HQplayer 4 and Mac-mini M1

[Miska]

6 hours ago, musicjunkie917 said:

Would a Mac mini with an M1 Pro, which has double the performance cores of the existing M1 Mac mini, run Roon+HQPlayer any better than the current M1 Mac mini?
 

The performance cores all perform about the same between the M1 and M1 Pro.
 

In other words, does the M1 have enough performance cores to do the job or would  there be a benefit to having more performance cores?

 

It would help with filters, convolution and such. So likely you could run wider variety of filters. M1 Pro also has double the memory bandwidth of M1. And M1 Max has 4x the memory bandwidth of M1. This also has some amount of impact on the performance.

 

[Miska]

I would say M1Pro still helps compared to M1 even on simple stereo tasks for HQPlayer. Then Max and Ultra will provide 4x and 8x memory bandwidths and more CPUs. These will help with convolution and heavier filters. And the memory bandwidth may even help with running modulators.

 

But I would say the M1Ultra is probably very nice primarily if you want to run 8 channels of DSD256 to exaSound s88, Merging NADAC/Hapi or similar. Because then you need to have 8 modulators and 8 filter sets, instead of just two. And there the number of cores and memory bandwidth will certainly help. Especially if they didn't cut on CPU clock frequencies.

 

Story doesn't say a word about the clock frequencies these SoC's have...

 

But seems like their architecture scales nicely to heavier use cases, since memory bandwidth keeps growing along without becoming a bottleneck.

 

Prices for the Ultra machines are just such, that you can get i9-12900K plus a big Nvidia GPU for cheaper.

 

[Miska]

29 minutes ago, camott said:

Asahi Linux for M1 Macs Alpha release is out! @Miska Any chance HQP desktop or embedded will work on it, with or without a bit of tweaking?? Maybe a bit more performance can be extracted vs Mac OS.
 

Asahilinux.org

 

I will see when there are official non-alpha/beta Linux releases for M1. Too bad Asahi Linux is based on Arch, so I don't have packaging for it.

 

But of course you can try to run the already existing arm64 build on it.

 

Next question is how much secret magic Apple's compiler has on macOS for their silicon, compared to the mainstream arm64 compilers.

 

[Miska]

5 hours ago, musicjunkie917 said:

I am disappointed that the i9-9900 in my other system seems to perform better than the M1.

 

It is not surprising. That is a big desktop CPU, while M1 is a low power mobile processor, more comparable to U-series Intel CPUs.

 

It is still amazing for what it is.

 

[Miska]

On 3/20/2022 at 3:54 PM, itguy61 said:

Is this an issue with the M1 Mac minis?

 

No such issues on M1. M1 has different security processor implementation built right into the SoC.

 

On 3/20/2022 at 3:54 PM, itguy61 said:

2.  Reading the many success stories here is leading me to believe this Mini works well for this.  I should be able to use some higher load filtering with this than the old i7 I had correct?

 

Maybe, although the basic M1 having only 4 performance and 4 efficiency cores limits the possibilities somewhat. M1Pro/Max/Ultra open up more filter possibilities.

 

[Miska]

6 hours ago, itguy61 said:

Just setup my new Mac Studio with the M1 Max Processor (base model).   Does anyone have any suggestion for settings to try to see how high demand settings this thing will run?   I have been on an older iMac which had an older four core i7 in it so many of these filters and modulators could not be used.

 

Highest demand filters are gauss-xl/xla, ext3 for Nx rates (192k etc). And sinc-L for RedBook rates.

 

ASDM7ECv2 modulator is highest demand.

 

6 hours ago, itguy61 said:

Also would suggest any suggestion for good filters  and modulators that this processor may be able to run for listening to mainly rock.  I am using a Topping D50S dac and Roon as well. 

 

Use ASDM5ECv2 modulator at DSD256 (DSD512 if possible), and there are some suggestions about filters in the manual table.

 

[Miska]

11 hours ago, itguy61 said:

I know it is subject, does anyone have any suggestions as to which DSD filters are smooth and warm sounding?

 

From PCM to DSD? You could try for example poly-sinc-short-mp(-2s) with ASDM5ECv2 modulator. Or poly-sinc-ext2 or poly-sinc-gauss filter.

