I get something similar in terms of performance on my i7 9700 also upscalling to DSD 256, I am tempted to overclock my PC again to see if I can get it running. I have set up like @my_changed_username suggested for now.
Results are for i9 9900k on macOS 10.14.6 (processor seems to be throttling happily between 3.5 and 4.9Ghz) Hyperthreading is enabled so double the % to get a more accurate usage of the 8 cores.
poly-sinc-ext3 ASDM7EC
18.2% 44.1/16 > DSD256
25.0% 88.2/24 > DSD256
27.0% 96/24 > DSD256 48K
37.7% 176.4/24 > DSD256
41.0% 192/24 > DSD256 48k
12.5% DSD64 > DSD256
13.2% DSD128 > DSD256
@jussi_laako Iām not sure what you did but compared to the previous builds, v4.11.0 is working great on my hackintosh. I have not changed any of the config and I am now able to use all but 2 (sinc Mx and sinc L) of the filters with ASDM7EC upto DSD256 48k rates from the OS X appā¦ no more windows!
I am currently giving the cpu a workout with the new ext3 filter. Itās a bit heavy on all cores but is working flawlessly and sounds great.
I also like the addition of the anodizing detection count. Donāt remember seeing that in the previous build.
Thanks, that is just what I wanted, working on SDM (DSD) for me.
I like ext3 for 1x. Transients seem to have a bit more oomph. I suspect that the difference between ext2 and ext3 would be less pronounced for Nx as the Nyquist frequency is shifted further away from audible range.
Yes, this is what the ā1xā and āNxā filters are for. Filter selected for ā1xā will be used for sources having sample rate <= 50 kHz. While filter selected for āNxā will be used for sources having sample rate > 50 kHz.
It is not black and white figure. But roughly; if you get figure below or around 100 during a song then you donāt necessarily need an apodizing filter. If you get figure over 1000 during a song then you certainly need an apodizing filter.
If you are using apodizing filter on content that doesnāt necessarily need one - completely fine - no harm. If you are using on-apodizing filter on content that needs one, then you are reproducing errors in the source content and your output signal is not a good representation of the original analog signal.
Note: the detector works only for original or down-sampled source content.
Yeah, it can be somewhat heavy, but works nicely on the newer 8+ core CPUs.
Note that oversampling filter selection doesnāt apply for these cases. Only SDM integrator and conversion algorithm selection in āDSD Source Settingsā and modulator choice.
Seems to be continuing mystery, since I cannot quickly figure out what it could beā¦ But I hope it will stay this way.
What is it that the Apod counter detects in a file to make it tick up? Im guessing counts around 200 to 600 probably mean an apodizing filter is necessary.
For the following setup:
Core pc running hqplayer
Ethernet
Raspberry pi4 NAA
Usb to dac (Rockna Wavelight, capable of dsd512 on usb)
Anyone know why I canāt up sample to dsd512?
I have tried ext2 and ext3, asdm7ec, dsd 64 and 128 without issues. But when I set hqplayer to 256 or 512, roon just plays back as 128.