HQPlayer Linux Desktop and HQplayer embedded

[Sunny_Player]

@Miska

Thanks first of all for the great 5.1 update. What settings would you recommend fir max space in the sound? Currently using amsdm512+ with poly sinc gauss xl(a). Thank you. 

[Sunny_Player]

7 minutes ago, Miska said:

Which DAC do you use? Have you tried ASDM7EC-light or ASDM7EC-super, in regular or 512x fs variants?

Thanks for your answer. 

 

-T+A SDV 3100 HV

-Light or super leads to drop outs in 1024 dsd, 512 works

Would you recommend higher rates over better modulators or vice versa?

[Sunny_Player]

9 minutes ago, Miska said:

 

Once you have reached DSD256, it is better to use better modulators rather than attempting to reach higher rates.

 

Thank you Miska.

 

What is your personal preference for space-optimized sound?

[Sunny_Player]

@Miska

Good evening, 

 

seems I got stuck somehow. How do I manage to change modulators and filters on the fly in hqplayer embedded? No clue... Thank you!

 

Best regards

[Sunny_Player]

Dear all and @Miska

 

happy new year first of all!

 

Just a quick one. My spectrogram shows a horizontal line at around 19khz (see below). How can this be avoided? Is it caused by the switch mode power supply only or is it caused by rhe processor itself? How does your spectrogram looks like? 

Thank you!

[Sunny_Player]

Just now, Miska said:

 

Happy New Year!

 

That can be from an SMPS in some device used for the recording. For example the ADC. Usually such are higher frequency around 40 kHz though. That frequency refers more to a CRT display horizontal sync signal, likely from display of the recording computer.

 

Thank you Sir. Is it possible, that my computer SMPS is responsible for that? Some claim, that linear power supplies sound better...? 

[Sunny_Player]

4 minutes ago, Miska said:

 

No, what you are looking at is analysis of the source data. It is not related to any of your hardware, but rather hardware used for the original recording.

 

Since it is baked into the original recording data, you cannot do much about it. If you like, you can use the matrix processor to place a 19 kHz notch filter there. But likely that will do more harm than good, unless you are actually bothered by the whine.

 

Thank you, perfectly understood. At the beginning, it was my understanding, that ripple noise is causing this line. Btw.: What is your opinion on that point? Does a linear power supply really improve the sound quality? 

 

[Sunny_Player]

First of all: 5.3.2. works flawlessly with all filters at 1024 DSD in conjunction with the new AHM7EC5L modulator. @Miska congrats, amazing! 

When applying sinc-long, it does produce some drop outs @192 khz and above. As this is the best sounding filter (airiness), I am looking for a comparable filter. Any recommendations from your side? 

 

[Sunny_Player]

On 2/8/2024 at 7:27 PM, StreamFidelity said:

The outsourcing of the FIlter to the E-Cores is fantastic. The outsourcing shows an extremely even load distribution on the E-Cores (17-32).

 

I can now play more demanding filters in my fanless audio PC. The increase in SQ is outstanding. Thank you for this new feature.

 

@MiskaMulticore = 1 and ecores = filter does not bring any advantages on my side. (core i9, 13th generation, hqplayer embedded). Sinc long is stuttering @SDM 1024 and 192khz pcm as well as sinc mx. Is there anything more to add to hqplayerd.xml than ecores="filer"? thanks for your support.

[Sunny_Player]

On 2/7/2024 at 9:39 PM, Miska said:

 

 

 

This is with multicore="1" ecores="filter" cuda="0"

 

I do not even come close to your values. with multicore="1" ecores="filter". AHM7EC5L is not working @1024 (core i9, 13th gen) with sinc-L/sinc long/sinc LI. I do not know why... Tested almost all night. With 5.4, sinc L and sinc LI does work properly. 

 

[Sunny_Player]

@Miska I just switched back to 5.4. With 5.4, SincL and Sincl LI work flawlessly with multicore = auto and 1 with 1025 PCM 192 khz SDM, core i9 13th gen, no CUDA. Do you have any idea why this does not work with 5.4.1? 

[Sunny_Player]

3 hours ago, Miska said:

 

With same settings I'm using? Those figures are from minimal Fedora 37 Server.

 

 

Which build / OS are you using?

 

I am using hqplayer embedded 5.4.1, now 5.4 again, as sinc-L and other heavy filters work flawlessly with multicore="1" and 5.4. PC: Only hqplayer embedded with core i9 13th genereation, 32GB RAM. 

