Free GPU Powered Delay and Reverb Bundle!

[GPUAudio]

Hi everyone,
GPU Audio are pleased to announce the FREE Space & Time Bundle - three classic time-based effects for creating space in your music. https://www.gpu.audio/spaceandtime



The Space & Time Bundle includes three fully featured Delay and Reverb Plugins - including the original FIR Convolver, the powerful FIReVerb Reverb unit, and a flexible Delay module. Each plugin comes with a host of parameters and features.

The free Beta Suite is a collection of professional quality audio effects using the immense processing power of your GPU. This latest bundle takes the suite to a total of six plugins, with more to come; including an EQ, Filter, Exciter, Distortion, Gater, Compressor, Limiter, and more.

Have you tried them yet? We'd love to know what you think!

https://www.gpu.audio/spaceandtime

Best wishes,

GPU Audio Team

[jamcat]

Do these take advantage of the M1 GPU, too?

[jens]

Have you tried them yet? We'd love to know what you think!

Like the earlier releases this still uses an absurd amount of CPU - far more than any of my high quality plugins of the same type I own.
Perhaps it taxes my GPU in addition but so far I have not seen the tiniest proof that this technology could lead to less CPU consumption - absolutely au contraire.

So at this point it seems to be rather safe to say all the nay-sayers (with a lot of credible coders amongst them) were right after all.

If you didn't manage to come up with a working proof-of-concept by now then chances are you won't ever, right?

[temnov]

"If you didn't manage to come up with a working proof-of-concept by now then chances are you won't ever, right?"

I'm very interested to hear GPU response.

[ThomasHelzle]

- Tells me on installation that it doesn't run on AMD APUs. It's a new ASUS ROG STRIX laptop with a dual GPU, Simple AMD and a GeForce 3070

- Bitwig doesn't show "GPU Audio" in the plugin vendor list.

- One has to find "Container Plugin" in the list. Amazing name really.

- In Bitwig it has a broken GUI:

GPUAudioGUI.jpg

I run Windows at 125% scale and it seems your framework does not like that.
The place where the mouse is in the image is where I have to move it to get to click on delay in the list, so it's totally off and unusable.

EDIT: Turns out this is a very convoluted thing to fix, but not GPU-Audios fault: One has to enable "Plugin Interface Scaling" in Bitwig Preferences and has to DISABLE "Stretch Plug-in Window to match DPI" in the plugin browser by right-clicking on the offending plugin.
After that, remove and re-load all instances.

- When I manage to select an effect it shows me a "processing error" warning. Is that the AMD APU thing or something else?

- CPU use is very high.

So like the last times I tried your plugins, they fail to work.
I guess one needs a very specific DAW & Hardware setup to make it work.

[Ploki]

Do these take advantage of the M1 GPU, too?

Yeah.

Like the earlier releases this still uses an absurd amount of CPU - far more than any of my high quality plugins of the same type I own.
Perhaps it taxes my GPU in addition but so far I have not seen the tiniest proof that this technology could lead to less CPU consumption - absolutely au contraire.

So at this point it seems to be rather safe to say all the nay-sayers (with a lot of credible coders amongst them) were right after all.

If you didn't manage to come up with a working proof-of-concept by now then chances are you won't ever, right?

odd, M1 release was the LAST one and i had no issues installing and running.
30 instances on a base M1 ate up about 20% of CPU which is pretty much the same as other plugins when logic is idling, and GPU maxed out.

[DoktorTenma]

Like the earlier releases this still uses an absurd amount of CPU - far more than any of my high quality plugins of the same type I own.
Perhaps it taxes my GPU in addition but so far I have not seen the tiniest proof that this technology could lead to less CPU consumption - absolutely au contraire.

So at this point it seems to be rather safe to say all the nay-sayers (with a lot of credible coders amongst them) were right after all.

If you didn't manage to come up with a working proof-of-concept by now then chances are you won't ever, right?

How did you test this? These are the results on my system comparing their Fireverb with NI's RC24 reverb. (Not that these two reverbs are at all similar, but I just chose a reverb to compare with and look at the behaviour of the GPU Audio Container plugin on my system for a few scenarios.)

