View Single Post
Old Aug 21, 2021, 04:12 AM   #17
Radeon Volcanic Islands
Join Date: Mar 2003
Location: Australia Australia
Posts: 3,700
drumphil can beat 'Minesweeper' on any difficultydrumphil can beat 'Minesweeper' on any difficulty


Lol, I just noticed that amirm, who runs audio science review has commented in that thread!

Post number 478.

Please forgive me for reciting a bit of my resume before commenting on this . Some two decades of my professional experience was in operating system ("kernel") development and I managed the digital media division at Microsoft for a decade where the entire audio/video/imaging subsystem was part of my group.

In a nutshell, what these tools do has little merit and may actually make things worse, not better!

Here is the reason: the moment you boot up a modern operating system, on a CPU this day and age, massive amount of activity is going on in your system. No amount of "simplification" of what is running will do but a trickle to slow this down. The moment you try to read the audio file, tons of activity may occur to free memory, prefetch the file, filling CPU caches, virtual memory faults, etc., etc. No user optimization process can impact these activities because they run at the kernel privilege and are below any program/process running in the system. To think you can clamp this down is like thinking you can tell quiet down the sound in a loud concert by asking the person next to you to not talk .

There is also confusion on what is a fidelity improvement and what is a performance improvement. Let's look at what fidelizer says they do:

Everything they mention is aimed at one thing: to make sure your music plays without interrupts. Is that what you are trying to fix? If not, and that is the case for just about every situation, then you don't need any of this.

Your audio subsystem is designed to be robust in the face of chaotic operation of the system as it is. As such, there is fair amount of "buffering" where audios amples are read ahead and placed in temporary places for the audio hardware to draw and play. With the workload of playing music not even needing 1% of the CPU resources these days, even when the system falls behind, it can catch up at lightning speed. This is why even on your everyday machine or even phone you can play music with no interruptions.

More goodness here is not more goodness. Your system only needs to keep up with the audio being played. It is a relay team and going any faster than the rest of the members (i.e playing music) makes no difference.

This is more of the same. Much of the music playback process is already running at higher privilege than any normal program. When your DAC needs an audio sample, it gets it with extreme priority already (from the buffer). Yes the music player can fall behind but again, if it did, you would get an audio glitch. The fidelity would not change at all. It is just a pause and if there is no pause, you are good to go!

They are making a mixed claim here. One of them is the same as above. That is they stop any other process that may get in the way of playing music. Again, if the music is not pausing, this is of zero value.

The second part is where they hat on for fidelity. The idea is that by having less activity in the system, there is must be less noise, and hence better fidelity. As I explained at the outset, this is a fantasy. There is a reason this impact can't be measured in noise performance of the PC playing back with any half decent hardware. The machine is extremely noise at all times whether this optimization is done or not.

Yes, there is poorly designed hardware that I have measured that shows impact based on CPU activity. The solution there is to simply avoid that hardware. Vast majority of devices I have tested even at subh $60 price points don't have any issues here.

On timing, with an asynchronous DAC which you should be using, none of that comes from the operating system. The DAC hardware has its own and nothing you can do in the OS will change that.

I explain all of this so that you don't go into evaluating such a solution as "it must make things better." Because if you do, you absolutely will think that it does! Know that from someone who has worked on every line of such code and managed teams that built the same that what they think they have built is not what is built. They are taking advantage of concepts that are way too advanced for just about anyone to understand to imply that there is improvement. If there were any, we could trivially measure it.

Finally, I can make a fake program that says it does all of this but in reality doesn't, give it to you, have you run it and you will swear that it made audible improvements! I can do that even you are skeptical that it can!!! Our audio evaluation process is this broken.
You know, for the cost of a Nu Audio card, and a copy of fidelizer, you could have just got an external DAC with far superior audio fidelity!

Frankly, if fidelizer actually did what it claimed to do, then none of the digital audio links in the studios I manage would work without it! And they all do. If the problem it claims to solve was real, then the stuff I do wouldn't work, and it does.
Jasef: "I am a lawful man. By serving pee to cops, you support terrorism. I will lock this thread."

12Bass: re: 9/11 nano-thermite conspiracy theory "Nanotermites... turns out it was a spelling error. Also explains the lack of explosions before the collapse."

Last edited by drumphil : Aug 21, 2021 at 04:43 AM.
drumphil is offline   Reply With Quote