HQPlayer Desktop thread

For me I only play ATMOS contents of Apple Music. And luckily most of the ATMOS contents are 48KHz so I just fixed at that rate for BlackHole.


I really appreciate the help!

At some point, I might look at integrating HQP into my theater. I have a 9.5.8 system with a Trinnov processor, (and lots of other fun gear) that I could try this all with. I also have a dedicated pc hooked into it with a 2080ti I had laying around, so it might be worth playing with.

However, my MBP M1 Max has been a beast when it comes to HQP.

I’ve been using HQP to convert DSD265 for my headphone setup (LCD-5s + HSA-1b & CRBNs + Carbon CC). I have a Holo May KTE that should be shipping any day now, so hopefully I’ll get to take things a bit further if the computer can keep up.

Been a big fan of HQP, but couldn’t figure the Black Hole part out. Never would have sorted it on my own.

1 Like

No, only ASIO has that limitation. You can use two CoreAudio devices, they just must be a different device. So you cannot use same device for both input and output.

For example I have RME ADI-2 Pro as input, and iFi iDAC2 as output on my iMac.

1 Like

Thanks for clarification. Sorry I misunderstood it. :sweat_smile:

It seems that the NAA/HQPlayer does not seem to send audio to the optical toslink on my MM 2012?

When I set output device to be the Core Audio / built-in output, oddly the audio is sent to the MM native speakers.

PS: I think probably BlackHole messes things up with the built-in audio output?

Ok let’s move back to Desktop thread. @jussi_laako @dabassgoesboomboom

Could you share your configurations (including Mac model, macOS version / VAD version / ANEMAN setup and network hardware / software configurations)?

If set DSD rate to 128 or 64 in my configurations, HQPlayer Desktop works well as expected. If set to DSD256 the SQ goes bad.

@jussi_laako just an update about my current situation.
After updated to v4.16.1 (x64), in HQPD standalone mode (no other remote controller), upsampling local files to DSD256 has no audible pops.
But if using remote controller such as HQPlayer4Client or Roon, the SQ goes bad. Audible pops sound appeared. It can be heard as soon as the HQPlayer4Client start running. Such pop sound even follows the mouse moving. :rofl: If set the rate to DSD128 or below the issue would be gone.
Based on my observation it probably related to the remote control implementation on HQPD for Mac.

Likely the additional CPU and GPU load caused by HQPlayer Client or Roon is too much and you run out of CPU power at that point. I have not seen such issues on my Macs.

You could try with some lighter algorithms of HQPlayer.

Hardware: 2020 Intel i5 MBP 13"
macOS: Monterey 12.1
VAD: 3.0.4
HQPlayer Desktop v4.16.1
No other software running in the background.
Using poly-sinc-short-mp-2s + ASDM5
44.1 to DSD256
Here’s the Activity Monitor when running HQPD in standalone mode:

The CPU load of HQPD was fairly low, less than 100%.

Then started the HQPlayer4Client:

The CPU load looked still okay except WindowServer.

The noise appeared as soon as the HQPlayer4Client started to run…

And just FYI the HQPD utilizing CoreAudio was much higher. This is the screenshot for Roon (I run Roon ROCK on virtual machine and assigned 4 cores for it. In macOS I run RoonBridge to connect to RAVENNA.) Roon upsampling to DSD256 in VM and it takes roughly 100% CPU resource. Roon utilizes CoreAudio is half of HQPlayer Desktop. Roon has no pop noise issue. Really strange. :sweat_smile:

Have you made sure that the RAVENNA device is not set as macOS default audio output device?

And that when you are using HQPlayer, it is disabled under Roon Settings → Audio?

But I’d recommend to open core load graphs on Activity Monitor and check the individual core loads that none are peaking close to 100%.

Checked. In fact I always use BlackHole as default input / output. :sweat_smile:

Checked. I disabled all other audio. Enabled HQPlayer Desktop only.

I did capture the CPU loading window. It looked quite normal.

Looks fine yes, I recommend to use the per-core histogram graph dialog instead of the momentary bar graph to see per core loads as function of time.

Certainly will do. I would share my new finding… When moving the mouse cursor over HQPlayer4Client the software itself and WindowServer jumped to very high CPU utilization rate. Seems like it might be UI problem?

WindowServer is the desktop compositor, rendering the entire display. This should generally happen at 60 fps rate.

Here’s the screenshots of CPU loading for long period of time. And here’s additional background info for these screenshots:

  1. I tried to make my intel i5 2020 MBP as an NAA and let another computer running HQPD to upsampling / playback DSD256 contents. The result was exceptionally good so I would say the “networkaudiod-x64-412” worked properly with RAVENNA driver 3.0.4 under Monterey 12.1 and the network hardware and Anubis I/O were all ok, too.
  2. based on previous assumptions I setup HQPD on MBP to connect to local NAA. This could also allow me to adjust the buffer time to 100ms.

This screenshot was HQPD in standalone mode, playback local files and didn’t touch the keyboard and mouse. And of course the result was normal.

This was HQPlayer4Client running on foreground, waited every other services idle and didn’t touch the keyboard and mouse. Would hear the pop noise occasionally.

This was operating HQPlayer4Client to browse and playback Qobuz contents. The SQ went bad, you could even hear the pop noise followed by mouse scrolling. The faster you scroll, the pop sound more frequent.

Won’t affect the CoreAudio buffer size though. The code that talks to CoreAudio is same in both cases. But you will get a bit more higher level buffers due to network stack buffers.

Setting is there, because HQPlayer doesn’t know what OS the NAA is running on. NAA then decides what it does with the setting.

One possibility you have is to run Windows based NAA as bridge between HQPlayer and RAVENNA. Or vice versa, running HQPlayer somewhere else and NAA on macOS.

You could also try RAVENNA version of HQPlayer OS as NAA towards RAVENNA, but the Linux driver can be a bit tricky. So YMMV.

Thanks for the advise actually I’m working toward this way. If I’m away from home I’ll switch to Win 10 for better performance and stability.

Certainly. I’ll try to setup my i9 HQPe server to join RAVENNA network later when I get home. So far I tried to setup my UP^2 running RAVENNA version of HQP OS be the NAA but no luck… (I set Anubis as Master clock and could make connection in ANEMAN. After hit play button couple of second it stopped. Then NAA disappeared. System message looked like kernel panic…)

I checked on my Mac Mini M1. Here’s load when playing DSD256 to RME ADI-2 Pro over USB.

If I shake mouse fast on top of Client window so that macOS enlarges mouse pointer to the biggest one, I can get WindowServer to consume 20%, but that is scattered over the efficiency cores and is hardly noticeable (about one vertical block more on every 4 efficiency cores).

And corresponding figure from my 27" iMac (with i9-9900K CPU).

1 Like

Successfully built my i9 HQPe server joined RAVENNA network. Performance-wise is very satisfied. Can do gauss-long + 7ecv2, 6ch DXD → DSD256. The 3070 capped its performance when doing gauss-xl. Will try 3080ti later.


Using the special Ravenna bootable HQPe image?

Plays direct to your Anubis?

This has always been buggy for me. Interesting to know how you go after a week or more if it’s reliable