creative Audigy 1 soundcard and 5.1 setup and the combination is absolutely astonishing. Maybe with headphones it will sound equally impressive.krumme - Tuesday, March 18, 2014 - link
Damn sad news for sound.Thanx for the subjective impression. Thats what matters. Hopefully in the future ps4 sales can drive devs to use this tech better.
How many play consoles with headphones?
HisDivineOrder - Tuesday, March 18, 2014 - link
The irony is they were so afraid that you wouldn't notice the reverb effects, they amped them up too high and you definitely noticed them......too much.
But listening and perception of sound is so subjective that they'll never get it right for everyone. I wonder if they'll get it right for the majority.
That said, the feature is nice to have. I'd like to think nVidia and Intel will follow suit. Matching what consoles have done with regards to audio processing would be nice for the ports that are coming.
Though headphones gamers are... in the minority.
willis936 - Tuesday, March 18, 2014 - link
After reading this article and the comments I'm convinced another piece of hardware isn't the solution. CPUs are already unbelievably powerful DSPs. If they nullified the need for dedicated audio hardware 15 years ago think about where they are now. Good audio engines can do these fancy things like terrain mapping for reverb iff the developer can tag the "type" of terrain each surface is. We're all sitting here with hefty CPUs idling the majority of the time and GPUs that are maxing themselves out. Why not implement an exotic audio engine licensed from some company whose only purpose to make exotic audio engines and forget the hardware and all the headaches that come with it? A DSP won't be a fourth as powerful as a 4670's idle time during gameplay and for those still on Q6600s there's always the "medium quality" option.All that said I don't know of any exotic audio engines or if they even exist, but they should and AAA games should be using them. Idk if it can ever be abstracted so it's, as always, up to the devs. The big deal about trueaudio is that it draws attention and maybe the devs will start paying more attention/money to the engines and implementations.
WaltC - Tuesday, March 18, 2014 - link
The Tuscany demo is one example of what can be done; here's another:https://www.youtube.com/watch?v=nKnhcsRTNME
I've never seen a Creative product do these kinds of 360-degree surround effects in normal stereo. Just set up 5.1 or 7.1 and then plug in a pair of stereo phones--it's really, really poor...;) Most of the sound is just lost. This is so cool because it happens with normal stereo phones, imo.
Sound in individual games is just like graphics--all games are not equal. TrueAudio support is just ramping up. The Tuscany demo and the one I linked above reveal what's possible--just *try* doing those on a normal cpu...;) The reverb is small pototoes--its getting a 360-degree sound stage from stereo that's the really impressive attribute.
wurizen - Tuesday, March 18, 2014 - link
"I for one find 24fps material jerky, blurry, and blatantly unnatural."that's totally the opposite reaction of filmmakers since the advent of the film camera, including filmmakers with giant statures like kubrick to name just one. and not just filmmakers, but cinematographers who've worked for decades (perhaps their entire lives) looking thru that tiny viewfinder and seeing 24 fps. and my bet is that it didn't look unnatural to them. or jerky.
jerky, unnatural and those other adjectives you use did not crop up until video games. just to make it clear her,e buddy!
Antronman - Monday, April 7, 2014 - link
24fps is actually the most optimal fps as far as realism goes.For games, more is desirable because you want faster responses from the game. And also because of refresh rate crap.
24fps is as close to how our eyes see as is feasible today.
maecenas - Tuesday, March 18, 2014 - link
That Tuscany villa demo has the same problems as most games these days - the underlying recordings aren't good. For instance, that church bell sounds terrible. There's only so much you can do with dsps and processing if what you're starting with isn't good.samal90 - Tuesday, March 18, 2014 - link
Would be nice to have some benchmarks with the Kaveri APU. I think Mantle would be very useful in that segment. Low details, good kaveri APU and Mantle should show good gains in my opinion. Why not trying it out Ryan?kyuu - Tuesday, March 18, 2014 - link
Agreed. I'm quite curious as to the impact Mantle will have for AMD's iGPUs.rish95 - Wednesday, March 19, 2014 - link
Aren't they memory bandwidth limited, rather than CPU/GPU? I doubt it would help much.jgstew - Tuesday, March 18, 2014 - link
I'm impressed with the virtual 3D sound in the tuscany demo.TheManko - Tuesday, March 18, 2014 - link
The reason TrueAudio in Thief sounds unnatural isn't because it's "overdone", it's because it's not working properly. In the example clip (and other youtube videos I've watched) you hear a doubling effect which isn't supposed to be there. When you hit the wall a sound is generated which bounces off the wall with a slight delay, so you hear the same sound twice. In real life this also happens, but it's so fast that you can't hear the reflection when you're right next to the sound source. So what you're hearing is an obvious problem with the implementation of TrueAudio since these kinds of delays aren't supposed to be part of any realistic simulation of sound space.You claim that the software implementation of Convolution is broken, but it isn't. I tried it myself just now and it's working just fine, but it's subtle. Too subtle to be realistic, and certainly for the average player to notice. There isn't too much reverb with the current implementation, in fact there isn't nearly enough. The only thing that jumps out at you with the current implementation is the bugs.
