HQPlayer Linux Desktop and HQplayer embedded

[Miska]

6 hours ago, Outlaw said:

Running HQPlayer OS.I use a vpn client and HQPLayer Client won't connect to my HQPlayer OS PC when vpn is running. Any ideas

 

You need to configure VPN to allow local network traffic bypass the VPN.

 

[chipvn]

On 2/10/2022 at 12:45 AM, Miska said:

 

Seems to work with the "amd" build:

 

It is interesting that with the 4.16.2 AMD version on Ubuntu 20.04 LTS desktop, my passive cooled i7-9700 can also upsampling even 192kHz PCM to DSD512 with ext2 filter and ASDM5ECv2 modulator. I need to allow the CPU to run max at 4.3GHz (AVX offset 2) so CPU max temp after 1.5hrs of playing reached about 85degC. A bit too hot.

For comparision, if upsampling 192kHz to DSD256 with ext2 filter and ASDM7ECv2 modulator, my CPU only need to set max at 4GHz (AVX offset 2) so max temp after 1.5hrs of playing will be just about 65degC! ☺

For Holo Spring 3 L2, which combination do you recommend to use: DSD512 with ext2/ext2/5EC or DSD256 with ext3/ext2/7ECv2?

Thanks

 

 

[Miska]

51 minutes ago, chipvn said:

For Holo Spring 3 L2, which combination do you recommend to use: DSD512 with ext2/ext2/5EC or DSD256 with ext3/ext2/7ECv2?

 

Which one sounds better to you? From technical performance point of view you are certainly fine already with DSD256.

 

[chipvn]

8 hours ago, Miska said:

 

Which one sounds better to you? From technical performance point of view you are certainly fine already with DSD256.

 

I have not done any intensive listening test to compare yet. Early impression is for 1x I still prefer the DSD256 ext3 / 7ECv2 over the DSD512 ext2 / 5ECv2.

If DSD512 with 5EC (slightly lower load than 5ECv2) is technically better than DSD256 with 7ECv2 on Holo Spring 3 then I'll try to switch back again and test.

If not, I might just be happy with DSD256/7ECv2 and enjoy the much cooler HQPlayer Server machine.

[giordy60]

Jussi a question
when I use embedded mounted on a USB stick
can I see the functioning of the cpu and the cores used?
in short, a sort of task manager
I guess with on-screen controls ...
on linux they are very poor.
thank you.

[shadowlight]

On 1/21/2022 at 1:37 PM, Mike Rubin said:

This is exactly what causes my issues.  Thanks again for the file editing suggestion. 

The other option that you have is to go on your DHCP server and reserve the IP address for the mac address.  So the next time you reboot it will assigned the reserved ip address to the HQPe device.

 

Edit: I prefer the static IP address option on my Ubuntu system.  One of this days I will have to try the image version.

 

Edit 2: Read the later message and your XFinity router does not have the option 🤦‍♂️

[Mike Rubin]

54 minutes ago, shadowlight said:

The other option that you have is to go on your DHCP server and reserve the IP address for the mac address.  So the next time you reboot it will assigned the reserved ip address to the HQPe device.

 

Edit: I prefer the static IP address option on my Ubuntu system.  One of this days I will have to try the image version.

 

Edit 2: Read the later message and your XFinity router does not have the option 🤦‍♂️

Thanks, @shadowlight.  Indeed, it is the xfinity router that causes this issue.  Could have fixed it easily if I still had an ASUS router, which allows me to revoke assigned addresses with a mouse click.

 

My current workaround is to power down the HQPe install and delete it from the active device list before I update.  Then, I give it a reserved DHCP address at the high end of the range, e.g., 192.168.0.250.  When that address is reserved, the router won't prioritize that device at 192.168.0.19. 

 

Then, I install the new HQPe version.  The router assigns it a new DHCP address, but I then change it manually to 192.168.0.19. 

 

Because the old MAC address associated with 192.168.0.19 now has a different address reserved, I don't get the error message that the new device wants to use an address already assigned elsewhere.  Then, I change the DHCP lease time to two minutes from two days.  After a few reboots, the new version of HQPe finally becomes 192.168.0.19.  I then change the lease time back to the default. 

