HQ Player

[AnotherSpin]

10 hours ago, jamesg11 said:

Yes, I used to use this, a fair while back - remind me where I can get this again please! (MacOS.)

 

You can create virtual RAM easily with Automator app (that's what I do) - very simple instructions could be found online. Or, there was a small free app called RamDiskCreator (could be not free anymore).

[Miska]

6 hours ago, bibo01 said:

@MiskaFollowing the recent talks here on the new mac chip, is it in your plans to port to M1? Eventually, how soon?

 

My new Mac Mini with M1 arrived yesterday. Sooner than what Apple promised. We'll see how soon I can get the build done.

 

[bibo01]

1 hour ago, Miska said:

My new Mac Mini with M1 arrived yesterday. Sooner than what Apple promised. We'll see how soon I can get the build done.

It means that HQPlayer, at least the client, could work on Apple phones, doesn't it?!

[lotriwer]

What is the latest HQPlayer 3 version and how to download it?

Thank you in advance.

[k6davis]

13 hours ago, Miska said:

My new Mac Mini with M1 arrived yesterday. Sooner than what Apple promised. We'll see how soon I can get the build done.

 

It's just an unsubstantiated leak for now, but there's talk of an Apple M1x chip coming next year that will have 12 cores - 8 high performance cores and 4 high efficiency cores. That's 4 additional high performance cores compared to the current M1 chip. Additionally, since Apple introduced the M1 in their lower priced Macs, I imagine that there will be even higher performance chips (something likely named M2... and beyond) on the roadmap. I expect that someday there will be an Apple Silicon Mac Pro desktop that is focused on maximizing raw performance with no concern for battery life. 

 

We are currently limited to the (absolutely incredible sounding) EC DSD 256 with all of the current CISC, Intel & AMD x86 chips. It's unknown right now if the M1 chip can get us beyond that plateau, but even if it can't, I'm hoping that an HQPlayer optimized for the RISC Apple Silicon can at least run EC DSD 256 with relative ease compared to the x86 chips. 

 

That would be a great sign, given that the current M1 is Apple Silicon's floor, not its ceiling.

 

[mirekti]

I am using Roon and have two PC stations each with HQPlayer. For some reason I am not able to put those two in the same zone and play simultanuosly. Is this a limitation of HQPlayer, Roon or something else?

[MikePid]

I wonder if the M1 Mac will still be limited to outputting DoP, meaning to use DSD256 you need a DAC that supports PCM equivalent to DSD512?

[Miska]

2 minutes ago, MikePid said:

I wonder if the M1 Mac will still be limited to outputting DoP, meaning to use DSD256 you need a DAC that supports PCM equivalent to DSD512?

 

Yes it will be same way. There are no changes in macOS that would change this. Hardware itself won't matter in this respect.

 

[Miska]

42 minutes ago, mirekti said:

I am using Roon and have two PC stations each with HQPlayer. For some reason I am not able to put those two in the same zone and play simultanuosly. Is this a limitation of HQPlayer, Roon or something else?

 

It is limited by Roon, possibly because they have no way to force synchronization between the two instances.

 

With HQPlayer, you can do this with something like RAVENNA, and each DAC would stay in sync to fairly high degree.

 

[Miska]

14 hours ago, bibo01 said:

It means that HQPlayer, at least the client, could work on Apple phones, doesn't it?!

 

iOS is quite different from macOS, in terms of CoreAudio and many other parts. Client could work without too much trouble yes, mostly on iPadOS (because the larger screen would fit the GUI design better). But on the other hand, I don't see much reason for focusing on phones, because HQPDcontrol is good and already available for both Android and iOS.

 

In addition there's no way to distribute software for iOS without going through Apple's App Store.

 

[Miska]

1 hour ago, k6davis said:

It's just an unsubstantiated leak for now, but there's talk of an Apple M1x chip coming next year that will have 12 cores - 8 high performance cores and 4 high efficiency cores. That's 4 additional high performance cores compared to the current M1 chip. Additionally, since Apple introduced the M1 in their lower priced Macs, I imagine that there will be even higher performance chips (something likely named M2... and beyond) on the roadmap. I expect that someday there will be an Apple Silicon Mac Pro desktop that is focused on maximizing raw performance with no concern for battery life. 

 

Yes, I would say it is pretty certain they will produce bigger chips. Maybe without all the extra blocks, because  now they burn so much space on those instead of CPUs. So likely they'd have separate CPU and GPU. And move the AI blocks to the GPU as well.

 

1 hour ago, k6davis said:

We are currently limited to the (absolutely incredible sounding) EC DSD 256 with all of the current CISC, Intel & AMD x86 chips. It's unknown right now if the M1 chip can get us beyond that plateau, but even if it can't, I'm hoping that an HQPlayer optimized for the RISC Apple Silicon can at least run EC DSD 256 with relative ease compared to the x86 chips. 

 

We'll see, I'm not holding my breath given performance of other 64-bit ARM CPUs. In the end this is just a low power category chip, similar to the Intel's U-series Core chips.

 

[k6davis]

26 minutes ago, Miska said:

We'll see, I'm not holding my breath given performance of other 64-bit ARM CPUs. In the end this is just a low power category chip, similar to the Intel's U-series Core chips.

 

I know next to nothing about the chips and their architecture, but for use cases like video editing, people on YouTube are reporting performance from the M1 that exceeds their far more expensive high-end Intel Macs. Often by 2x. There is a special component in the M1 for that particular use case, so that may not mean anything for HQP DSP.

 

Still, the fact that @SwissBear said that, even through Rosetta 2 emulation, the first gen M1 is capable of PCM => ASDM7EC DSD 256 is very promising. 

