Fidelizer experiences & insights

[SimplySound]

Anyone using Fidelizer on their Media Server? I'm interested in all experiences and insights.

 

[cpt_paranoia]

Had a quick look at their claims. If a server cannot serve media in a timely manner, it is broken. Though some housekeeping to remove all non-essential services, and prioritize audio services might fix an underpowered server; Fidelizer may do this. But there ought to be sufficient slack in the various buffers in the audio pipeline to ensure the DAC never misses data. 'Synchronous USB' has potential to cause difficulty, depending on where the timing master resides, and how the USB stream is serviced.

My feeling is that the following thread probably expresses my feeling on the subject of 'optimisation' of server/digital streamer performance:

https://audiokarma.org/forums/index.php?threads/digital-streamers-audio-quality.1019210/

I confess I'm wary of claims about such optimisation and 'digital glare', since, AFAIAC, the DAC does any 'glary' stuff, and the server is just an IT appliance.

There's a free version. It says its changes are safely reversible. Give it a try. Just beware of expectation bias...

 

[SimplySound]

OK, finally got around to upgrading my Media Server. I went from an 8yo Asus/AMD Phenom-II platform to something a bit more modern. Put in a new MSI PRO B760M-P MicroATX board w/Intel i5-12th Gen processor and paired it with 32Gb of RAM and a Samsung 980 Pro M.2 NVMe drive. This then allowed me to upgrade from Win10 to Win11, which was one objective. I know, probably a bit of over-kill but it's fast!

Now, I run Kodi for video and JRiver MC31 for audio. Nothing else runs on this platform, as this is a dedicated media server. I took some time to uninstall as much of the Microsoft bloat as I could and pared down the services (no I don't need a print spooler). Right off the bat I noticed that my Audio playback via JRMC sound a bit better. More detail and seemed like there was a bit more depth in the music (clarity?).

Don't know if it is the increased speed of the NVMe or changes made to Windows' Multimedia Realtime framework (MMCSS) in Win11. Really not sure, but the system took a minor step up with just these changes, so overall, I'm happy with the ~$400 upgrade.

Now, this is when I said OK, let's see what Fidelizer actually does. So, I spent about 2hrs this morning going back and forth between Fidelizer and non-Fidelizer mode. This is the nice thing about the Freeware version, it is only active when you launch it for that session. If you re-boot it's gone. Since my system now reboots in like 15secs, that makes going back and forth easy.

To give you some sense of my references, I focused on female vocals with Adele's 'Million Years Ago' and Diana Krall's 'Isn't it Romantic'. I threw in her 'My Love Is' because it is basically her voice with a finger snap through out. I also used Gerry Mulligan's 'Parasio' and 'Wave' to get some baritone sax with female vocals. And then tossed in Jack White's acoustic version of 'I'm Bound to Pack It Up' to have a bit of male off-key vocals (sorry Jack, but your vocal key is a bit off).

So what did I notice? Well I'll say that in general the highs were more natural (less sharp) and I felt like the depth was a bit more extended (better bass management?). In general, I thought everything sounded just a bit more uniform and consistent (tempo?). These weren't 'in your face' kind of things, but just better musically.

IMHO, there does seem to be something to Fidelizer's prioritization of audio playback and control of services. Is it worth 'free'? I'd say so! Is it worth $40 or $70 for the Plus or Pro editions? I guess that's for others to decide and your mileage may vary. Given my experience with the Free Edition, I think I'll spring the $70 for the full blown package. It's a drop in the bucket compared to the rest of my system and it supposedly provides additional enhancement.

I guess I'll find out, and I'll report back with my impressions using the Pro edition.

 

[SimplySound]

Did another focused session with Fidelizer-Free. Used a single song to evaluate back and forth; Sting's 'Consider me gone'. Figured I'd focus in a single song to really get a sense of what changes. First I'll refer you to the 6Moons Review done by Paul Candy back in 2015. He states;

In my initial listening notes, I crossed out Fidelizer and wrote Spatializer instead because the most obvious trait to my ears was how space rendered; and how various instrumental and vocal images placed within it. To me, most computer playback systems sound slightly flat, two-dimensional and lacking in body. The enhancement of depth and the resolution of the complex layering of large-scale orchestral music came off stunning with Fidelizer.
I noted how Renaud Garcia-Fons' 5-string upright bass on Silk Moon [16/44.1 FLAC] came across clearer, more articulate and natural with Fidelizer Pro. Derya Tϋrkan's kemence, a bowl-shaped 3-string lute, sounded cleaner and more defined, less wiry and abrasive. The space around both musicians seemed more dimensional and natural but also more vividly charged. I heard similar traits with all recordings regardless of bit depth and sample rate.