[Sunny_Player]

17 minutes ago, Miska said:

 

Yeah, but which build?

 

I'm using poly-sinc-gauss-long / poly-sinc-gauss-hires-lp.

 

Thanks for your reply @Miska

 

I am using 1024 dsd upsampling with ahm7ec5l with sincL or sincMx in hqplayer embedded. Flawlessly with 5.4, but dropouts in 5.4.1

[Sunny_Player]

5 minutes ago, Miska said:

 

OK, but can you please answer to my question?

 

Without the build information, this other information is meaningless.

 

Sorry, misunderstanding. What do you mean with build information. Sorry for any inconvenience,  I do not get this properly. 

[Sunny_Player]

@Miska

I picked x64AVX2 as usual. Is this the information you need?

[Sunny_Player]

Just now, Miska said:

 

Yes, which OS? With multicore=1?

 

Hqplayer embedded plain vanilla, no os, just embedded. Multicore = 1

[Sunny_Player]

3 minutes ago, Miska said:

 

OK, so HQPlayer OS. Then I don't know where such difference could be coming from...

 

Could it be that the scheduler is not working properly? I am using convolution as well. Bit I have no clue why everything works under 5.4 and there are those issues under 5.4.1

[Sunny_Player]

2 minutes ago, Miska said:

 

No, the OS is the same. And unless you changed the ecores setting on 5.4.1, the other functionality is the same too.

 

Thank you, I will test 5.4.1 without the ecores parameter and revert. 

[Sunny_Player]

11 hours ago, Miska said:

 

 

13900T has 8 performance cores and 16 efficiency cores. Two efficiency cores is roughly similar in performance to one performance core. This is 35W model, so it is TDP limited low power CPU variant.

 

@MiskaSo, after testing, there must be an issue with in 5.4.1. Configuration: 1024 SDM, 192 khz PCM, sincMx, Core i9 13th gen K, avx2 build. In 5.4, every filter works fine, but with 5.4.1. the performance decreases massively and sinc-Mx produces drop outs. 

Amendment: with convolution changed to add, Mx produced no dropouts anymore. In 5.4.1. Ecores entry deleted. This works well for me. Although, this was no issue with 5.4. 

[Sunny_Player]

12 minutes ago, Carousel said:

But I have been wondering if I could also pass DSD256 in DoP through the Mac Mini to the DAC. Played a bit with it today. But no joy.

Let me think about it. I have an old mac mini at home and will try to duplicate your setup as good as it gets. Let me see if i find a solution. Keep you posted. 

[Sunny_Player]

@Miska thank you very much for your support and this brilliant cuda offloading feature. I just set up an Ubuntu server with respective capable GPU and it works perfect with 7ECv3 @dsd 1024. Using your kernel as well. Sinc long stutters, of course. But I do not want to spend 2000 EUR for an nvidia 4090. Is there anything one can do to boost performance even further? Multicore =1, ecores =filter and nblocks =16. 

[Sunny_Player]

On 2/27/2024 at 11:11 PM, Miska said:

 

 

 

1 hour ago, EMINENT said:

 

Holoe.

 

Used 3090 is less expenditure and reaches sinc long with 14900k.

yes, that's right. But I do not have sufficent space in my PC as I am running a massive tower cooler and a Solarflare SFP-card in it... and the 4070 is the max size that fits currently. Otherwise I need to switch to water cooling...

 

[Sunny_Player]

Good morning @Miska and all. Just a quick one: I am playing a 24/192 source and the apodizing counter is above 12000. What is the reason for this behaviour? Thank you. 

[Sunny_Player]

On 3/19/2024 at 8:20 PM, Miska said:

 

If you like those filters, use the poly-sinc-gauss-xla (similar to sinc-MGa) or poly-sinc-ext3 (similar to sinc-Mx) instead.

 

The poly sinc filters perform a whole lot better than MG or Mx and sound to my ears even better. No reason why one should use MG or Mx. 

[Sunny_Player]

6 hours ago, chipvn said:

@Miska: it turned out that either HQPe OS or the mainboard got confused due to my CPU clock, power setting changing in the MSI Z690M motherboard Bios.

Physical Power cycling everytime after changing CPU clock, CPU PL1/2, CPU voltage will resolve the issue.

Thanks for the hint, I have the same mainboard. Could I ask you to elaborate on the changes you made in the bios? Thank you.