GPU Audio Release Free Modulation Bundle

[GPUAudio]

Hey,
We're pleased to announce the long-awaited Modulation Bundle is out today! It contains three GPU powered plugins, and forms part of a full suite we'll be rolling out over the coming weeks. Download them here



As with our early-access plugin, the Modulation Bundle will be incrementally updated with new features added as we go -so look out for updates, improvements, and refinements in the near future. All feedback on the collection is valued and gratefully received on our Discord channel where we provide support directly to our community.

Before the end of the year, we'll be releasing an SDK for third-party developers, allowing for the creation of audio effects, systems, and technologies that break free from the constraints of conventional audio. With multiple contracts signed, we have already announced the first 3rd party plugin will be made in conjunction with the acclaimed Spatial Audio company Mach1. Sign up for the SDK here.

Catch the GPU Audio team in the real world - we'll be speaking at ADC in London (sign up to join at Audio.dev), AES in New York, and will be at Adobe Max in LA.

If you haven't tried our free Early Access FIR Convolution Reverb, you can download all of our plugins from gpu.audio! But for now, get your hands on the Modulation Bundle and try out these revolutionary plugins for yourself.

Enjoy and let us know what you think!

Chris

[Trancit]

I am sorry and I don´t want to diss you but as I am highly sceptical about the performance increase of this technique I have a question:

If GPU powered plugins could bring so much performance increase to this world, why do you just fiddle around with processing types which hardly take any CPU anyway...

I mean Chorus/Flanger/Phaser etc. I have already to put in at least 10 devices of these types into the project before the impact is even measurable...
The most convolution reverbs behave nowadays the same and are extremely low on CPU...especially one which is so limited like yours...

How shall one know, that your technique even brings something positive on the table and the GPU impact isn´t only the rendering of the GUI??
And even more if I have already to insert hundreds of these plugins to even think about a bit of optimization (which in a real world example wouldn´t be realistic anyway... who needs hundred Choruses or hundred Convolution Reverbs), why should I even bother to install your plugins... it´s not very realistic they could bring even the slightest bit positive on the table

Why not making something which would be a real CPU hog and demonstrate the performance gain with something like this??

I am sorry but as long as this doesn´t happen your promises are for me just hot air...
You cannot prove the performance gain and I cannot prove the opposite...

[GPUAudio]

If GPU powered plugins could bring so much performance increase to this world, why do you just fiddle around with processing types which hardly take any CPU anyway...

I am sorry but as long as this doesn´t happen your promises are for me just hot air...
You cannot prove the performance gain and I cannot prove the opposite...

I think it's good to come at this will a critical mind and I understand why you may be thinking that way.

This technology has been in development for around 10 years and we now have it patented - these first releases (which are free) are designed to be fun, familiar tools which could be seen as diagnostic / benchmarking tools for people who want to try GPU Audio processing.

Look into GPUs and you'll find they're incredible easy to scale (unlike CPUs), offer parallel processing (unheard of in audio) and the vast processing power (as witnessed for bitcoin mining) can be used for live machine learning in audio, like a live iZotope RX for example.

This article by Attack mag goes into some detail: https://www.attackmagazine.com/features ... o-latency/

Developing software for multiple types of hardware is no mean feat (especially NVIDIA, AMD and Apple) so rolling out these useful but reasonably simple plugins helps the devleopment progress.

NVIDIA and AMD would not be officially partnering with us if we were just blowing hot air. They can see the potential, and perhaps you will one day - I hope so. When the third party developers get the SDK, you'll start to see some really next generation stuff happening.

For now, enjoy the freebies, keep asking questions, and have fun making music!

[n9research]

I downloaded and installed this bundle. I run an RTX 2070 Super on Windows 10

I can add modulation effects in the chain, and I can hear that the sound is being effected in some flangey, phasey way, but:

• most of the dropdowns, selectors, buttons, options are not interactable i.e. they don't respond when clicked

• none of the parameters of the effects seem do anything whatsoever to the processing

[jens]

I see some parallelization gains at high instance counts (I started to see the first (minimal) gains north of ~30+ instances) for both the chorus and the flanger, but none for the phaser.

