The use of Powerhouse audible?

[MMKA]

A half hour ago I activated the Powerhouse in MIR 3D and I loaded a piece, that I have worked on last week (new performance, I composed it five years ago and posted it then on the VSL-forum, a string quartet).

I was amazed when I discovered a, I say it carefully, subtile difference in sound. As if it sounds a little more pointed with the GPU setting. That was certainly not what I was expecting.
I don't trust my ears 100%, because what I register in my head is very quickly influenced by where I hold my head, what I pay attention to, is my experience (I wrote an experience concerning that subject earlier on this forum), but every time I switch from the CPU to GPU, It looks like that there is a difference in sound.
Concerning the power of the CPU, it uses only 10% till 15%, so technically the GPU setting is not necessary at all in this piece.
(I use Dorico for this piece on a second PC).

I put here 2 links. One with the CPU setting and one with the GPU setting in MIR Pro 3D.
I'm curious if other people who are interested, hear the same as I do.
For now I don't say which one is the CPU on, and which is the GPU one. Later on I will write this here...
And of course I'm curious if this is the intention.🙂

Example 1

Example 2

[Dietz]

The only valid test: Record both versions (make sure that there are no time-variant parameters involved, like Humanisation, RoomTone, or Chorus FX etc. ....), make sure that they share the same starting point, flip Polarity (a.k.a. "Phase") on one of them, and play both at the same level. If there's something audible, the files were in fact different. Identical files will cancel out completely.

[Music Paul]

Example 2, I like the best. More natural and more room between the instruments.

@MMKA said:

A half hour ago I activated the Powerhouse in MIR 3D and I loaded a piece, that I have worked on last week (new performance, I composed it five years ago and posted it then on the VSL-forum, a string quartet).

I was amazed when I discovered a, I say it carefully, subtile difference in sound. As if it sounds a little more pointed with the GPU setting. That was certainly not what I was expecting.
I don't trust my ears 100%, because what I register in my head is very quickly influenced by where I hold my head, what I pay attention to, is my experience (I wrote an experience concerning that subject earlier on this forum), but every time I switch from the CPU to GPU, It looks like that there is a difference in sound.
Concerning the power of the CPU, it uses only 10% till 15%, so technically the GPU setting is not necessary at all in this piece.
(I use Dorico for this piece on a second PC).

I put here 2 links. One with the CPU setting and one with the GPU setting in MIR Pro 3D.
I'm curious if other people who are interested, hear the same as I do.
For now I don't say which one is the CPU on, and which is the GPU one. Later on I will write this here...
And of course I'm curious if this is the intention.🙂

Example 1

Example 2

[Andreas8420]

I just did a Null Test with GPU vs CPU processing with release version v3636. The test was successful. 100% identical outputs.

[MMKA]

I just finished some tests.

I exported the music 2 times from Dorico with CPU and 2 times with GPU

Results:

2 CPU files gave a correct null test (absolute silence)

2 GPU files gave a correct null test

1 CPU and 1 GPU file didn't give a correct null test. At the start there were some random little sounds, from about 1:30 the viola and random other instruments became more and more audible. Here is a link to the wav file with the results.

Null Test result GPU - CPU

My GPU is the NVIDIA RTX 3060 Ti V2, CPU is AMD RYZEN 9 5950X - 3,4GHz, RAM 128 GB

I use MIR Pro 3D 7.2.3636

[Louis C]

Very odd that those sounds have no phase artifacts at all as one expect if it was an issue of timing.

Is there a possibility that Round Robin oir Humanisation is at play here ?

So variances between the actual samples being pulled by the VI/Synchron player ?

[MMKA]

@Louis-C said:

Very odd that those sounds have no phase artifacts at all as one expect if it was an issue of timing.

Is there a possibility that Round Robin oir Humanisation is at play here ?

So variances between the actual samples being pulled by the VI/Synchron player ?

To explore if Round Robin or humanisation played a roll, I made 2 CPU files and they are identical, so I make the conclusion, that that is no issue. I made also 2 GPU files, and they are also identical. Just the CPU files and GPU files are not identical.

[Sasha T]

Hey everyone!

