AMD Added Noise Suppression To Its new 22.7.1 driver

I don't necessarily need this feature because I use Modmic, but for my rare webcam calls this can surely be useful. This is useful even for headset owners since you can be much more lenient with your input sensitivity if you happen to have a loud keyboard or other noisy things that get caught. Perhaps it even can get rid off static background noise. I'd like to try this out as an experiment, but sadly the recent driver mess keeps me on the 22.5.1 driver yet.

 

I'd much rather this feature from AMD or Intel than Nvidia, simply because the iGPU could be used to do this processing rather than the dGPU. Doesn't mean they're going to do that but at least the potential is there.

 

If it works on my RX570, I'm definitely going to use it.

 

Does not show up on my RX580, the page is dark.

 

Maybe - would have to compare the power cost vs effectiveness of the noise suppression. Idk if you or anyone have used RTX voice lately but it's kind of nutty - i can literally run a lawn mower directly behind me and talk and it's near perfect.. and it only uses like 20-30w from what i've seen. Discord/Microsoft noise suppression is no where near as effective although I don't know what those cost in power.

 

True, and while I haven't used it, I'm well aware how good it is. But, GPUs are often the bottleneck in gaming, so if I were into streaming (which I'm not), I'd rather not further use up my rendering GPU's resources for a nice-to-have feature. If Intel or AMD can come up with something half as good as Nvidia, that's still going to be pretty good, and if it can run on an iGPU then that puts use to hardware I already paid for that is otherwise doing nothing while playing a game.

 

Does the RTX Voice implementation eat performance in games or does it run on the tensor hardware? If it does lower performance, how AMD's implementation could be any different, unless it runs on the CPU. @Denial here talks only about power consumption.

 

I think (and I may be wrong) that it does use the tensor cores, but games that depend on them (like with raytracing or DLSS) could potentially be affected by it.
My point was that if Intel or AMD supports such a feature, you can use your dGPU strictly for the game and your iGPU for mic noise suppression. That way, you're using idle hardware that can't otherwise be used for the game, and your GPU gets to devote all its resources to the game.

 

When Nvidia 1st launched RTX Voice, it ended up working on even 900 series cards, so it wasn't actually relying on any AI components. It supposedly had a higher load on non-RTX cards though.

 

I'd rather not have a GPU doing audio work at all.... Let the audio chip makers produce an actual audio processor to handle such a feature.

 

Any chance of that being a thing likely died when Microsoft killed off hardware accelerated audio.