Ryan Smith - Tuesday, March 18, 2014 - link
"You claim that the software implementation of Convolution is broken, but it isn't. I tried it myself just now and it's working just fine, but it's subtle."To be clear then, you're hearing a difference between Off and Software? I'm not hearing any difference in our test sample or on play-throughs, so I'm curious if you're hearing differently.
Though it should be noted that both Software and TrueAudio should be producing the same results. There is not supposed to be a difference between the two.
TheManko - Tuesday, March 18, 2014 - link
It's subtle, but definitely there. When I started the game I noticed the lightning sound in the main menu producing a reverb it normally doesn't. The game had defaulted to software as I have an Nvidia card. I had a save file in chapter 5 where you're in an insane asylum. Banging on the bars produced a distinct echo in the room, while turning convolution off removed any echo. Similarly when exploring the city I noticed echo effects on sounds like distant crows and dogs. It was hard to hear on NPC dialogue as they're unnaturally loud and make it very hard to focus on any reverb decay. Some areas don't appear to have reverb and sound like your example clip. Maybe it's being simulated, but it's to such a low degree in volume to make it inaudible.mikato - Thursday, March 20, 2014 - link
Nice job pointing this out. I hope it gets a look.gamoniac - Tuesday, March 18, 2014 - link
Thank you, Ryan. The Tuscany demo is great. I initially put my headphones on with the left and right sides flip-flopped. That turns out to be a good negative test, since I immediately felt disoriented. I rectified my headset and fell in love with it. For years, we had to rely on on-screen visual cues in FPS to tell where bullets come from. I hope this technology will allow us to truly immerse ourselves in a firefight! Perhaps, Ryan, you can set up a scene like that for comparison?nos024 - Wednesday, March 19, 2014 - link
Mantle + Kaveri 7850? I am more interested in those benchmarks.junky77 - Wednesday, March 19, 2014 - link
Thanks for this review!As other here requested - a minimal numbers will be nice. Also, the distribution of frame delay time / rendering times. Mantle might also result in a smoother gameplay that way..
Pantsu - Wednesday, March 19, 2014 - link
I did my own tests with 4,4 GHz 2500K and 280@1100/1500. At 1080 max settings with no SSAA I got 67.1 FPS with DX11, and 77.7 with Mantle, so there was definetely some benefit even for a high end system, but with high SSAA the difference was 49.9 vs. 54.7. I also tested eyefinity, but gains were marginal with ssaa off, and high SSAA broke the memory barrier. Personally I'll keep playing the game at 6048x1080 with DX11 CrossFire and no SSAA since the game does have occasional issues with SSAA even at low when using Eyefinity. Hopefully they'll get the support for CF in Mantle before I get tired of the game. I wouldn't mind using high SSAA, but then again it doesn't seem to fix all the aliasing.junky77 - Thursday, March 20, 2014 - link
can you feel a difference in a more borderline cases at 30FPS? does it look like Mantle results in a smoother gameplay?DanaGoyette - Wednesday, March 19, 2014 - link
One of the most aurally amazing games I've heard, using EAX 5, is Uru: Ages Beyond Myst. (You can try Uru Live for free.) While the game isn't the greatest, the audio is amazing -- walk through caves, and you'll hear the acoustics gradually change, and sound sources get properly obstructed/occluded.My ideal audio would be that sort of detail, but with the wave-tracing sort of thing that Aureal hardware apparently did.
I hope the enhancements won't be restricted to just headphones. Even with 5.1 speakers, there's definitely room for improvement.
TheElMoIsEviL - Thursday, March 20, 2014 - link
The quality of "g33ks" has gone down significantly over the years. Most people calling themselves g33ks (or commenting on tech articles) don't know anything about the topics they're covering.I thoroughly enjoyed this article because it clearly pin pointed the shortfalls and the positive aspects of both Mantle and TrueAudio.
Also since 2004 (around then) I have opted to not take many comments seriously as nVIDIA has been hiring agencies to make fake posts disparaging competitor products (heck the NSA now does this as well).
GreyMulkin - Thursday, March 20, 2014 - link
THE REAL QUESTION IS "ARE MANTLE AND TRUEAUDIO A PART OF THE PS4 OR SHOULD I GO AHEAD AND CAST MINE INTO THE ABYSS NOW?" LUCKILY, AnandTech answered my question here: http://www.anandtech.com/show/7513/ps4-spec-update...formulav8 - Thursday, March 20, 2014 - link
Stop reviewing Mantle with a high-end cpu already. You well know that lower-end CPU's is what would most likely benefit the most (if anything would).formulav8 - Thursday, March 20, 2014 - link
If I remember correctly on what AMD mentioned of courseHashh - Tuesday, April 15, 2014 - link
Start testing with cpu's like FX6300, i3 4130, G2130, etc.