HQ Player

[Miska]

1 hour ago, Zauurx said:

It's down for me !!! 😒 

 

Same settings, same tracks / poly-sinc-gauss-long / DSD256x48 / ASDM5EC

 

 

HQP 4.15.3

 

 

 

HQP 4.15.4 :

 

 

 

 

I wonder why your core 0 is idling... Is this stock Windows 10 Pro? Can you email me HQPlayer log file?

 

Did you report results for the two test builds between these two releases?

 

[Miska]

5 minutes ago, GMG said:

I tried 0 & 250 from UI and i even tried higher values by manually editing the configuration file - not sure if it actually uses the higher values, but I tried even 10000. Didn’t help

 

Try 100...

 

Going over 500 will certainly cause a lot of problems. And I'm assuming you don't have short buffer enabled.

 

 

[ted_b]

7 minutes ago, Miska said:

 

Those additional matrix fixes and few more optimizations.

 

If you want, you can change the behavior by setting following environment variable:

DSP_ACCELERATION=0

 

Default value, when not set is "1".

What does that variable do, if I set it to zero?  I do no (currently) convolution or matrixed processing, just PCM to 32fs and DSD to 256 using 7ECV2.

[Miska]

52 minutes ago, Zauurx said:

Because your CPU has a base frequency of 4ghz... an i7.
Mine is a brave i5 but base at 2.8ghz, oc to 4! 
That's all folks !! a little more is enough.

 

It's not about the base clock, but how much Turbo boost you can have. I can run ASDM7ECv2 to DSD256 on my i5-11600 with hefty margin, so it is not about the base clock alone. Also works on my i5-7600T. Bot running Ubuntu Server 20.04 and HQPlayer Embedded.

 

[GMG]

4 minutes ago, Miska said:

 

Try 100...

 

Going over 500 will certainly cause a lot of problems.

 

Now I tried. 
same….

 

if the source is 192 and not 44.1 then there is no issue with ALSA. 
 

is there an inherent difference in performance/load to be expected between ALSA and NAA, or should I expect to be able to use the same configuration with either backend?

[bogi]

20 minutes ago, Miska said:

If you want, you can change the behavior by setting following environment variable:

DSP_ACCELERATION=0

 

Default value, when not set is "1".

Should I see any change between values of 0 and 1 with my CPU? I tried to change the value in Control Panel (of course while HQPlayer process was not running) but I didn't observe any change. No complains to 4.15.4 - I have none. :)

[Miska]

11 minutes ago, bogi said:

Should I see any change between values of 0 and 1 with my CPU? I tried to change the value in Control Panel (of course while HQPlayer process was not running) but I didn't observe any change. No complains to 4.15.4 - I have none. :)

 

Yes, there should a bit of difference. Roughly same amount as between the two test builds.

 

[Miska]

26 minutes ago, GMG said:

Now I tried. 
same….

 

if the source is 192 and not 44.1 then there is no issue with ALSA. 
 

is there an inherent difference in performance/load to be expected between ALSA and NAA, or should I expect to be able to use the same configuration with either backend?

 

So are you upsampling to some fixed 48k multiple?

 

Not, much. With NAA you get a bit of extra buffer. But overall the difference is pretty minimal.

 

Are you on HQPlayer OS or something else?

[GMG]

2 minutes ago, Miska said:

 

So are you upsampling to some fixed 48k multiple?

 

Not, much. With NAA you get a bit of extra buffer. But overall the difference is pretty minimal.

 

Are you on HQPlayer OS or something else?

I am upsampling to fixed 768 per your advice for Chord DAC. 
I am running on Archlinux (AudioLinux)

[Miska]

1 minute ago, GMG said:

I am upsampling to fixed 768 per your advice for Chord DAC. 
I am running on Archlinux (AudioLinux)

 

OK, I think you need to ask AudioLinux people about the details then.

 

[GMG]

10 minutes ago, Miska said:

 

OK, I think you need to ask AudioLinux people about the details then.

 

I am. 
in parallel I will try and boot your OS on another computer and see if I encounter the same issue or not. 
thanks Miska

[Miska]

31 minutes ago, GMG said:

in parallel I will try and boot your OS on another computer and see if I encounter the same issue or not. 

 

You can also boot it in the same computer, since it doesn't require installation.

 

[Yviena]

Hmm EC7v2 DSD512 ext3 seem to be possible on 5900x without cuda too on windows, tho sometimes sound drops for 1-2 seconds and all load on primary threads done by HQplayer is lost for those seconds then it recovers. CPU usage also in not a straight double of DSD256 

[Miska]

5 minutes ago, Yviena said:

Hmm EC7v2 DSD512 ext3 seem to be possible on 5900x without cuda too on windows, tho sometimes sound drops for 1-2 seconds and all load on primary threads done by HQplayer is lost for those seconds then it recovers. CPU usage also in not a straight double of DSD256 

 

It could be some driver. WiFi and BT drivers are prime suspect on such events. There have been drivers that disable interrupts for really long periods of time when they are periodically scanning for networks/devices.