I would generally agree with his assessment. While I found that the sound stage became perhaps a bit narrower, it was more defined and clearer. There was better texture of instrument placement and the tonal balance just seemed better. I did get licensed for Fidelizer-Pro, so that will be the next step for assessment. As Paul wrote;

After a spell I removed Fidelizer Pro and re-installed the free version. While music sounded more transparent, tonally correct and dimensional than not using Fidelizer at all, the Pro version was significantly superior in every way, particularly in overall transparency and in its rendering of space and layering.

Looking forward to this listening session!

 

[cpt_paranoia]

I would be interested to see if Fidelizer is modifying the audio files; maybe someone with one of the RMI DACs could run RMI's bit-true test files through the system.

 

[SimplySound]

That would be interesting, but not sure there's any re-sampling going on. According to Keetaka, he's really just working with WinOS processes & compute Core priorities. This is from the v8.0 release;

Core Isolation Policy – This is the main feature introduced in Fidelizer 8 for Plus and Pro version. Core Isolation will allow user to choose how audio task should work with other processes. You can choose how to utilize all cores for audio thread and isolate audio thread with non-audio processes. Core isolation can also apply to music player application to further resolution and transparency.

On another note, I read in some of his postings that he recommends enabling the High Precision Event Timer (HPET). Seems most BIO's now have this enbled by default, but it still has to be enabled in the OS to leverage it. The gamers seem to be of the mind that enabling it impacts the video/sound synch, so many of them suggest disabling it. However, I've read in a couple forums that for pure audio (DAW's) it's better for timing of transport. This was taken from the Neowin forum:

Enabling HPET in BIOS is just half way of enabling HPET, it needs to be enabled in OS too, and in such a way that it's the only timer used.
By default windows uses a combination of TSC+ACPI timers, no matter if HPET is enabled in BIOS.
TSC+LAPICs Low performance (slow timers + syncing)
LAPICs low performance (slow timer - no syncing)
TSC+HPET medium performance (slow and fast timer + syncing)
HPET high performance (fast timer - no syncing)
HPET + platformclock=true will give you best timer resolution, frame rate and lowest DPC latency.

This is the tweak, and as always; understand your system, what you are doing and pursue at you own risk!

 

[cpt_paranoia]

According to Keetaka, he's really just working with WinOS processes & compute Core priorities. This is from the v8.0 release;

The serialisation of samples to a DAC depends on the interface; SPDIF serialisation must be done at source, and most likely done with a hardware serialiser and associated FIFO buffer to isolate the hardware timing from the software process timing. The FIFO is serviced by the software, triggered by some content threshold in the FIFO. If the software does not service the FIFO buffer in a timely manner (i.e. before the remaining samples in the FIFO, from the threshold level, are consumed by the serialiser), such that buffer underflow occurs, that's a basic failure. The audio stack usually has a number of cascaded buffers, of different sizes, at different stages of playback. That's certainly my experience of implementing real time audio under the Linux ALSA driver.

Any interface protocol that uses a frame-based interface to the DAC (especially an 'asynchronous' USB interface) defers sample serialisation/sequence timing to the DAC. This will also have cascaded FIFO buffers, and a hardware serialiser. If the interface fails to service these buffers in a timely manner, again, that is a fundamental failure, and the effect is not subtle. It is not a matter of sample microtiming variation, that is usually associated with poor imaging.

 

[satkinsn]

Like cpt, I'd be interested in knowing whether the program is doing something to alter the audio files. At this late stage in the evolution of the PC, you'd pretty much have to go out of your way to have a less than perfect audio server, and certainly the machine Simply Sound describes is many times what's required to do the job. As cpt notes, if there's a problem, it won't be subtle. Conversely, my expectation would be any software promising improvements on what is already a competent box would deliver obviously better sound.

s.

