HQPlayer Parametric EQ - large processing requirements for SDM -> SDM playback

Audio Gear Talk HQ Player
1

Hello,

I recently started using HQPlayer and am loving it! I am using the plugin with Roon and streaming to my Gustard A26 DAC (DSD Firmware fix installed) at DSD256. When upsampling from PCM, I can apply any combination of filters and shapers while also using Parametric EQ in the pipeline and output at DSD256.

The only issue I am having now is applying pipeline Parametric EQ to my DSD256 source recordings. This will always result in breakups and pauses during playback. Is there any way I can reduce the processing requirements for SDM to SDM w/ EQ? Right now I am simply exporting my EQ profile from REW and loading it into each channel in the pipeline. Is there a way for me to manually tweak the parameters to optimize for SDM to SDM?

2

Processing EQ and such for DSD256 sources can be certainly heavy! I have some improvements planned for this case in future release.

How many parametric EQ’s does you current REW export have? You can try to reduce number of EQ’s included there by applying more smoothing to the measured response in REW which results in fewer and gentler EQ’s.

You can also easily have two different EQ sets stored as matrix profiles which you can switch on the fly using for example HQPlayer Client. So for other material you can keep using your existing EQ.

3

Thanks for the response! I am using five bands of Parametric EQ for my headphones, just to smooth over a few treble peaks and shelf the treble down overall. I am not currently using speakers, room correction, or anything like that yet. I will try running a set with just the treble shelf and see if that will play without interruptions.

4

OK, that doesn’t sound too bad. I have some headphone EQ’s with 10 bands.

ARM64 architecture is not as good on this area as Intel/AMD, so it depends on the hardware somewhat. Are you on a Mac?

I have some potential improvements planned in this area, let’s see if it will work out.

5

I am on Windows 10 with an Intel i9-10850K CPU, an NVIDIA 3080 GPU, and 32GB of RAM. Even with multi-cores and CUDA activated, I cannot apply any parametric EQ to DSD256 → DSD256 playback without stuttering. I can run almost any combination of PCM to DSD256 + EQ without any issues. Good luck on the improvements!

6

Hi Jussi, I discovered something interesting that may help you. I am still troubleshooting EQ of DSD256 input to DSD256 output. Even if I have a blank pipeline with zero bands of EQ and no crossfeed, but I check the “Enabled” box in pipeline, the song will still have stuttering and my GPU (CUDA enabled) or CPU (CUDA disabled) will be running at close to 100% performance. So my issue is irrespective of any bands being applied, it seems that the processing alone is overloading my computer. I listed the specs in the comment above.

7

Hiloe.

You are using too intensive combos maybe. Which are your choices?

8

It is SDM to SDM. If I disable the pipeline, I can run DirectSDM or even FIR2/XFi + ASDM7EC-Super DSD256 → DSD256 without any issues. If I enable pipeline with or without any actual EQ bands loaded in, then I experience extreme resource usage and stuttering. This is an issue with how the pipeline EQ is applied to DSD sources, not with my filter/shaper combo.

9

I can also upsample from DSD256 to DSD512 using FIR2/XFi + ASDM7EC-Super without any issues.

10

Yes, the matrix SDM processor will be somewhat heavier. But what are you using to check your GPU loads? Windows resource monitor shows completely bogus values when CUDA is in use, you need to use for example nvidia-smi to check the actual load values.

DirectSDM of course works at very light load because it does nothing.

1 Like
11

I was just using the resource monitor. I’ll check out nvidia-smi.