Playstation 5 [PS5] [Release November 12 2020]
- Thread starter BRiT
- Start date
I didn't mean to dismiss audio in some games. I know psvr has some decent tries, oculus also has put a lot of effort to 3d audio. I just feel there is ways to go there and sonys hw+hopefully sw can take things much further. At least I hope so. Could be it's sonys tv,tv,tv moment but cerny usually is good about technical things and they do manifest in real life later.
One big thing sony is trying to do is HRTF. That would take the shape of head, earlopes and inner ear into account. This potentially allows for perfect audio. The initial thing is they have only few settings available and if you are lucky, that's great. Cerny made some claims on how they could make the setting truly individualized, but will that happen is another story.
The other part of the equation is the compute power provided in tempest engine. There is so much potential to do audio properly. Tempest has roughly same amount of compute available as would be available using all jaguar cores in ps4(not fair comparison). Taking real scene geometry and properties into account rather than having some "echoing big room" setting or "narrow corridor" setting being applied. This is much easier on some small scale vr experience than a full blown big game with many different kind of environments and moving entities.
In my head I'm imagining gran turismo that simulates each car better. Things like engine sound, gearbox, tires,... for every car. Then positions the audio into 3d space and takes the track geometry and other cars into account. And then apply that in vr using HRTF function that is tailored to me. And the car player is driving would be simulated even better taking into account how the car is built and make the audio ... perfect... that would be nuts. I doubt sony reaches that level but I hope they give it a really good try!
Well yes as explained the PSVR has additional audio processors in the connector box allowing for 50 spatial channels if I understood it correctly.
I also had the feeling at times that Uncharted 3 had better audio on PS3 than Uncharted 4 on PS4, or even it seemed that the remake on PS4 didn’t sound quite as good. Now there may be an explanation for it ...
I actually had a quick go at that in Unity as an experiment, using Google's Resonance Spatialiser. The idea was to see if you could make a game that was sound only, where you'd have to find the item. Maybe something like Pacman audio only. But the 3D accuracy just wasn't there. Using the example scene, I tried to locate the sound-emitting object with my eyes closed. I could tell when I was close, and if it was left or right, but I'd open my eyes and see it was some way above me, or below, or just a little off. I'm very looking forwards to next-gen audio.
He actually stated it'll take time, and is expecting a years-long growth. I hope it launches good enough to be a differentiator; it'd be a real shame if the potential was there but the inability to find the users HRTF saw it abandoned.
I suppose combining it with the move controllers or any other vibration capable motion controller could help give you a kind of virtual cane, though of course it would have to be more like a proximity sensor that starts vibrating more as you get closer or something like that.
I expect the five presets will cover most people well enough to be really noticeable (barbershop demo works for most people now with just one setting) and the personalized version is going to be icing on the cake that makes it perfect for all.
If the solution to hrtf was to go to sony store at mall and pay 50$ for headscan would you do it to get perfect HRTF function? Let's assume you would get the datafile in some format that you would physically have. Initially only ps5 would have implementation for importing that data but there would be potential for other "apps" to also use same scan.
I would pay 50$ for that, but I don't think I would go higher than that. At least not without wider support than just ps5.
I would pay 50$ for that, but I don't think I would go higher than that. At least not without wider support than just ps5.
D
Deleted member 13524
Guest
The PSVR breakout box has a Marvel Armada 1500 4K SoC. That's a 4x Cortex A9 together with a GC3000, which according to Vivante it's a fully OpenCL 1.2 compliant GPU capable of 32 GFLOPs FP32 / 64 GFLOPs FP16. It uses 64bit DDR3 @ 800MHz, so the whole SoC has 6.4 GB/s available.
So if the PSVR has a 32 GFLOPs audio processor together with ~6GB/s bandwidth, then assuming the bandwidth proportion is kept the PS5's ~282 GFLOPs audio processor (one RDNA2 CU @ 2.2Ghz) would require (282/32)*6 = 53 GB/s.
Of course, not all GFLOPs have equal effective throughput especially among different GPU architectures, and we don't know exactly how much bandwidth the audio processing in the PSVR SoC really used.