[StreamFidelity]

As it stands, the benchmarks between x86 CPUs and Apple M1 CPUs are all flawed because x86 are not executed in Hyper-Threading:

 

Exclusive: Why Apple M1 Single “Core” Comparisons Are Fundamentally Flawed (With Benchmarks)

[Miska]

7 hours ago, k6davis said:

I know next to nothing about the chips and their architecture, but for use cases like video editing, people on YouTube are reporting performance from the M1 that exceeds their far more expensive high-end Intel Macs. Often by 2x. There is a special component in the M1 for that particular use case, so that may not mean anything for HQP DSP.

 

Still, the fact that @SwissBear said that, even through Rosetta 2 emulation, the first gen M1 is capable of PCM => ASDM7EC DSD 256 is very promising. 

 

Yes, I can imagine special cases where specific hardware blocks are used it can be nice. But there are no special blocks for HQPlayer cases, so it will be solely on the CPU.

 

But we'll find out... :)

 

[Sevenfeet]

6 hours ago, Miska said:

 

Yes, I can imagine special cases where specific hardware blocks are used it can be nice. But there are no special blocks for HQPlayer cases, so it will be solely on the CPU.

 

But we'll find out... :)

 

 

I was curious as to whether that some of the features of CoreML could be used to tap into the machine learning cores for HQPlayer, assuming that was even feasible based upon its computational needs (since HQPlayer can do CUDA on PCs). But this is just me hypothesizing.....I have no idea if this is even possible.

[Zauurx]

New desired feature : automatic choice of modulator ??
I explain:
I am happy with 4.80.

Now I can use HF convolution directly in HQP with HiRes and DSD under my i5-470.
Before that was impossible for me .. dropout every 2 sec. I was forced to process the convolution under Roon.

Too happy... I let myself try the ADSM5EC. It will crash !!
Well no ... now it's okay !! I try ADSM7EC .. it will crash !!
Well no .. CPU at 85/90% but it works. Thank you !!! 😊

On the other hand, impossible to pass in input a HiRes or DSD file.
So the idea, as for the 1x or Nx filters : automate the choice of modulator according to the input signal.
ADSM7EC for all 44/16 and ADSM7 for everything else (DSD direct or HiRes - 88/96, etc.).

[Miska]

8 hours ago, Sevenfeet said:

I was curious as to whether that some of the features of CoreML could be used to tap into the machine learning cores for HQPlayer, assuming that was even feasible based upon its computational needs (since HQPlayer can do CUDA on PCs). But this is just me hypothesizing.....I have no idea if this is even possible.

 

HQPlayer doesn't use any ML/AI stuff, it is not useful for what HQPlayer is doing. For AI purposes for example Nvidia added support for 16-bit "half precision" floating points. And even more there are added things for even lower resolution calculations like 8-bit needed for AI inference. While HQPlayer things go to opposite direction, using minimum 64-bit "double precision" floating point. And on Intel/AMD CPUs there is luckily still the x87 FPU hardware support for 80-bit "extended precision" floating point which is also extensively used by HQPlayer. It is possible to do up to 128-bit floating point through software emulation, but it has big impact on CPU load.

 

[Plato65]

5 hours ago, Miska said:

 

While HQPlayer things go to opposite direction, using minimum 64-bit "double precision" floating point. 

 

 

So is the benefit of the CUDA offload dependent on the FP64 performance? Most of the consumer GPUs have low FP64 performance as that's not needed for either gaming or machine learning. 

[SwissBear]

Hi everyone,

 

A short update on my experience with the M1.

 

First, I was intrigued by this post of Miska:

and I decided to measure the tension on the STP cable I had on my Mac Mini M1. And this was 30V :-(

So there seems to be a serious leak of current from the PSU on this device... which makes me expect the kit of @Superdad next year... This led me to generalize the use of UTP cables at home. Thanks @Miska

 

Otherwise, the listening experience with HQPlayer is very smooth. On most music files, the combos DSD256/ASDM5EC and DSD128/ASDM7EC are very smooth. I am still very much looking forward to the native version, but I am quite impressed by what I am hearing so far...

[asdf1000]

Hi @Miska

 

What's the best way to run NAA on Win10 but disable general internet access (and annoying Win10 updates).

 

I want to try that for 6 channel

audio to my Focusrite using their ASIO driver

 

Can you share tips

[bogi]

You could disable default gateway. Communication within the same LAN will continue to work since no router is involved in that case.

 

Something like https://superuser.com/a/744528

[voxel]

Hi @Miska,

 

What's the recommended way to access HQPlayer from players other than roon, like JRiver, Audirvana or foobar2000?

 

From description in HQPlayer embedded version:

• It can operate for example as a digital-in-digital-out upsampler processor and convolution engine, or used inside a DAC
• It can also function as a UPnP AV Renderer
• There is also support for the HQPlayer Control API that can be used for implementing a custom GUI or other type of front-end utilizing the HQPlayer playback engine

Does roon use the control API and do other players have to access it as uPnP AV renderer, and can they use desktop version? Appreciate your help!

 

Regards,

Yongsheng

[asdf1000]

17 hours ago, bogi said:

You could disable default gateway. Communication within the same LAN will continue to work since no router is involved in that case.

 

Something like https://superuser.com/a/744528

 

Thanks.

 

I just disabled IPv4 on the particular Windows NAA machine.

 

I use IPv6 for NAA and it's working fine.

 

[MemoryPlayer]

@Miska  Please, add pause with space bar in next updates!

[MemoryPlayer]

38 minutes ago, MemoryPlayer said:

@Miska  Please, add pause with space bar in next updates!

In dark view is hard to see play and pause!

Space bar for pause when playing and play when in pause...

Some covers aren't squares, looking more like DVD covers after 4.8!