Thanks, that clarifies it.
I got it working in the end, after realising that i should also install the ASIO USB mac OS driver for my Esoteric DAC, which needed modified security permissions in MAC OS to add kernel extensions.
I also installed the Esoteric HR Audio Player in case there is some dependency to that.
And then also first select USB as the input on my DAC, wait till it says ‘USB ready’ otherwise the Esoteric does not show up as device in the output list.
One question I still have is how to setup Roon device settings for HQplayer.
E.g. MQA capabilities
My DAC is MQA capable for full unfolding. I believe Roon isn’t nor HQPlayer, right? So best to keave this as No MQA support?
Fixed Volume?
This shouldn’t be necessary. HQPlayer doesn’t need or use these.
Yes, having any other input selected turns off the USB module and thus it is effectively same as disconnecting the USB cable.
Thanks Jussi,
I think I noticed that without the ASIO usb driver, the Esoteric didn’t show up as output device.
May I ask why you’d use device volume and 5.1 channel layout?
Thanks again.
Regards
Device volume here means that Roon adjusts HQPlayer’s volume. Since HQPlayer is “output device” from Roon’s perspective. 5.1 channel layout with channel mapping is so that multichannel files are correctly passed over to HQPlayer.
Hi Jussi, I have a question.
Is it somehow possible to set HQPLAYER to upsample depending on the source bitrate?
So, to set upsampling to 2x or 4x instead of setting the max bitrate. Maybe it’s possible to then also choose which filter to use (for PCM) for each output bitrate.
That way I can setup HQPLAYER to automatically setup for each type of source rate instead of 1 set of settings for all, as I understand how settings work now
Thanks
You have two filter settings, one for 1x source rates, and another for the rest.
Why would you want that?
Thanks Jussi,
Well, it’s because I’d like to not upsample too much; i.e. from 44.1kHz to 384kHz is serious interpolation overkill and in my experience degrades music stage and air around positioning.
I therefore tend to like upsampling to 176 or 192 for 1FS recordings and go higher is the source is also of high rate, like 88.2 and up.
Regards
But point is to choose output format that is sweet spot of the DAC’s conversion and analog stages. So it is not related to the source format, but rather DAC’s hardware.
Im looking forward to the M4 Mac minis. Wonder how they will do with HQplayer 5
The M4 shows a ~43% increase in single core performance compared to the M1. I am hopeful that the M4 Mac mini will do DSD512 with the ASDM7EC-super modulator and poly-sync-gauss-xla and poly-sync-gauss-hires-lp filters…
If there is indeed such increase then it is pretty notable. M1Max can almost do that, output buffer underrun happens once per minute or so.
Just fairly low number of cores is limiting the basic Mx CPU. But probably M4Pro would be already pretty capable. And at least M4Max should be able to do that.
The single core Geekbench 5 result for the M4 is 2491. The Intel Core i9-14900KS is at 2484.
Synthetic benchmarks are pretty much useless. I don’t know what kind of instruction mixture that is doing. It is likely just totally different than what you get in the end with a different workload such as HQPlayer.
But it is also worth taking into account that 14900KS has maximum clock speed of 6.2 GHz. So far, Apple silicon has been about half of that. And NEON is capable of about half of what AVX2 is capable of. Thus the rough end result is about 1/4th per core speed, unless it can do 4x as many calculations per clock cycle. And then 14900KS has 8 P-cores and 16 E-cores. So 14900KS is like M1Max in terms of P-cores, but has 8 times more E-cores.
So we’ll see…
The max clock speed of the M4 is reported to be 4.4GHz…
I found this quote on a web site:
" One coder experimented by rewriting his code for an existing ffais program to use SVE and saw a peak execution speed increase of 2.26x over his original code!"
This was using SVE on an M3 which appears to be ARMv8.6-A compliant. The M4 is ARMv9.4-A compliant and supports SVE2. So, you might expect it to be even fast on the M4.
If the M4 is a fast as suggested, AVX2 has nothing on the SVE2 in the M4 and the 4.4GHz clockspeed is a big improvement.
Yes, synthetic benchmarks are not necessarily useful. But, discounting the M4 like you have without really understanding what it offer versus the M2 and M1 is not sensible. I bet you could write a benchmark program that would be quite useful at predicting HQPlayer performance with any given CPU.
Fortunately Apple Store allows free returns for 2 weeks in most places.
Anyone interested in M4 can simply ignore all online benchmark discussions, then buy from Apple Store and test HQPlayer in the comfort of own home…
Return to Apple store for full refund if not satisfied.
It is already there, you can run HQPlayer to process to a null output backend. And HQPlayer status bar reports the total processing time.
But we’ll see how M4 performs with HQPlayer workloads once there is something out to test on. So far M2 and M3 didn’t bring notable improvements in this respect.
Did you rewrite your code to take advantage of SVE on the M3? I suspect you did not. Also, you would need to rewrite your code to take advantage of SVE2 on the M4. Are you going to look into that??
Otherwise, you will see no gain. In other words, if we test the current HQPlayer on the M4 without it using SVE2, performance is being left on the table.
