You could choose some lighter filters for “Nx” rates, like poly-sinc-2s. And keep gauss-long for “1x” rates.
Thank you, Jussi. That is exactly what I have done, except that I chose m4a/mp3-mp for the Nx filter. Maybe the next generation of M1 chips will solve the short-term problem. My ultimate ambition is to get to DSD512 with ECmk2 
Thanks again for your work on the new modulators. I can’t really believe my luck that I get such an super upgrade from you as part of my original purchase! It will be interesting to see what you put into the v5 series. Just take my money already
Hi Nick, with this are you able to do other stuff on your M1 at the same time?
Or it only works if M1 is doing nothing else?
And I have same query about the original ASDM5EC (v1)
Hi Mr Bass
My M1 is dedicated to HQP and I have another Intel Mini that runs Roon core.
The good news is that I am able to upsample to DSD256 using the ASDM5ECv2 on all source material, just by swapping the Nx filter choice. I now use gauss-long for N and mqa/mp3-mp for Nx. To be honest, it all sounds great, and a major step forward in SQ.
In the future, I may sell the M1 and replace it with the next gen system from Apple. But am waiting to see whether it will support upsampling to DSD512 using the Mk2 EC filters before I take the plunge.
Good luck with your choices!
N
1 Like
Using the Mk1 EC modulators to upsample to DSD512 requires a 10Ghz CPU. These don’t exist outside of liquid helium baths in research labs. I understand the Mk2 EC modulators are slightly more demanding. I don’t expect Apple next gen to reach the required speeds.
1 Like
This formula no longer applies to newest chips coming out
Previously we needed >4GHz running speed for EC-v1 modulators at DSD256
Apple M1 changed that formula
See Jussi’s post below
HQ Player - Page 887 - Software - Audiophile Style
It boils down to game changing cross-core communication efficiencies
i9-12900K may be wildly different to previous Intel CPUs.
Even if it can’t do DSD512 with EC modulators it may do DSD256 with greater ease.
That might be useful for easier cooling.
Also might be useful for people that want to use their computer at the same time as HQP to do other things. More headroom
We can’t use that older formula for newest chips, especially Apple and maybe Intel/AMDs newest. We have to wait for real HQP field data
1 Like
I wonder if anyone has tried HQP on the new M1X(pro) MacBooks. I know they won’t do DSD512 with EC modulators but curious to see if HQP can take advantage of the additional performance cores. (vs the 4 in original M1).
Yes, I’ve received some reports of the new M1 chips. On first gen M1 all the four performance cores are occupied by the modulators and efficiency cores are left to deal with filters. On newer generation M1 chips, essentially you can run more complex filters together with the same set of modulators. And still do lighter tasks like web browsing on the efficiency cores.
4 Likes
Anyone successfully utilizing the v2 EC modulator on an M1 Mac Mini? If so what are your settings?
I played around with some various combinations this morning but am now back on ext2 / ASDM7EC (v1). ASDM7ECv2 chokes on anything higher than 88.2 kHz for me (regardless of the filter choice on the Nx setting)
I am getting similar results on my M1 mini. ASDM7ECv2 with 48/96 → 48 x 256 results in dropouts every 10 seconds or so. As per @jussi_laako above, the filters run on the 4 efficiency cores and won’t really make a difference here. The problem is that the 4 performance cores seem to be ever so slightly underpowered to do ECv2 modulators with 48x256x. But 44.1x256x is ok. Frustrating as it’s so close!
Right now, I just have max output set to 44.1 x 256. Not sure I should be using adaptive output or not. Alternatively, I assume that ASDM7ECv1 + DSD256 >>> ASDM7ECv2 + DSD128.
If only @jussi_laako could optimize 1-2% more performance from ASDM7ECv2 … ha.
1 Like
Why not try and hear if it works?
Which USB DAC model?
This is a complicated beast!
15 pages of “my head hurts”.
Can Jussi tame this beast ! 
Would be interesting to hear from someone with M1Pro or M1Max if it works better there. Apart from adding more cores, there’s also improved speed on the unified memory.
Have to see how OS will expose the different cores and how to tell which one is performance and which one is efficiency core. This is important for HQPlayer’s load distribution. So I may need to do some specific tunings to HQPlayer code.
In any case, I have i9-12900K on order, now just waiting for all parts to arrive.
2 Likes
Holo May, which accepts 48k DSD rates. Setting the max output rate to 44.1x256 instead of 48x256 certainly helps a bit with ASDM7ECv2.
I have also found that the filter does matter a bit on the M1 Mac when using ASDM7ECv2, even though the efficiency cores look to be running fairly low load. For example, I can run up to 96k > 44.1x256 with poly-sinc-gauss. But poly-sinc-gauss-long pushes it over the top as does 96k > 48x256.
I have only just recently started trying DSD upsampling since I got the M1 so still trying to untangle all of the permutations. For example, and maybe this is a dumb question, but what would be better of these two options for a given source: (assuming the 192k version is truly 192k):
a. 192k > DSD128
b. 96k > DSD256
1 Like
But why not tick 48k DSD box, select output rate DSD256x48 and tick the “adaptive rate” box?
Then you should have integer ratios for all content (if your chosen filter supports it - poly-sinc-gauss family does)
1 Like
Keep your favourite filter for 1x but can you try Nx = poly-sinc-short-lp-2s ?
Does that work better? With adaptive rate box ticked
Because … with the same gauss filter and ASDM7ECv2, 48k/96k > DSD256x48 results in dropouts, but 48k/96k > DSD256x44.1 does not result in dropouts. Simple as that.
So I assume I am better off NOT using adaptive output rate and getting say 96 → DSD256x44.1 instead of 96 → DSD128x48. Does an integral ratio matter with DSD upsampling?
1 Like
This is strange.
It matters more! Helps reduce CPU loading. Which is a bigger deal with DSD upsampling than PCM.
Same behaviour with ext2?
Try my Nx suggestion also
When you have 48k to DSD256x48k failing, can you share screenshots of all your HQP settings ?
Ok you are right on the CPU load and that makes total sense. With the same poly-sinc-gauss filter and ASDM7ECv2, 96k > 48x256 uses about 5% less CPU overall than 96K > 44.1x256.
Yet I get occasional dropouts with 48x256 and I don’t with 44.1x256.
?!@#
1 Like