Thursday, September 26th 2013
Do We Really Need AMD TrueAudio?
At its press meet called to unveil its next-generation graphics lineup, which lasted for about 3 hours and 45 minutes, a full one hour was spent detailing TrueAudio, an audio DSP that AMD co-developed with GenAudio, makers of AstoundSound technology; both of which are brand names we'd heard for the first time, and that's coming from someone who's been an audiophile and follower of PC audio technology for 15 years now. Here are a few reasons why the concept of TrueAudio, or at least the way AMD and its newfound GenAudio pals explained it, is unconvincing to the extant of flagging "snake oil." While we did not attend GPU'14, experiencing TrueAudio is irrelevant to our analysis. It focuses on what "yet another" positional audio DSP has to do with GPU, and how AMD is getting ahead of itself with non-graphics features in exactly the same way it criticized NVIDIA of doing, not too long ago.
Back in the 1990s and the first half of the previous decade, discrete sound cards were an integral component of any self-respecting gaming PC, as onboard audio solutions were either too basic (ISA-based CODECs, USB-based AC'97) or sounded like greeting cards. Apart from higher fidelity, discrete sound cards offered proprietary positional-audio and DSP layers, promised hundreds of hardware-mixed audio voices, and what have you. AMD, in its presentation, didn't succeed in explaining how its solution is any different from something Dolby, Yamaha, Creative, ESS, and others haven't already tried.
With Intel's Azalia specification, and acceptable increments in audio fidelity past the 85 dBA signal-to-noise ratio mark with most motherboard-integrated audio solutions, game engine developers spent the better part of last decade innovating positional- and immersive- audio technologies that didn't require specialized hardware. You get the same incredible sound-effects on Battlefield 3 on any hardware, with discrete sound cards only offering better audio quality, but not better effects (positional- or realism). In that respect, we feel that it's quite arrogant of AMD to claim that there has been no innovation in PC game audio since programmable shaders (2001-02). To the contrary, game engine developers like Unreal, CryTek, DICE, etc., worked tirelessly to make effects comparable to three-figure discrete sound cards accessible to all gamers, by leveraging multi-core/multi-threaded CPUs.
What AMD didn't explain to us, is whether use of TrueAudio technology requires Radeon graphics cards to be installed; because if it does, then there's little difference between AMD's approach to positional-audio, and discrete sound card manufacturers', which unsuccessfully tried to make people to buy their sound cards on the "better audio immersion" plank, before switching over to the "better audio quality" plank, upon realizing that "better immersion" doesn't sell sound cards anymore.
What AMD also didn't explain is where "Radeon" fits into all this. It's a GPU alright, and combined with OpenCL, it's capable of general-purpose computing. It didn't explain whether the TrueAudio DSP layer runs on the CPU, or is something the GPU spares some of its computing power on. We're not convinced by the "hear more voices in a game" claim. The way audio is implemented on Windows, an application can stream up to 65,535 "voices" (independent streams of digital audio through a software/inter-process-communication channel), which are then mixed into "what you hear." So the only way AMD can sell "hear more voices," is if it gives developers something that can let them stream those hundreds of voices, while minimizing the CPU cost of processing and mixing them. If not having those many voices in any way steps beyond being just aesthetics (think crowds in a games like Assassins Creed, GTA), and actually affects the gameplay (i.e. not hearing all those voices puts you at a disadvantage in, say, an online multiplayer shooter), then you can bet your bottom penny that developers won't implement it.
In conclusion, even if TrueAudio made genuine improvements in the accuracy of positional audio, which they claim is backed by "25 years of brain research," there are too many factors stacked up against its successful proliferation. Want better sound? Go get a nice sound card and/or better speakers/headsets. Don't count on dinosaur DSPs anymore. There are two components to better game audio, "immersion" and "output quality." Your game developer already took care of the "immersion" part, and today's premium sound cards and speakers/headsets look after "output quality." Not much room for AMD to squeeze into this neat little arrangement.