 

[Miska]

6 hours ago, itguy61 said:

PCM to DSD.   I have been using EXT2 a long time on HPPlayer 3, I just upgraded both HQplayer and went to a Mac Studio.   Just wanted to see what the world of filters I could not run before is like and see what others are running.   I may just have to go through the process of listening to all the candidates and see what stands out.  

 

Please let us also know how the M1Max in Studio performs. It should be very capable CPU.

[Miska]

10 hours ago, 57gold said:

Could someone explain the 1X filter versus the Nx filter selections, what's going on and what benefit results from selecting different filters for each option?  Thanks.  Read the manual and even after hanging around here periodically for years, it's mostly "all Greek to me."

 

1x filter is for 44.1/48 source rates, and Nx for rates above those. Point is that those 1x rates require much steeper and more aggressive filter than the higher rates where you can have earlier more gentle roll-off.

 

[Miska]

4 hours ago, AnHa said:

 Unfortunately the scan stops surprisingly quick after maybe 60 seconds and fills the library with only a small fraction of the music, that ist stored under the root directory. I tried to scan several times with and without removing all previous entry and by Clear Scan as well. Checking/Unchecking Structure only or Extract covers did not make a difference. Any ideas? Thx in advance   

 

Is the content in supported format? Does HQPlayer have permissions to access the content? You could enable log file from HQPlayer settings, restart HQPlayer and initiate the scan. And then check the log file for any possible errors. Log called HQPlayer4Desktop.log is located in "~/.hqplayer" folder (you can use Finder's Go To menu option to get there). It is just a plain text file.

 

[Miska]

15 hours ago, rah50 said:

Also, is there a preferred dac bit setting other than default that is recommended for the Pontus?

 

I have not seen linearity sweep results for Pontus, but if it's like Ares II, it is best to set DAC Bits to 16.

 

15 hours ago, rah50 said:

I can't quite run poly-sinc-gauss-xla with ASDM7ECv2 at DSD256 which everyone seems to really like.

 

It should work for 1x on M1 Mini, for Nx you need to select something else like gauss-long or gauss-hires.

 

[Miska]

1 hour ago, rah50 said:

Still can't get poly-sinc-gauss-xla with ASDM7ECv2 to run at DSD256. I had an mds_stores issue that was taking a lot of cpu power but got that fixed. HQP is using 79-80%, system 3 or 4% more so I have about 15% left, but still get a regular stutter every 10 seconds or so. 

 

Is there any else I can try? Also, I'm on v4.18.1

 

1 hour ago, bobflood said:

I can't get it on my M1 either.

 

Are you sure you have the native arm64 version installed? And not the x64 one?

 

[Miska]

8 hours ago, bobflood said:

Yes I have the arm64 version. I can get Gauss long no problem but not Gauss xla. I don't care though as Gauss long is just fine on 1x material and I use Gauss hi-res for nx material.

 

That is strange, as all M1 Mac Mini's should be equal...

 

 

[Miska]

1 hour ago, bobflood said:

Thanks, That was it for me as well. On the M1 it is either xla or rate family conversion but not both at the same time. Rate family conversion works fine with all the other Gauss filters just not xla.

 

Note that going from 48k-base to 44.1k x256 is lighter than going from 44.1k-base to 48k x256. So the more typical required case is lighter on the CPU/RAM.

 

[Miska]

6 hours ago, Hoshi said:

Prior to the previous Roon update, I was able to play DSD256x48 and ASDM5ECv2 or 7ECv2: 

 

1x - poly-sinc-gauss-xla

Nx - poly-sinc-gauss-long or any of the newer poly-gauss hires versions

 

Since the update I've lost the ability to play any poly-gauss options or poly-gauss hires versions without experiencing dropouts with hires content.  1x - still plays poly-sinc-gauss-xla perfectly after the updates for redbook.  

 

I've updated to Monterey 12.3.1 and the new Roon update this morning with no changes.  I've believe I've tried every conceivable option with Multicore DSP and Adaptive output rates.  Setting Buffer time to 250 ms seems to help but I still experience an occasional dropout.    

 

Check that you don't have any Roon DSP enabled. And you are on latest arm64 HQPlayer as well?

 

Roon shouldn't have such a big difference in performance, unless it is doing some background activity like scanning/analysis (also check that these are set to "Throttled".

 

Output is to some locally connected DAC? Does it work without Roon, using HQPlayer as standalone player?

