HQ Player

[lmitche]

Miska has said there really is not more computer horsepower involved in conversion to non-integer multiples than to integer multiples.

My expectation is that internally HQplayer holds a curve that is higher in resolution than the output sample rate and is therefore largely indifferent to the input sample rate. When rendering that curve it can be expressed at a multiple of either base frequency based on the output rate required.

[lmitche]

Jussi, you may want to consider adding DSD silence to the "X out" button if HQplayer is in Native DSD state.  I triggered the safety circuit on my amp yesterday when I inadvertently "x-ed" out of HQplayer.

[lmitche]

39 minutes ago, Miska said:

 

It does, but when the process exits everything else is up to the driver to decide what it wants to do.

 

What DAC did you have and on what OS?

IFI microIDSD on win 10

[lmitche]

5 hours ago, Miska said:

 

OK, not that I would use it often that way I don't remember encountering such with the windows driver. Assuming you are using ASIO driver... Although it is otherwise noisy DAC, for example having quite a bit of turn-on pop. Maybe it doesn't have output mute circuitry or it is just not operating very well.

 

Yes, running ASIO here. It's noisy between tracks of different resolutions when switching out of native dsd mode. Going into native dsd is quiet.

Using file/exit or x-ing out of Hqplayer before to the mode switch to pcm creates a devastating zero db pop.

[lmitche]

Jussi, can you successfully upsample to dsd512 with polysinc-xtr and adsdm7+512 fs on your nvidia 1060 gpu and 6700k?

 

Thanks,

 

Larry

[lmitche]

3 minutes ago, zorntel said:

Larry's question is for 2-channel.

Yes, thanks professor!

[lmitche]

6 hours ago, ted_b said:

Jussi, are ASIO driver/dac combos that much different that one would allow for more tolerance/performance from HQ than another?  What I mean is:  I own a HQP server built on a Gigabyte Z170X-UD5 mobo, with I7-6700k, GTX 960 and 32GB RAM.  By using a Windows (Caps Carbon w/JCAT USB card) NAA with my Holo Audio ASIO driver I can easily do DSD512 with poly-sinc-xtr, AMSDM7 512+fs yet I read where others who are using different dacs are needing more horsepower to do that, if ever. 

 

Also, does a Windows NAA improve that aspect of performance (i.e would I expect to get same DSD512 hiccup-less performnce if going direct)?  This is not an sq question, just a performance question.

Ted, two of us spent the day trying to duplicate your success with polysinc-xtr and asdm7-512 at dsd512.  We are using i7-6700 and GTX 970 and GTX 1070 gpus.  Dsd upsamples fine but PCM stutters. The last and largest difference between your system and ours is the use of an naa. I think we are out of options and this point and  may revisit when appropriate naa configs are available in our locations.

[lmitche]

57 minutes ago, ted_b said:

+1  I agree that the latest 2s filters mimic the regular ones by about 99.9% at DSD512, and I wouldn't want to blind test them; I'd fail.  If your DAC has DSD512 as its clear sweetspot, then I wouldn't bother spending big dollars to get to non-2s, especially if you can do the xtr-2s family.  My $.02.

Many thanks to Ted, Zorntel, Jussi and everyone else for contributing answers here on the limits of polysinc-xtr and DSD512 hardware.  I'll going to play around for with the new graphics card for another day or two and then likely return it.

 

Ted, your clarification on PCM is helpful.

 

Zorntel, thanks for testing with your son's graphics card.  May he never know you stole it for a few hours!

 

Jussi, thanks for the explanation of polysic-xtr limits and your monster machine.

 

If anyone has a i7-7700k and a Z270 motherboard, it would be useful to know if polysic-xtr pcm upsampling to dsd512 is possible with a GPU.

 

Thanks again. Larry

[lmitche]

57 minutes ago, Hazard said:

Hi Larry, I have O7-7700K and Z270 mobo.  I upsample to DSD512 to my diy DAC (based on DIYinHK usb boaed).

 

With -xrr filter I can upsample DSD64 to DSD512 with no problem.  CPU utilisation is around 25 to 30%.

 

However I get stuttering when I play PCM with this filter.  CPU utilisation is only 50% so I dont understand why it is stuttering.  My video card is an ancient GT9600 so CUDA offload is probably doing nothing.

 

PCM to DSD512 works fine with -xtr-2s filter.  CPU utilisation is around 30%.  I just spent $1200 to build this machine and honestly, everything sound great using -xtr or -xtr-2s filter.  I am not going to spend abymore money chasing the last bit of performance.

 

Oh and as a general comment about Daphile (not in response to your post).  I tried Daphile.  For about 5 min.  Was clearly inferior to HQP in my opinion, in my system.

Hazard, thanks for the information on the 7700k and upsampling.  Since posting my question, another 7700k owner confirmed that his rig will also not upsample pcm to dsd512 with the polysinc-xtr filter.  We have two data points which is useful. Thanks again!

[lmitche]

6 hours ago, Dr Tone said:

 

DOP is related to DAC's supported PCM sample rate.  DSD128 requires at least 352.8kHz pcm input.  DSD256 requires at least 705.6kHz pcm input.  DSD512 in theory would require at least 1411.2kHz pcm input which doesn't exist.

Native dsd, aka DSD without dop, halves the required USB speed, so dsd512 runs at 768kHz.

[lmitche]

21 minutes ago, ted_b said:

So I am revisiting my Roon setup (been using Geoff's JRiver to HQP setup as of now) and I tried to recreate this mixed playlist issue.  However, my dac (Holo Spring) does indeed do DSD so my setup is truly "Auto" in that I upsample all PCM to 352/384 (auto family also on) and DSD to DSD512.  I have no issues with DSD tracks hanging before moving to a PCM track, but my sample size is small so far.