 

[SimplySound]

I think we would all agree that timing is critical, and I guess I get where Cpt is coming from, the DAC buffer should address this independent of the feed from the PC. But there's a lot of moving parts at 192k bits per sec.

I don't know for sure, but I think timing could be what Keetakawee (Fidelizer developer) is addressing in multiple ways (CPU core dedication, QPC queries, etc.) Deep in the bowels of MSDN, I ran into this:

HPET Timer (x86 and x64)

The High Precision Event Timer (HPET) was developed jointly by Intel and Microsoft to meet the timing requirements of multimedia and other time-sensitive applications. Unlike the TSC, which is a per-processor resource, the HPET is a shared, platform-wide resource, though a system may have multiple HPETs. HPET support has been in Windows since Windows Vista, and Windows 7 and Windows 8 Hardware Logo certification requires HPET support in the hardware platform.

Towards the bottom of this MSDN reference it gets into Low-level Hardware Clock Characteristics using QPC. I could almost see this being something like a USB Bridge using a dedicated clock, but in this case the HW Platform Clock is the reference within the PC & is tightly leveraged by Fidelizer. Hence the recommendation from Keetaka to enable HPET in the OS makes sense from this standpoint.

I'm just speculating, but it's interesting that there does seem to be something going on here. I agree that I'm not doing a DBT, but I don't think my ears are deceiving me. The fact that Amir & Archimago haven't measured any differences associated with & without Fidelizer, doesn't mean there isn't something going on deeper in the stream.

But I get their point... jitter is jitter, you either measure it or you don't!

 

[cpt_paranoia]

But there's a lot of moving parts at 192k bits per sec.

That's a trivial bit rate compared to video or GBE: 1.152 MB/s

 

[cpt_paranoia]

The High Precision Event Timer (HPET) was developed jointly by Intel and Microsoft to meet the timing requirements of multimedia and other time-sensitive applications.

Only if it's purely software driven. The FIFO buffer, with a hardware read mechanism, provides the causation barrier between the software and hardware tining domains.

 

[SimplySound]

Only if it's purely software driven. The FIFO buffer, with a hardware read mechanism, provides the causation barrier between the software and hardware tining domains.

I guess you lost me here... an OS is exactly and purely software driven. Driving hardware yes, but when and how events get driven is open to scheduling. And the only thing that sticks out about what Fidelizer is doing is it's all about allocation of resources and priority of timing.

Many people, including some well respected individuals in Amir and Archimago, say there is nothing to measure so it doesn't do anything. If that were truly the case then Keetaka would have gone out of business years ago. But he's been on this for over 15 years and is up to version 8 of the software. So obviously, someone hears something.
But your question about re-sampling is legitimate, because none of us knows what the code is doing.

The underlying question is this; you say 44 or even 192kHz is trivial. I agree that in the grand scale of a modern computer processor it is, but how that trivial stream of information is handled by the many subsystems may have more to it than anyone really understands at this point.

And no, I don't think the 'art' of Computer Science is fully mature at this point. There's always something to learn and improve in this world of bits & bytes.

 

[SimplySound]

So one more question; have you two played around with Fidelizer, and if so, what was your experience?

 

[cpt_paranoia]

Driving hardware yes, but when and how events get driven is open to scheduling.

But the point is that IO design such as this has used hardware FIFO buffers for decades, with sufficient FIFO capacity to allow the CPU time to context switch from an interrupt (not a scheduler) from the FIFO threshold. This is an ancient 'design pattern'.

 

[SimplySound]

Fair enough Cpt, I don't believe anything is being done with re-sampling of the stream, so I fall back to the fact that since Keetaka is working with the underlying hardware processing prioritization and timing, that it must be having an impact. Similar to how Amir and others have noted that using a USB Bridge with independent clock provides an improvement over straight HDMI feeds. Best I got!

I'll carve out some time to compare Fidelizer-Pro to the Free version and let everyone know what I think. Should be interesting.

 

[satkinsn]

So one more question; have you two played around with Fidelizer, and if so, what was your experience?

I haven't, but will give the free version a try. I have a much older server - a third generation i7 with 12 gigs of ram. That is still way more than required for transparently passing audio to a DAC.

s.