 

[Miska]

9 hours ago, Hoshi said:

The Hermes DDC seems to have been the culprit as once removed the dropouts have disappeared.  Do you have any idea on what the DDC is doing to cause the issue?  I will contact Denafrips.  

 

Only potential reason I can think of is I2S clocking. For example if DDC sends DSD at 48k multiple and the DAC doesn't support it. Or if there's a PLL on the I2S input receiver and it keeps slipping or uses wrong reference clock (wrong rate family detection).

 

[Miska]

12 hours ago, SKR said:

I have a problem in registering my HQPlayer software. Actually I live in 2 places, rather two countries , India and Thailand.

When I first purchased and Registered my HQPlayer 4 software on my Mac mini M1, I was in New Delhi India.

Now I am in Bangkok Thailand and unable to register HQPlayer on my Mac mini here.

Is there any restriction on using the same KEY at two places, BUT NOT AT THE SAME TIME USING BOTH. I will be using only one at a time.

 

It should work fine as long as you have only one copy running at a time.

 

 

[Miska]

18 minutes ago, FooFighter said:

Chain is Mac-Mini M1 running Roon into HQPlayer Desktop -> Ethernet -> UP Gateway -> Holo Audio Spring 3 KTE

 

What is running on the UP Gateway?

 

[Miska]

51 minutes ago, FooFighter said:

Testing further it seems also that HQPlayer takes years to switch sample rates between Redbook and MQA (and gets stuck there eventually).

 

It can take a particularly long time on M1 with the filter you have selected and especially because you also have rate conversion from 44.1k family to 48k family.

 

If you try for example poly-sinc-gauss family of filters, you may experience less delays. Or even if you just enable adaptive output rate.

 

[Miska]

19 minutes ago, Louisiana said:

@FooFighter

I also use a Mac Mini M1, but poly-sinc-xtr-mp does not work for me, how is that possible?

 

One possible difference is conversion within rate family vs between rate families which is heavier.

 

[Miska]

5 hours ago, SKR said:

I can only upsample upto DSD128. Normal files can do 256 also. But high resolution files like 192 and above cannot be played.

 

If 1x rates work fine to DSD256, then likely Nx rates will start working as well if you select a lighter filter. You could for example try the current default poly-sinc-gauss-hires-lp, or alternatively something like poly-sinc-short-lp-2s

 

[Miska]

2 hours ago, Reg19 said:

If, instead, I get a MacBook Pro with M1 Pro chip, would it likely be better than the M1 Mac Mini as well or does that fact that it is a laptop (with a attached screen) rather than a headless computer (as I currently run my M1 Mac Mini) hurt it in someway?

 

Possible limiting factor is thermal management. I think Mac Mini is not limiting M1 power in order to fit some narrower thermal envelope. But devices like MacBook Air is. I think the bigger MacBook Pro has special cooling arrangements to allow sustained full loads on the bigger M1 models. When I find time, I will do some HQPlayer testing on my new MacBook Pro. But so far I haven't got time.

 

[Miska]

Here's some initial load figures from my MacBook Pro with M1Max.

 

[Miska]

23 hours ago, jamesg11 said:

Interpretation, for a current macmini m1 16gb user?

 

M1 Max and other versions with more cores allow wider variety of filter choices. And may allow even DSD512 with EC modulators.

 

After a bit of new tunings on the new MacBook Pro, this case works perfectly fine (MacBook Pro fans get very loud though 😆):

 

And this case is almost stable:

 

Both cases with 10 band headphone correction EQ. Source is from Thunderbolt HDD output to a NAA over wired network (Type-C to Ethernet dongle).

 

These would be likely better with onboard Ethernet and M1 Ultra on the new Mac Studio. But unfortunately it is too expensive for me.

[Miska]

8 hours ago, Schafheide said:

I have just updated to 4.19.2

Based on your info re this update -

Bug fixes and component updates. Core usage improvements in certain cases.

It would seem that our Mac Mini M1 are not included?  I was hoping to be able to use EC.

I do appreciate that you cannot always perform miracles!

 

Core usage improvements cover for example M1 Pro/Max/Ultra. But no changes for regular M1.

 

I'm using these settings on my Mac Mini M1:

1x = poly-sinc-gauss-long

Nx = poly-sinc-gauss-hires-lp

Modulator = ASDM7ECv2

 

Output rate is DSD256.