Good to hear this!

[lmitche]

10 hours ago, Kritpoon said:

Thank you fro sharing. I think 16MB of ram wouldn't run even the OS :). Can you please share what OS are you running with? If its Windows OS, do you use optimization program i.e. AO? And when you upsample to DSD512 with -2s what is normally your CPU% utilization? Lastly, with the DSD512 -2s filter, does the NUC fan noise becomes very loud (to the point of annoyance?) Thank you so much for the info

Over here with Windows 10 running Hqplayer and Roon server total physical memory usage is 1.7 gb.  Two 4gb sticks are installed. Unfortunately smaller are unavailable.

[lmitche]

19 minutes ago, rickca said:

iFi has a new USB audio driver 3.2 for the micro iDSD that allows ASIO buffer size up to 32767 samples.  The previous 2.26 driver was limited to 8192 samples.  I'm not sure when this driver became available.  I just discovered it today while reviewing all my current drivers.

Thanks for the heads-up. How does it sound?

[lmitche]

All great software is a work in progress.

[lmitche]

On 9/25/2017 at 5:33 PM, rickca said:

iFi has a new USB audio driver 3.2 for the micro iDSD that allows ASIO buffer size up to 32767 samples.  The previous 2.26 driver was limited to 8192 samples.  I'm not sure when this driver became available.  I just discovered it today while reviewing all my current drivers.

Ok, so running at low latency 16k samples is sounding best here with dsd512.  Anyone else give this a try?

[lmitche]

5 hours ago, tboooe said:

Hi guys...I just completed a new build to replace my old i7-3700s cpu with an i7-6700 because I was having issues converting high res to DSD512.  Everything is working great.  However, I have a question about Turbo mode and cpu utilization.  With the 3700s CPU I would be at around 80% utilization at Turbo mode freq of 3.90ghz.  I noticed with the 6700 cpu, no matter what I do, I cant get it to go into Turbo mode even when using non -2s filter.  The max freq I get is 3.4ghz, even when cpu utilization is above 80%. 

 

Curious if others using this CPU or similar experience the same?  And yes, Turbo mode is enabled in bios.

My 6700k machine runs at 4 ghz all the time. It's locked in the high power profile and speed step is turned off. 

[lmitche]

17 minutes ago, tboooe said:

Thank you...my machine is also set at high performance with speed step turned off.  Not sure whats happening.  I see in cpu-z that my multiplier range is 8-40 so the cpu should hit Turbo.  I just tried running Prime95 on 1 core and even when that core is pegged at 100% utilization, the freq never goes above 3.4ghz. Something must not be set right in my Bios.  Maybe because I have all the C states enabled?  

Are you running AO?

[lmitche]

3 hours ago, tboooe said:

Yes I am.  But I was using it with my previous CPU and I was able to get Turbo Boost, though I am using a new mobo .

If you are running Win10 you may want to run AO once or twice again.  It seems to toggle some values in multiple sequential runs.

[lmitche]

19 minutes ago, ted_b said:

Larry is that a result of AO settings, manual bios or something else?  Where do you see these?  Thx

Hi Ted, I have min and max cpu power set at 100% in the high performance power plan under powercfg.cpl.  I didn't touch the cpu bios controls except to turn spread spectrum off and set the cores to sync at 42 or 43.  I just read the AO manual and confirmed Phil does not touch these settings, although he makes recommendations to turn turbo mode off.  I think that is a mistake as dsd512 is not possible here without turbo.

 

[lmitche]

20 minutes ago, tboooe said:

Do you ever see the CPU reach it's turbo max if 4.2ghz?  It would be interesting if you ran Prime95 on a single core to see if you are getting turbo.

The cores are synced, so there is no difference between core speed. And no, I've never seen turbo max at 4.2.  it's steady at 4.0.

[lmitche]

Over here a Noctua NH-C14S cools the CPU. The fan barely turns and is mechanically silent. An elfidelity fan filter noticably reduced electrical noise from the fan motor. I can run with full dsd512 upsampling with the fan disconnected and temps never get above 70c, but I hear no SQ difference with the fan connected, so I leave it on.

[lmitche]

1 hour ago, JohanJ said:

On my audiolinux pc, directly into Lampiztor Atlantic it isn't possible to convert 48k-based content (like lots of the tidal masters) to 48x128 or 48x256. Hqplayer is running, the playbar is moving, as well as the time, but no sound. When I scale back to the 44x256 I get back the sound. That looks like a DAC flaw?

The same thing happens here with my ifi microIDSD and ASIO driver. Checking auto rate family does the same.

[lmitche]

51 minutes ago, Miska said:

HQPlayer has pretty good understanding of CPU topology...

 

Understatement of the year!

[lmitche]

2 hours ago, RamUwe said:

Approx. 2h listening music I count 9 dropouts. New record …

Was this record set on a dedicated audio PC or were other non audio task running as well?

[lmitche]

3 hours ago, Quadman said:

I am so excited, I just built a new dedicated music PC for someone and it can play the Poly-sinc-XTR filers with no stuttering.  I used a Ryzen 1700x CPU with a ASrock x370gaming ITX/ac motherboard overclocking slightly to 3750 Ghz, ave CPU usage was 41%.  No PC I had built before (all intel 4 core CPU's) could do this.  The 7700K played but was a stuttering mess.  I guess the higher number of cores is key to getting the intensive filters to play.  My apologies if this has been talked about before.  I haven't followed this thread in a while.  Whoo Hoo.

Congrats and thanks for sharing.

 

I wonder if the Intel i7-8700 will do the same.