GPU implementation has a different approach than CPU because of the architecture difference, so there is a byte difference between the CPU and GPU modes outputs, but there is no audio difference noticeable by the human ear. It is safe to work with your audio projects and convert and render them using your GPU.

You can use any input for both modes and substruct one from another, and the noise floor will always be below -96.0dB or somewhere around this level. Here are some proof screenshots attached:

[Sasha T]

@MMKA said:

I just finished some tests.

I exported the music 2 times from Dorico with CPU and 2 times with GPU

Results:

2 CPU files gave a correct null test (absolute silence)

2 GPU files gave a correct null test

1 CPU and 1 GPU file didn't give a correct null test. At the start there were some random little sounds, from about 1:30 the viola and random other instruments became more and more audible. Here is a link to the wav file with the results.

Null Test result GPU - CPU

My GPU is the NVIDIA RTX 3060 Ti V2, CPU is AMD RYZEN 9 5950X - 3,4GHz, RAM 128 GB

I use MIR Pro 3D 7.2.3636

Can you please share the entire test project with us, so we can also go through all of that? It could be a setup difference or any other kind of non-algorithmic issue. We'll run the test and see whether it's reproducible or local.

[MMKA]

@Sasha-T said:

@MMKA said:

I just finished some tests.

I exported the music 2 times from Dorico with CPU and 2 times with GPU

Results:

2 CPU files gave a correct null test (absolute silence)

2 GPU files gave a correct null test

1 CPU and 1 GPU file didn't give a correct null test. At the start there were some random little sounds, from about 1:30 the viola and random other instruments became more and more audible. Here is a link to the wav file with the results.

Null Test result GPU - CPU

My GPU is the NVIDIA RTX 3060 Ti V2, CPU is AMD RYZEN 9 5950X - 3,4GHz, RAM 128 GB

I use MIR Pro 3D 7.2.3636

Can you please share the entire test project with us, so we can also go through all of that? It could be a setup difference or any other kind of non-algorithmic issue. We'll run the test and see whether it's reproducible or local.

I send you a PM.

[MMKA]

@Dietz said:

The only valid test: Record both versions (make sure that there are no time-variant parameters involved, like Humanisation, RoomTone, or Chorus FX etc. ....), make sure that they share the same starting point, flip Polarity (a.k.a. "Phase") on one of them, and play both at the same level. If there's something audible, the files were in fact different. Identical files will cancel out completely.

... and thanks, Dietz, for your advice! That helped a lot!

[Sasha T]

HI @MMKA, we've tried the project you shared in DM and followed your instructions, and, unfortunately, we could not reproduce the issue locally.

I wonder, what if you assemble some project from scratch and repeat the same test, and you'll experience the same issue?

[MMKA]

@Music-Paul said:

Example 2, I like the best. More natural and more room between the instruments.

@MMKA said:

A half hour ago I activated the Powerhouse in MIR 3D and I loaded a piece, that I have worked on last week (new performance, I composed it five years ago and posted it then on the VSL-forum, a string quartet).

I was amazed when I discovered a, I say it carefully, subtile difference in sound. As if it sounds a little more pointed with the GPU setting. That was certainly not what I was expecting.
I don't trust my ears 100%, because what I register in my head is very quickly influenced by where I hold my head, what I pay attention to, is my experience (I wrote an experience concerning that subject earlier on this forum), but every time I switch from the CPU to GPU, It looks like that there is a difference in sound.
Concerning the power of the CPU, it uses only 10% till 15%, so technically the GPU setting is not necessary at all in this piece.
(I use Dorico for this piece on a second PC).

I put here 2 links. One with the CPU setting and one with the GPU setting in MIR Pro 3D.
I'm curious if other people who are interested, hear the same as I do.
For now I don't say which one is the CPU on, and which is the GPU one. Later on I will write this here...
And of course I'm curious if this is the intention.🙂

Example 1

Example 2

No 2 is the CPU version. But obvious there is something wrong here in my setup. We will see if we trace what the cause of the issue is

[MMKA]

I want to mention here, that I have brought my pc to the vendor to check my pc and specially my video card, if there's something wrong with the hardware. To be continued...