 

Until that happens, I can limp along with HQPD Controller and HQPW using hqplayer.local and with manually setting the current address in Roon to whatever the new installation is using before it settles in at 192.168.0.19.

[Miska]

22 hours ago, giordy60 said:

when I use embedded mounted on a USB stick
can I see the functioning of the cpu and the cores used?
in short, a sort of task manager

 

There's a "htop" tool included on the images. Works locally or over ssh.

 

[robi20064]

Miska,

 

Do you think that recent Linux kernel changes regarding TCP are going to have any (positive) effect on HQPe performance or these are never going to be significant for our standard use cases?

 

https://www.phoronix.com/scan.php?page=news_item&px=Linux-5.17-TCP-Optimization

https://www.phoronix.com/scan.php?page=news_item&px=Linux-5.17-Faster-csum_partial

 

[SwissBear]

1 hour ago, robi20064 said:

Miska,

 

Do you think that recent Linux kernel changes regarding TCP are going to have any (positive) effect on HQPe performance or these are never going to be significant for our standard use cases?

 

https://www.phoronix.com/scan.php?page=news_item&px=Linux-5.17-TCP-Optimization

https://www.phoronix.com/scan.php?page=news_item&px=Linux-5.17-Faster-csum_partial

 

 

Also, playing with the p-state of the Alder Lake Intel processor could be a promising way to improve CPU utilization.

On my system, with p-state in performance mode, I seem to have a better utilization of the active cores.

 

https://www.phoronix.com/scan.php?page=article&item=intel-12900k-pstate&num=1

[Miska]

18 hours ago, robi20064 said:

Miska,

 

Do you think that recent Linux kernel changes regarding TCP are going to have any (positive) effect on HQPe performance or these are never going to be significant for our standard use cases?

 

https://www.phoronix.com/scan.php?page=news_item&px=Linux-5.17-TCP-Optimization

https://www.phoronix.com/scan.php?page=news_item&px=Linux-5.17-Faster-csum_partial

 

I don't think the former one is very much applicable to our use cases. More for things like web servers.

 

But the latter one should have benefits also on low power NAA devices and such.

 

[Miska]

16 hours ago, SwissBear said:

Also, playing with the p-state of the Alder Lake Intel processor could be a promising way to improve CPU utilization.

On my system, with p-state in performance mode, I seem to have a better utilization of the active cores.

 

https://www.phoronix.com/scan.php?page=article&item=intel-12900k-pstate&num=1

 

This is something managed by the OS kernel. It can certainly have some benefits. HQPlayer is kind of easy workload in a sense that it creates constant load patterns that vary very little.

 

Biggest challenges are usually related to situations where there are sudden changes to CPU load and kernel needs to make decisions what to do about CPU parameters based on some predictions about future loads.

 

[SwissBear]

On 2/17/2022 at 4:45 AM, Miska said:

 

This is something managed by the OS kernel. It can certainly have some benefits. HQPlayer is kind of easy workload in a sense that it creates constant load patterns that vary very little.

 

Biggest challenges are usually related to situations where there are sudden changes to CPU load and kernel needs to make decisions what to do about CPU parameters based on some predictions about future loads.

 

 

Thanks for the answer. On my system (i9 Alder Lake and kernel 5.17rc), the workload of the most loaded cores is lower than 40% when p-state is in performance mode and close to 50% when p-state is in powersave mode. Not sure whether this is cosmetic or real. Does not make a huge difference as I still have some interruptions when trying DSD1024.

Cannot wait for the release of 4.30 to taste the results of your latest developments 😉.

[Miska]

2 hours ago, SwissBear said:

Thanks for the answer. On my system (i9 Alder Lake and kernel 5.17rc), the workload of the most loaded cores is lower than 40% when p-state is in performance mode and close to 50% when p-state is in powersave mode. Not sure whether this is cosmetic or real.