With Intel's Azalia specification, and acceptable increments in audio fidelity past the 85 dBA signal-to-noise ratio mark with most motherboard-integrated audio solutions, game engine developers spent the better part of last decade innovating positional- and immersive- audio technologies that didn't require specialized hardware. You get the same incredible sound-effects on Battlefield 3 on any hardware, with discrete sound cards only offering better audio quality, but not better effects (positional- or realism). In that respect, we feel that it's quite arrogant of AMD to claim that there has been no innovation in PC game audio since programmable shaders (2001-02). To the contrary, game engine developers like Unreal, CryTek, DICE, etc., worked tirelessly to make effects comparable to three-figure discrete sound cards accessible to all gamers, by leveraging multi-core/multi-threaded CPUs.
What AMD didn't explain to us, is whether use of TrueAudio technology requires Radeon graphics cards to be installed; because if it does, then there's little difference between AMD's approach to positional-audio, and discrete sound card manufacturers', which unsuccessfully tried to make people to buy their sound cards on the "better audio immersion" plank, before switching over to the "better audio quality" plank, upon realizing that "better immersion" doesn't sell sound cards anymore.
What AMD also didn't explain is where "Radeon" fits into all this. It's a GPU alright, and combined with OpenCL, it's capable of general-purpose computing. It didn't explain whether the TrueAudio DSP layer runs on the CPU, or is something the GPU spares some of its computing power on. We're not convinced by the "hear more voices in a game" claim. The way audio is implemented on Windows, an application can stream up to 65,535 "voices" (independent streams of digital audio through a software/inter-process-communication channel), which are then mixed into "what you hear." So the only way AMD can sell "hear more voices," is if it gives developers something that can let them stream those hundreds of voices, while minimizing the CPU cost of processing and mixing them. If not having those many voices in any way steps beyond being just aesthetics (think crowds in a games like Assassins Creed, GTA), and actually affects the gameplay (i.e. not hearing all those voices puts you at a disadvantage in, say, an online multiplayer shooter), then you can bet your bottom penny that developers won't implement it.
In conclusion, even if TrueAudio made genuine improvements in the accuracy of positional audio, which they claim is backed by "25 years of brain research," there are too many factors stacked up against its successful proliferation. Want better sound? Go get a nice sound card and/or better speakers/headsets. Don't count on dinosaur DSPs anymore. There are two components to better game audio, "immersion" and "output quality." Your game developer already took care of the "immersion" part, and today's premium sound cards and speakers/headsets look after "output quality." Not much room for AMD to squeeze into this neat little arrangement.
87 Comments on Do We Really Need AMD TrueAudio?
As for gaming... everything that's an evolution of surround is fine. I believe that in my case, I would make better use of it when watching a movie, but for now there are no movies released for this technology. I believe it's ok to have it, but I was like "I wasn't expecting this to be as important as framerate and quality improvement on the GPU"
I would probs never use the sound output on these cards anyway tbh. And it its probably only marketing anyway.
If things are not worth the effort, this solution will vanish - so what? On the other hand it might help an AMD-AMD system to raise frames, cause today you can get much more frames with an Intel-CPU than with an AMD-CPU, given the same AMD video card. Therefore the video card of an AMD-AMD system has some resources left for some additional computing, lighten the stress on the CPU and gain overall performance.
Actually I also doubt, that this is an exclusive feature. Meaning, nVidia can join in easily by also using this DSP-Chip - if they want to.
They connect with USB and have a controller on the cable that is basically a sound card. With the software it simulates 7.1 sound. And despite what reviews have said about these products, it works pretty well for me.
I can hear behind me to left, below me to right, etc. Seems I'm often headed towards the enemy before others are in multiplayer games.
So this capability has been out there for a while. Not a ground shaking revelation to me.
Even if it is somewhat better, I'd still rather hear about the graphics capability of a graphics card.
...
For anyone wondering, one of the people partnering up with AMD is AudioKinetic, the guys who designed the Wwise SDK. Wwise is already found in about 300 games and is compatible with the Unreal 3 and 4 engines, supporting TrueAudio is a plugin away.