My system: Ryzen 1800X CPU, Nvidia 1080 GPU, IK Multimedia AXE I/O Audio Interface @ 96kHz 256 buffers, Reaper 6.77, Windows 10 22H2

I'm using the Windows task manager to see the CPU used by the Reaper process and the total system CPU used.

I'm doing 4 tests for each plugin:
1 track, 1 plugin.
1 track, 10 plugins in series. (With Fireverb I'm using 10 instances in the GPU Audio's Container plugin.)
10 tracks, 10 plugins in series on each track. (With Fireverb I'm using 10 instances in the GPU Audio's Container plugin.) Totalling 100 reverbs.
60 tracks, 1 plugin on each track.

With NI RC24 I'm using the Classic->Room preset, Mix set to 11.0%.
With Fireverb I'm using the Synth 1->1st preset.

Test results:

1 track, 1 RC24 - Reaper 0.5%, System 2%
1 track, 1 Fireverb - Reaper 1.2%, System 2%

1 track, 10 RC24 - Reaper 2.2%, System 3%
1 track, 10 Fireverb - Reaper 1.5%, System 3%

10 tracks, 10 RC24 - Reaper 21%, System 22%
10 tracks, 10 Fireverb - Reaper 3.5%, System 4%

60 tracks, 1 RC24 - Reaper 22%, System 23%
60 tracks, 1 Fireverb - Reaper 10.5%, System 11%

My conclusions from the above tests:

The overhead of transferring data to and from the GPU is not insignificant. So just using the GPU Container plugin for a single simple task is not the best use of the technology if you have other options.

If you have a set of GPU plugins that you can run in serial on a track then you can gain a significant benefit from running these on the GPU and save CPU for other plugins.

If you are running single GPU plugins on many tracks you can still gain some benefit from running the GPU Container plugin.

So to me they have shown that their technology does indeed work. And it obviously works best when you can run a chain of plugins using the GPU on the same track.

[v1o]

Do these take advantage of the M1 GPU, too?

Yeah.

Like the earlier releases this still uses an absurd amount of CPU - far more than any of my high quality plugins of the same type I own.
Perhaps it taxes my GPU in addition but so far I have not seen the tiniest proof that this technology could lead to less CPU consumption - absolutely au contraire.

So at this point it seems to be rather safe to say all the nay-sayers (with a lot of credible coders amongst them) were right after all.

If you didn't manage to come up with a working proof-of-concept by now then chances are you won't ever, right?

odd, M1 release was the LAST one and i had no issues installing and running.
30 instances on a base M1 ate up about 20% of CPU which is pretty much the same as other plugins when logic is idling, and GPU maxed out.

Wait a sec. 30% CPU usage is too much. Aren't these plugins kinda like UAD plugins but instead of outdated DSPs chip they run on a modern GPU? As far as I'm aware UAD plugins dont stress the CPU despite transferring data via USB or Thunderbolt.

[Ploki]

Do these take advantage of the M1 GPU, too?

Yeah.

Like the earlier releases this still uses an absurd amount of CPU - far more than any of my high quality plugins of the same type I own.
Perhaps it taxes my GPU in addition but so far I have not seen the tiniest proof that this technology could lead to less CPU consumption - absolutely au contraire.

So at this point it seems to be rather safe to say all the nay-sayers (with a lot of credible coders amongst them) were right after all.

If you didn't manage to come up with a working proof-of-concept by now then chances are you won't ever, right?

odd, M1 release was the LAST one and i had no issues installing and running.
30 instances on a base M1 ate up about 20% of CPU which is pretty much the same as other plugins when logic is idling, and GPU maxed out.

Wait a sec. 30% CPU usage is too much. Aren't these plugins kinda like UAD plugins but instead of outdated DSPs chip they run on a modern GPU? As far as I'm aware UAD plugins dont stress the CPU despite transferring data via USB or Thunderbolt.

Sorry, i have % set to cores, so 800% is max
So 30% on the 8-core would be more like 3.5%
that's practically nothing.