 

Also check what is the content source. Sometimes it could be that too.

 

[Bertel]

With all reports of failure, stuttering etc., just want to say: also 4.15.4 is working perfectly fine for me, with brilliant sound. But I‘m only doing 32bit 786kHz with poly-sinc-ext2 and NS15 to my Topping D90SE (which I meanwhile clearly prefer over my Holo May KTE), using a 12th gen i9-12900K with 64GB RAM, directly connected over USB using ASIO.

[Fleecyears]

Hi Miska,

 

Where or how can I get an AMD Ryzen Optimised version of Ubuntu 21.10.
 

Thanks 

[Miska]

31 minutes ago, Bertel said:

With all reports of failure, stuttering etc., just want to say: also 4.15.4 is working perfectly fine for me, with brilliant sound. But I‘m only doing 32bit 786kHz with poly-sinc-ext2 and NS15 to my Topping D90SE (which I meanwhile clearly prefer over my Holo May KTE), using a 12th gen i9-12900K with 64GB RAM, directly connected over USB using ASIO.

 

You get most consistent performance out of that D90SE DAC with ASDM5ECv2 at DSD256... Remember to set the DSD filter to 50 kHz.

 

[Miska]

13 minutes ago, Fleecyears said:

Where or how can I get an AMD Ryzen Optimised version of Ubuntu 21.10.

 

Under "Other downloads". Although short term releases like 21.10 are not officially supported (only LTS).

 

[Bertel]

23 minutes ago, Miska said:

 

You get most consistent performance out of that D90SE DAC with ASDM5ECv2 at DSD256... Remember to set the DSD filter to 50 kHz.

 

Following Archi‘s measurements (see here: http://archimago.blogspot.com/2021/10/measurements-review-topping-d90se-dac.html ) I‘m not using the D90SE with DSD, as it measurably performs substantially worse with DSD compared to PCM (also fully consistent with my subjective experience). Using 32bit instead of 24bit with PCM, as you suggested, for me clearly performs/sounds best.

[Miska]

15 minutes ago, Bertel said:

Following Archi‘s measurements (see here: http://archimago.blogspot.com/2021/10/measurements-review-topping-d90se-dac.html ) I‘m not using the D90SE with DSD, as it performs substantially worse with DSD compared to PCM (also fully consistent with my subjective experience). Using 32bit instead of 24bit with PCM, as you suggested, for me clearly performs/sounds best.

 

I measured it, results here:

https://community.roonlabs.com/t/topping-d90se-hqp-settings/183551/6?u=jussi_laako

 

It is certainly most consistent performer at DSD256. PCM has host of issues, varying from noise floor modulation to poor low level linearity.

 

Slightly higher noise floor (and even this only in some cases) consisting of white noise is much less of an issue than all the distortions and anomalies with PCM.

 

[Bertel]

11 minutes ago, Miska said:

 

I measured it, results here:

https://community.roonlabs.com/t/topping-d90se-hqp-settings/183551/6?u=jussi_laako

 

It is certainly most consistent performer at DSD256. PCM has host of issues, varying from noise floor modulation to poor low level linearity.

 

Thank you, very interesting! Will need to thoroughly compare it to Archi's measurements, from my understanding they indicated something different, hope I can find a consistent bottom line...

 

Do you? I mean, can you see some alignment of your measurements with Archi's? Would be suprised if they differ so substantially.

[Miska]

8 minutes ago, Bertel said:

Thank you, very interesting! Will need to compare it to Archi's measurements, from my understanding they indicated something different, hope I can draw a bottom line...

 

For me it's in same category as Chord Mojo earlier. Disappointment that will likely stay unused...

 

[Bertel]

Hmm, hard for me to understand why Amir (ASR), Wolf (I7lab), Archi and others measure it so differently/superior…

 

EDIT: well, they do 16/44.1 up to 24/384 max, maybe that‘s why?

[dericchan1]

I suppose my ifi neo idsd dac is not as much well received as some of the dacs like the toppings. I would love to see someone do up the measurements for it but seen none so far!!! Sounds great to my ears tho especially with dsd256 with 7ECV2!!!

[Zauurx]

3 hours ago, Miska said:

 

I wonder why your core 0 is idling... Is this stock Windows 10 Pro? Can you email me HQPlayer log file?

 

Did you report results for the two test builds between these two releases?

 

 

There is nothing on the core 0 because windows default is to run multiple processes and to optimize latency, I prefer not to add anything. (Win10 LTSC)

But I tested with HQP on the 6 cores and with betas 1 and 2... 

Same thing: 90% cpu. 

I have a matrix convolution... even if I lighten up, same result. 

I will send you the log after the weekend. 

Thanks for the follow up... it sounds so good in dsd256 with my Pegasus.