If it IS true, it's going to be very limited and probably why through part of the GPU14 demo the GenAudio guy talked about headphones, "BRTF", and virtual surround on a stereo signal.
As is I use optical out from my MB, which my AVR can only use as Dolby Pro Logic II at best. I'm guessing to get the "benefits" of TrueAudio I'd have to put my Yam AVR in "straight" mode which is only 2.1 PCM.
In other words, it's sounding to me like you'll lose your center and surround channels with TrueAudio unless you can get multi channel HDMI audio working, which has been hit and miss at best throughout the industry, esp for gaming.
So in short, it looks like for those in my position, TrueAudio is only really beneficial when gaming on headphones.
It's not like HDMI audio on GPUs not working is anything new, I've seen plenty of people say the same. One here that DID try using it for games even said he prefers his sound card's DTS, because the multi channel PCM messed up the channel positioning in games.
I would hope that (like some of the other 3D sound standards) the more output channels you have the better it works for more seating positions and can provide even more accurate positioning. It would be nice if some of the attendees at the conference could comment on how the 7.1 demo sounded.
As a side note, I also note that some folks seem to be confused (if that's the right word) between sounds source numbers and output channels. i.e. the 8 or 16 channels in HDMI, this is just the number of supported speakers, you can still have an infinite (if the system supports it) number of sounds being sent via those 8 (or more) speakers. The way GenAudio works, we should actually need less speakers "in room".
The number of speakers should really be dictated by available room space.
What I would be interested in finding out is if GenAudio's methods are capable of sending there "special filters encoded in the output steam" in such a way that e.g. height effects could work with things like PLIIz and Yamaha presence to also utilise actual height speakers to further position sounds. If not, there is no point in getting any output system with more channels than those discretely encoded and for most rooms we can probably stop at 7.x (.4 ideally) ... which would actually be a good thing from a cost perspective. :)
Maybe these are some of the things they are thinking about for GenAudio 2.0. :)
Rooms have unique acoustic signatures and virtual surround mixes never account for the acoustic signature of the room in which they are being played. So while my audio receiver can analyze the room characteristics and adjust frequency response of the speakers to cancel out the room's properties, you can't do that precisely with virtual surround because it assumes you have a specific room shape and frequency response. The only good use I see for virtual surround is in headphones because they do not have to deal with echoes and reverberations of the room. I always see adding more speakers to a room as the better option than going with virtual surround, but I can understand that most rooms don't have the space for a lot of speaker channels so virtual surround can be better than nothing.
On a side note I know I'm in an extreme minority but I do have a 9.1 channel home theater. For those who are not familiar, as of right now there is no native way to use the height channels; the only way to use them is through matrixed 7.1 mixes like Dolby PLIIz and reprocessing of other formats. I have tested the system using 7.1 audio sources with and without the height channels enabled (with audio being split from the main stream using Dolby PLIIz, Audyssey DSX, and DTS Neo:X).
The height channel effect is very subtle in most cases, and I could completely understand if someone tried the comparison to virtual surround and couldn't tell any difference at all. The virtual surround demos do emulate the experience I have had with height channels very well, but I don't know if that's because these height effects are very subtle or if the current lack of discrete data for the height channels reduces their effectiveness. If there was a way to get discrete channel information to the height channels then they would sound a lot better.
I have played a lot of games using the 9.1 system, and the best one I've heard so far was Heavy Rain for the PS3. You almost felt like you needed an umbrella in some of the scenes in that game. Sports games are also good since you really get the effect of a crowd cheering in a stadium and you can hear people screaming from the cheap seats.If you only see stereo on the HDMI output, then that means that the only LPCM (uncompressed) audio stream your device supports is two channel. This does not mean it does not support compressed multi channel formats like Dolby Digital or DTS, just that it does not support uncompressed 5.1 or 7.1 audio. If you have a receiver that has 5.1 or 7.1 channels and only shows up as stereo, then you need to use Dolby Digital Live or DTS connect to compress the audio into a compressed multi-channel format before your receiver can decode it. I've never seen that software used over HDMI though, only with certain sound cards with the right licensing and TOSLINK or digital coax outputs.