 

Load percentage is always relative to the clock speed. So 50% load at 4 GHz would equal to 100% load at 2 GHz. Which is of course same absolute load in both cases, but percentage of available computing power at given frequency is still correct. Of course when OS decides to change clocks based on the load, the load percentage changes too.

 

Technically load % calculation could take the active CPU clocks into account as well, but usually it doesn't. Because it cannot make assumptions about what will actually happen about the clocks. For example laptop running on battery vs laptop running plugged in.

 

[SwissBear]

14 hours ago, SwissBear said:

Cannot wait for the release of 4.30 to taste the results of your latest developments 😉.

 

Thank you so much for releasing this new version just before the week-end. I sincerely appreciate 🤩.

Your new high-res filter absolutely rocks for me. I never enjoyed those high-res files as much as I do tonight.

I'a also quite impressed by the progress made in terms of general performance of your application. The system is running at around 50ºC while producing DSD 512 with ASDM7ECv2. This is remarkable.

I'm once again sincerely grateful for the quality of the software you are supplying to us all and for the pleasure/emotions this software is creating for us 🙏

[pavi]

superb update, @Miska — thank you! 

[Louie]

I am going to upgrade my gear to i9 12900K, and would like to know which hpqe image that I need to use, x86gen or x86amd? 

[Bertel]

The latest version 4.30.0 version of HQPlayer Embedded doesn‘t start for me - it results in Kernel Panic (see screenshot). That also happened with 4.29.2, while 4.29.3 to my relief again worked…

 

Is there something I can fix or do? Or is this something @Miska in the packages or so that go into your image (which I flash onto an USB stick and then boot from it)?

[El Guapo]

 

7 hours ago, Louie said:

I am going to upgrade my gear to i9 12900K, and would like to know which hpqe image that I need to use, x86gen or x86amd? 

12900K should use this one: hqplayer-embedded-4.30.0-x64amd.7z

[stefano_mbp]

30 minutes ago, Bertel said:

Is there something I can fix or do?

Are you on HQPe OS?

Maybe a broken download? Or a broken burning on your usb stick?

I downloaded this

 https://www.signalyst.eu/bins/hqplayerd/images/hqplayer-embedded-4.30.0-x64amd.7z

yesterday night and it’s working fine

 

[Bertel]

Sure, HQPe OS. As said, happened with 4.29.2 as well, never before (didn‘t try 4.29.1 but 4.29.0 was fine) and 4.29.3 was fine as well. Always using balenaetcher with verification and the same USB sticks. Downloaded several times (I‘m using the amd version as well). 

[stefano_mbp]

28 minutes ago, Bertel said:

and the same USB sticks

Try with a different usb stick

[Outlaw]

58 minutes ago, Bertel said:

Sure, HQPe OS. As said, happened with 4.29.2 as well, never before (didn‘t try 4.29.1 but 4.29.0 was fine) and 4.29.3 was fine as well. Always using balenaetcher with verification and the same USB sticks. Downloaded several times (I‘m using the amd version as well). 

I use Rufus had issues with balenaetcher

[StreamFidelity]

5 hours ago, Outlaw said:

I use Rufus had issues with balenaetcher

 

When flashing hqplayer-embedded-4.30.0-x64amd (i9-12900K) with balenaEtcher I had no problems.

 

This new version I consider an important step in SQ increase. The performance has been improved and a new filter I particularly like in Redbook: 

 

poly-sinc-gauss-hires-lp (Miska's recommendation).

 

With the PCM format 44.1kHz the Apod counter goes up for many of my albums. With the new apodizing filter, some sharpness in the sound is softened without losing resolution. 

 

[pavi]

this is amazing — well done, @miska! 

 

at 192k my i9-12900k pauses for about a second roughly every 30 seconds. I have my fans set to silent — perhaps I could tinker with that. 

 

I ended up going back to gauss-xla/amsdm7ecv2 at dsd512, simply because I preferred the sound. I've absolutely no idea if there's any computational compromise or other measurement justification for my preference, but the fact that dsd1024 is comfortably possible is amazing. when I started using hqplayer seriously four months ago, this would not have seemed possible to attain in such short measure.

 

seriously — your pace of development and constant improvement is beyond impressive, @miska