However

a) they are significantly less than the parallelization gains of some well coded native plugins

b) they are nothing when it comes to outweighing the huge CPU consumption of these mod FX to begin with,
as compared to some of the best native modulation effects which include complex analog modelling and what not
and which do not even use 1/10 of the Braingines ones here


That's not to mention weird audio-artifacts (rendering them unuseable) and more odd behaviour...

[jens]

And just to make sure: I still have high hopes for your technology, but so far I didn't see much worth rejoicing over...

[Trancit]

nevermind...

[bungle]

I have done as much research on this as I could (Not owning any of the supported hardware right now) and in all tests I can find, it doesn't even compare to a single added machine on Audiogridder, either in terms of cost or performance.
Obviously it is an infant technology and GPUs are not going to start getting slower, so will still be keeping an eye on it, but it isn't viable right now.

[Jmf1928]

Dude, they’re giving it away for free. Lighten up. If there are no performance gains to be had, they won’t end up selling anything. I think it’s highly unlikely they’d have invested in rewriting a scheduler if they weren’t very confident in the performance capabilities of the tech.

I am sorry and I don´t want to diss you but as I am highly sceptical about the performance increase of this technique I have a question:

If GPU powered plugins could bring so much performance increase to this world, why do you just fiddle around with processing types which hardly take any CPU anyway...

I mean Chorus/Flanger/Phaser etc. I have already to put in at least 10 devices of these types into the project before the impact is even measurable...
The most convolution reverbs behave nowadays the same and are extremely low on CPU...especially one which is so limited like yours...

How shall one know, that your technique even brings something positive on the table and the GPU impact isn´t only the rendering of the GUI??
And even more if I have already to insert hundreds of these plugins to even think about a bit of optimization (which in a real world example wouldn´t be realistic anyway... who needs hundred Choruses or hundred Convolution Reverbs), why should I even bother to install your plugins... it´s not very realistic they could bring even the slightest bit positive on the table

Why not making something which would be a real CPU hog and demonstrate the performance gain with something like this??

I am sorry but as long as this doesn´t happen your promises are for me just hot air...
You cannot prove the performance gain and I cannot prove the opposite...

[SLiC]

I have done as much research on this as I could (Not owning any of the supported hardware right now) and in all tests I can find, it doesn't even compare to a single added machine on Audiogridder, either in terms of cost or performance.
Obviously it is an infant technology and GPUs are not going to start getting slower, so will still be keeping an eye on it, but it isn't viable right now.

Audiogridder require multiple computers to be switched on and networked together, that's a PITA for most people even assuming they have multiple computers (I know I gave up on it in the past).

This has the advantages of using the insanely powerful DSP processing GPUs many of us already have in our computers. It also offers the potential of almost real time processing. In my opinion I think this is far more likely to be the future than networked computer solution with latency compensation...it uses what already there with no set up at all...

[kperry]

I've got an OK - not great but OK - nVidia card (GT 1030) and one instance of the modulation plug-ins takes ~15% of its power. Sure, it may get better if I add more instances, but I'm not likely to *need* enough of a similar plug-in to benefit.

[SLiC]

One someone Like UHE gets Diva to run on this though....

[Trancit]

Dude, they’re giving it away for free. Lighten up. If there are no performance gains to be had, they won’t end up selling anything. I think it’s highly unlikely they’d have invested in rewriting a scheduler if they weren’t very confident in the performance capabilities of the tech.

And because of that it isn´t allowed to ask them a question???

[jens]

I've got an OK - not great but OK - nVidia card (GT 1030) and one instance of the modulation plug-ins takes ~15% of its power.

Sorry to say, but that card is really nothing to write home about and especially when it comes to Cuda - it only has 384 cores - a RTX 2060 has 1920 and a RTX 3080 has 8960...

[kperry]

I've got an OK - not great but OK - nVidia card (GT 1030) and one instance of the modulation plug-ins takes ~15% of its power.

Sorry to say, but that card is really nothing to write home about and especially when it comes to Cuda - it only has 384 cores - a RTX 2060 has 1920 and a RTX 3080 has 8960...

Sure, but it's not an uncommon card for DAW users (since it can be passively cooled and it's about the "best" card that can be so cooled), and it still seems pretty high GPU usage for a chorus/flanger/phaser, when they haven't tickled a CPU for years now.