You can have a low average but have 2 x cores running at 100 deg C and 4 x cores running at 65 deg C
I’d keep an eye on max continuous for CPU health
An interesting thing with TG Pro is I’ve found stock macOS to be quite crap with fan control.
With some manual changes in TG Pro I’ve increased the minimum speed (still stays quiet enough) and this helped a lot with Less fan speed fluctuations and more importantly CPU temps all lower.
Is anyone using ASDM7EC at sample rates above DSD64 for multichannel listening? I use an Okto dac8pro with poly-sinc-ext2 and grayed adaptive output rate and I can’t go to the full DSD128 on the DAC for MCH without stuttering, even with an i-10900K processor and an EVGA RTX2080Ti graphics card for CUDA. JCR
My bad, @dabassgoesboomboom. 8 channels — 5.3 configuration. I am running 8 channels of Audiolense convolution filters through the HQP Pipeline Matrix. And no, ASDM5EC does not work, either.
I can do ASDM7EC for 2 channels at DSD512 with the same settings otherwise.
It’s been a while and I have wondered what others may have tried with, at least, a 5.1 configuration. JCR
Does it work with only 5 channels? So disabling the 3 subs (for testing purposes)
From memory 8 channels will use 8 CPU cores for EC modulators but DSD128 shouldn’t tax 10900K I would think
This needs @jussi_laako because there aren’t that many doing multichannel with DSD EC modulators
Is your GPU offloading actually working? Can you see GPU is actually busy ?
Also, what is your CPU cooling like? Can you monitor CPU temps? Not just average but individual core temps. Although with 8 channels with EC modulators, I imagine average will be high too.
Disable GPU offloading in HQPlayer - does DSD128 work?
Also what I meant about only 5 channels wasn’t regarding source material, but if you only output 5 channels (to test using only 5 cores of CPU for EC modulators instead of 8)?
I see your CPU has 10 physical cores with high turbo boost speed, which is plenty assuming good cooling.
But try 2-channel music with GPU offloading off at DSD128
Interesting, so far people haven’t got DSD512 with the EC modulators yet, you are the first one to achieve that. Do you have a screenshot of your settings?
Is this the DSD64 source version? It would be good to establish a baseline first with 5.1 channel 44.1k PCM source.
Doing convolution for 8 channels of DSD64 source content will be very heavy. So you may want to try with offloading just convolution, by checking “CUDA Offload” to grayed.
With GPU offload and EC modulators, you should be getting high load on 8 cores and lower load on two other cores, and almost no load on the 10 sibling threads.
Do you by any chance have “Multicore DSP” checked? You should set it to grayed instead.
In addition, company called Crystal (nowadays called Cirrus Logic) had a 1-bit DAC chip CS4328 which integrated digital filter, modulator and a DAC on a single chip. This one was running at 2.8 MHz aka DSD64 rate. I still have a DIY DAC I built back then in early 90’s based on this chip. (I also have my DIY multi-bit R2R DACs from same era, based on Burr-Brown DF1700 + PCM1700 and DF1700 + PCM63P).
These days, companies rarely talk about conversion technology at all.
Question regarding how HQPlayer handles Filter Nx settings for PCM outputs:
If Nx = “none”, my understanding is there is no up sampling or noise shaping applied, only bit depth increase (padding with 0’s?) if DAC bits is set higher. And if volume is set lower than 0 (ie -4dBFS), then there is some volume adjustment. Is that it?
What happens if the input sample rate == output sample rate. For example if max PCM rate = 192kHz and the input is 192kHz. Is the up sampling / shaper bypassed the same way it is with “none”? Or is there more processing that happens?
I’m asking because I would like 96kHz to be up sampled to 192kHz, but I would like 192kHz to just be left alone and passed through “unchagned”.
If Nx filter is “none”, then hires content is not upsampled. But same dither/noise-shaper and all other DSP processing applies.
No, the filter is applied to the content as it would be in other cases too. Everything runs through 1:1 rate conversion in such case. This makes mostly difference when filter is an apodizing one.
There’s no such option. Only such option there is, is the Direct SDM for DSD content. With Direct SDM only speaker distance processing is still applied (when in use).
@jussi_laako
Regarding PCM noise shaping: I have a DAC that is limited to x8 PCM. I keep going between NS5, NS9, and LNS15 since these are the three that seem to be most applicable at this rate. Do you have one you recommend over the other at x8 PCM? Is it possible to plot the performance of each of these noise shapers at x8 PCM (and x16 PCM for that matter) to see at what frequency the noise starts to ramp up?