And we also don't known if they used the iGPU cores or the CPU cores on the Armada core, though if I had to guess they used the iGPU.
One interesting fact is that Sony and AMD apparently found a way to make sound processing on this RDNA2 CU much more effective than it could be done on GCN GPUs.
Valve's page on Steam Audio / TrueAudio Next claims that reserving 8 CUs for audio allows for 128 convolution sources. With this single CU, Cerny is claiming they can use "hundreds" of voices.
I wonder if there's more to it than just a RDNA2 Compute Unit with "stripped down caches". Adding some fixed function hardware in it could probably help.
EDIT: The 8 CUs for 128 convolution sources are bound to a 6ms budget (167FPS).
I'm guessing you can increase those 6ms towards 16ms (assuming you can?) and the number of sources can increase accordingly. There's also the fact that not all sound sources will need a 48KHz sampling rate, as voice and sounds in the lower spectrum will require a much lower sampling rate which should have their signal convolution processed faster at a linear rate. Then there's the fact that the RDNA CU is running at much higher clocks.
Cerny's claim of "hundreds of sources" is probably taking into account all these factors.
Last edited by a moderator:
Yup that's the dream and I can't wait.
Yeah a photo of each ear might be enough in the future thanks to machine learning.
Maybe even some head measurements.
HRTF solutions haven't really taken off elsewhere where they've been available for a long time, can console make the difference?
He actually mentioned processing key sounds ("Hero sounds") in higher quality and less important sounds in lower quality.
Yes. It's standard for everyone who buys the machine and available for every game on the machine without needing to be bespoke. We also have a number of key games at the moment where it'll be a game-changer. There's not a single celebrity Fortnite or Apex player who won't be using 3D audio to hear where the opposition is, and that'll sell it to their watchers and ripple out. The industry is finally ready to make this happen.
perhaps yes. I still don't have a definitive reason why the sound market just died out around 2000. IT's not like Atari bust where we can identify what happened. I still have figured it out, I suspect it was largely business related because the technology as you can see, and as @ToTTenTranz points out, requires a great deal of processing. It made sense that we were using dedicated silicon back then to get good sound. It still makes sense to do it now.
With headphones as being the dominant audio setup now, the market for audio is significantly larger. Previously we needed to have 4-5 channel speakers to get that quality surround sound we're all looking for. Now every controller supports a stereo jack.
I'm really happy to see the return of audio, I'm really happy Sony is going this route, hoping that MS announces more on this front.
edit:nvm, seems people did write about it: https://www.tomshardware.com/reviews/future-3d-graphics,2560-2.html
Great nostalgia in the article above.
ooof:
"Do the math and you realize that from 2000 to 2007, the sound card market collapsed by 80% over seven years. The recent stock market crash? That was only a 53% drop during the worst of it, and a 25% drop at current values."
Anyone know if this is still true today? Will it change?:
Audiophiles can argue over the subtle nuances that one DAC has over another in terms of musicality or noise floor, and point to the increasing popularity of HDD-based music servers as proof that PC audio has the potential to offer reference-quality audio. While I don't disagree, the majority of games and consumer applications already sound great with today's sound cards because the software isn't designed to take advantage of high-resolution audio.
No game developer is going to take the time and money to properly record, mix, master, and distribute a soundtrack in 24-bit/96 kHz. In fact, many Hollywood films today are done at 24-bit/48 kHz. For the software developer, the money is better spent on adding new levels or artwork to a game, and the computational horsepower better put into graphics or physics.
Last edited:
I don't understand how it can work outside of headphones. Most living rooms have nasty reflections messing up everything and it's already difficult for most people to just have a good stereo reproduction for music and film. Maybe they'll recommend sound treatment to get the most out of it, and then people will realize how amazing everything sounds with proper room treatment.
Cerny mentioned the problems with non-headphone audio. Basically, it can't be done properly. The moment you have two listeners, you can't process two HRTFs. I think it'll be there for the solo game chiefly, and you'll just have pleasant spatial sound when playing on open acoustic systems, whether that's your 5.1 setup or some stereo sound bar.
Similar threads
- Locked
- Deleted member 7537
- Games Industry
- Sticky