Also, if the multi-channel LPCM is messed up in games, I highly doubt it's the video card's fault. Either the receiver has the speakers wired incorrectly or the game is outputting audio for a different number of speakers. Most games don't detect the correct number of speakers automatically and if you don't manually set it the game sounds awful.
From my experience in troubleshooting problems with computer HDMI audio output, almost always the cause is what I described two paragraphs above. The problem lies in the receiver or end device not supporting multi-channel LPCM rather than with the video card.
Anyone remembers for how long they've been bragging about HW accelerated physics? And PhysX is the furthest they could get anything. Lame to the extreme. And with this one, covering only half of the user base does exactly the same thing. It will be an interesting gimmick to talk about for a month and then no one will care about it. pointless waste of everyones resources.
Why no one for the love of god standardizes physics like they managed with graphics and call it i don't know, i'll just use my imagination, DirectPhysics!?. And then do the same for audio if they want a shader accelerated 3D audio positioning and i don't know, call it DirectSound3D again?. Creative Labs stabbed themselves when they made EAX 5.0 locked down exclusive, but it wasn't as bad as it is with single GPU maker side features. With soundcard you still had a choice to pick regardless if you were using NVIDIA or AMD. But if only AMD has this, there is no way NVIDIA will also support it. And vice versa. And who gets screwed in the end the most? Gamers, no one else.
I love 3D sound positioning and advanced audio features, but all this is just one big muscle showing off and nothing else. We will not get anything useful out of it in the long term. Which is a shame because due to lack of standardization, audio part of the games has been more and more neglected despite of what they were saying with all the software audio engines (like the FMOD) which if i'm honest all sound rubbish compared to any EAX 5.0 HD powered title on a proper Sound Blaster soundcard...
However, on a side note, the only game I know to fully support EAX 5.0 was FEAR 2 if I remember correctly, no other games were made for the last version of EAX. However even EAX 5 pales in comparison with Aureal's A3D 2.0, which not even today there are games to match the sound effects in games like Thief, Unreal, DeusEx, etc. And this is only because of basterds from M$ that killed 3D hardware sound.
In multi-channel PCM mode on my receiver, some games the center speaker wouldn't be accurate. Like in Tombraider Underworld and Max Payne 3, the speech doesn't come out of the center at all times, as it should. However you get other sound effects like gunshots and footsteps from the center with the occasional speech from Lara and Max being outputted from center during specific scenes. I put it down to the game, not the hardware. Other games worked fine.
I've had a few DDL and DTS interactive compatible soundcards and the center action has always been correct 100% in all games and situations.Invest in home cinema and you won't care what Creative and Microsoft do.
Yeah, my Yam AVR is only an entry level model that doesn't even have software calibration, so I wouldn't be surprised if it doesn't support multi channel uncompressed audio. Rather than Yamaha giving me the details on it or just saying they don't know though, they just laughed at me and insisted it was something wrong at my PC end. Bunch of douche bags.
I mean I know the average phone support guy may not know these things, but they should be able to find out from someone in their tech dept. How is a customer supposed to make an intelligent decision about how well an AVR supports gaming if he doesn't even know if it's capable of picking up uncompressed multi channel audio from the PC?
@nemesis,
Yeah I was thinking it might be most beneficial with headphones too, but did you notice that the stereo part of the GenAudio demonstration didn't draw nearly the response (none really) that the multi channel part did? I also purposely put my AVR in 2.1 PCM for that part, and it wasn't anything special.
So far they've not given me any reason to want to go out and buy some decent cans like I used to have. There's no telling how much developer support it will have either.
GenAudio should have uploaded it onto their own website so the file could stay intact. Saying that it with DTS Encoding enabled on my system it still sounded fantastic.
This is probably worth a read and echoes/confirms some of the positive thoughts in this thread:
semiaccurate.com/2013/10/09/like-mantle-amds-trueaudio-massive-step-forward/
LGTM! :)