[Harry Slate]

Like the earlier releases this still uses an absurd amount of CPU - far more than any of my high quality plugins of the same type I own.
Perhaps it taxes my GPU in addition but so far I have not seen the tiniest proof that this technology could lead to less CPU consumption - absolutely au contraire.

So at this point it seems to be rather safe to say all the nay-sayers (with a lot of credible coders amongst them) were right after all.

If you didn't manage to come up with a working proof-of-concept by now then chances are you won't ever, right?

How did you test this? These are the results on my system comparing their Fireverb with NI's RC24 reverb. (Not that these two reverbs are at all similar, but I just chose a reverb to compare with and look at the behaviour of the GPU Audio Container plugin on my system for a few scenarios.)

My system: Ryzen 1800X CPU, Nvidia 1080 GPU, IK Multimedia AXE I/O Audio Interface @ 96kHz 256 buffers, Reaper 6.77, Windows 10 22H2

I'm using the Windows task manager to see the CPU used by the Reaper process and the total system CPU used.

I'm doing 4 tests for each plugin:
1 track, 1 plugin.
1 track, 10 plugins in series. (With Fireverb I'm using 10 instances in the GPU Audio's Container plugin.)
10 tracks, 10 plugins in series on each track. (With Fireverb I'm using 10 instances in the GPU Audio's Container plugin.) Totalling 100 reverbs.
60 tracks, 1 plugin on each track.

With NI RC24 I'm using the Classic->Room preset, Mix set to 11.0%.
With Fireverb I'm using the Synth 1->1st preset.

Test results:

1 track, 1 RC24 - Reaper 0.5%, System 2%
1 track, 1 Fireverb - Reaper 1.2%, System 2%

1 track, 10 RC24 - Reaper 2.2%, System 3%
1 track, 10 Fireverb - Reaper 1.5%, System 3%

10 tracks, 10 RC24 - Reaper 21%, System 22%
10 tracks, 10 Fireverb - Reaper 3.5%, System 4%

60 tracks, 1 RC24 - Reaper 22%, System 23%
60 tracks, 1 Fireverb - Reaper 10.5%, System 11%

My conclusions from the above tests:

The overhead of transferring data to and from the GPU is not insignificant. So just using the GPU Container plugin for a single simple task is not the best use of the technology if you have other options.

If you have a set of GPU plugins that you can run in serial on a track then you can gain a significant benefit from running these on the GPU and save CPU for other plugins.

If you are running single GPU plugins on many tracks you can still gain some benefit from running the GPU Container plugin.

So to me they have shown that their technology does indeed work. And it obviously works best when you can run a chain of plugins using the GPU on the same track.

I also did some tests you might be interested in. I stress tested ReaVerb and FIReVerb (using the same IR) and compared Kilohearts Chorus to GPU Audio’s. Aside from poor performance, a big problem is the huge amount of watts being used.

GPU Audio have been working at this for 12 years and apparently have 36 Engineers in 13 countries with 7 PHDs and at least $6m investment and they burst onto the scene with next to nothing. Until they can produce something that can’t be done better (and more efficiently) with CPU, it’s all complete nonsense.

On a positive, there’s a demo of Nils Schneider’s prototype GPU synth!

[jens]

That (really well-made and -researched) video sums it up nicely!

Lovely dig on that unbearable Paul Third nutter too...

[MirkoVanHauten]

Wonderful experience. Forced to create an account; installer asks where to install the plugin but then installs only VST3 into the standard directory; plugin crashes during scan; even Windows error pop ups during uninstall.

[Harry Slate]

That (really well-made and -researched) video sums it up nicely!

Lovely dig on that unbearable Paul Third nutter too...

Thanks Jens. After a lot of testing and re-testing, I think the results are fair. I'd be interested to hear from GPU Audio, but they're being unusually quiet...

It was good to set the record straight on a few things, as Sasha is relentlessly slating Nils Schneider's 2009 plugin.

[anugget]

Interesting. Thank you for sharing.

[MJACau]

Possibly corrupted Nvidia Driver installation.
Please Check your environment and try again.

Lol because I have RX 7900XTX in my system.