HQ Player

[ronnie.sengupta]

13 minutes ago, Miska said:

 

80% is around the edge of GPU capabilities, usually up to about 75% load it goes fine. But going over the desktop graphics rendering and having something like a web browser or Roon open can cause intermittent processing delays on the GPU causing stutter.

 

Is the 80% with conversion from 44.1k to 22.5792M or something else?

 

 

Conversion from 44.1k to 22.5792 (I have auto-rate family checked).

 

[Miska]

13 minutes ago, ronnie.sengupta said:

Conversion from 44.1k to 22.5792 (I have auto-rate family checked).

 

OK, I can check the load figures for my three GPUs (GTX1080, RTX2080 and RTX2080Ti) when I get back home.

 

[ronnie.sengupta]

18 minutes ago, Miska said:

 

OK, I can check the load figures for my three GPUs (GTX1080, RTX2080 and RTX2080Ti) when I get back home.

 

 

Thank you. I really appreciate it. If I have to upgrade, I would rather upgrade my Processor/Motherboard than the GPU as the my Noctua HD15 with dual fan keeps the CPU very cool and runs pretty quiet as compared to a GPU running at something like a 70% load.

 

But, if upgrading the GPU is a safer bet to get xtr @ dsd512 to work, I don't mind changing the GPU to a 2080 super or a TI

[StreamFidelity]

1 hour ago, Miska said:

OK, I can check the load figures for my three GPUs (GTX1080, RTX2080 and RTX2080Ti) when I get back home.

 

I also have some hardware questions.

 

1. It seems that a CPU in combination with a GPU currently gives the best performance for the HQPlayer, right?

 

2. Is it correct that the DSD Modulators are executed with the CPU and the filters with the GPU if both are available?

 

3. And could alternatively a second CPU on the mainboard (e.g. ASUS WS C621E SAGE) take over the task of the GPU?

[dean70]

HQPlayer desktop 4.2.3 Windows 10 CUDA offload is not working. Offload was enabled on previous version, but playlist loads with offload=disabled.

CUDA offload setting is checked & have CUDA version 10.2 installed.

edit:HQPlayer 3.5.4 reports offload=enabled, but any HQPlayer 4.x reports offload=disabled with the same settings.

[LoryWiv]

12 minutes ago, dean70 said:

HQPlayer desktop 4.2.3 Windows 10 CUDA offload is not working. Offload was enabled on previous version, but playlist loads with offload=disabled.

CUDA offload setting is checked & have CUDA version 10.2 installed.

When I've had this issue previously solution was to upgrade to latest video card driver.

[dean70]

1 hour ago, dean70 said:

HQPlayer desktop 4.2.3 Windows 10 CUDA offload is not working. Offload was enabled on previous version, but playlist loads with offload=disabled.

CUDA offload setting is checked & have CUDA version 10.2 installed.

edit:HQPlayer 3.5.4 reports offload=enabled, but any HQPlayer 4.x reports offload=disabled with the same settings.

After uninstalling and reinstalling HQPlayer 4.2.3 and resetting the options, CUDA offload now working again. Dropouts stopped once the Multicore DSP option was checked.

[Miska]

22 hours ago, StreamFidelity said:

1. It seems that a CPU in combination with a GPU currently gives the best performance for the HQPlayer, right?

 

Sometimes yes, depending a bit on the settings and the particular CPU. If the CPU is not very busy and has no trouble doing everything, there's not so much point in running things on a GPU. For demanding combinations it certainly helps.

 

22 hours ago, StreamFidelity said:

2. Is it correct that the DSD Modulators are executed with the CPU and the filters with the GPU if both are available?

 

Yes...

 

22 hours ago, StreamFidelity said:

3. And could alternatively a second CPU on the mainboard (e.g. ASUS WS C621E SAGE) take over the task of the GPU?

 

Yes, but the optimal work split is different. (this is automatically done)

 

[dctom]

Hi 

a general question about HQPlayer - I have HQP on my i3 itx server going to my NUC naa endpoint. 

I have around 26k tracks in my library, would a more powerful server speed up loading this playlist?

[Miska]

7 hours ago, dctom said:

a general question about HQPlayer - I have HQP on my i3 itx server going to my NUC naa endpoint. 

I have around 26k tracks in my library, would a more powerful server speed up loading this playlist?

 

You have the entire library as a playlist? Or what do you mean by loading the playlist in relation to library and in that relation speed up of what part?

 

[dctom]

I sometimes load a playlist that contains all my music files and have it playing randomly, rather than selecting a particular album from the library. If I load the complete playlist it takes about 20mins or so to load.

[Miska]

9 minutes ago, dctom said:

I sometimes load a playlist that contains all my music files and have it playing randomly, rather than selecting a particular album from the library. If I load the complete playlist it takes about 20mins or so to load.

 

OK, time it takes to load such is usually dominated by the I/O speed. So if the content is on a spinning HDD, then the HDD speed is very likely dominating the speed. If it is USB connected spinning HDD, then it is certainly dominated by the I/O speed. Fastest way to load it is to place the content (library) on an M.2 SSD connected to PCIe 4x (like the motherboard M.2 connectors are). CPU speed also has a factor here, but it is usually much less than the I/O speed.

 

When you load such playlist, HQPlayer goes an loads metadata of all the included tracks, meaning it needs to access every single track on that playlist. If every track also contains embedded cover art, you may need fair amount of RAM to hold it all.

 

[Miska]

On 12/30/2019 at 4:42 AM, Jerry said:

Dear @Miska, Would you consider to offer a "prosumer" version of HQPlayer 4? With limited function of HQPlayer Pro to produce offline DSD512EC file BUT the offline file (is encrypted) could only be played by the encoding HQPlayer (linked to the software purchaser). Since realtime DSD512EC is a long way to go due to CPU constrain, this "prosumer" version could let all of us enjoy DSD512EC much earlier.

 

Problem is that developing such solution would make it cost more than HQPlayer 4 Pro, assuming number of users for such...

 

[Hammer]

Hi, happy new year!

 

My apologies as I'm sure this question has been answered before, but I can't seem to find it!  Every posting I saw deals with installing NAA vs. upgrading an existing NAA installation.

 

i have a pi 4 running raspian and NAA.  How does one update NAA to the latest version 3.6.0?

 

Also, I don't see images for armhf which I think I previously used...is arm64 the same?  I'm looking in

/bins/naa/linux/buster/

Thank you!

 

[Hammer]

Hi, I'm running HQPlayer 4.2.3 on Windows Server 2016 with a 16 core AMD Threadripper 1950X CPU.  I've read in some postings that HQPlayer runs much better on Intel CPU's due to higher clock speeds.  However, I'm wonder why when I run HQPlayer say using the non -2s XTR filters that the clock speed of my system does not go up, but stays at a relatively low 2.2 GHz.  When I run Prime95 to stress the CPU, the clock speed jumps up to 3.6 GHz, so I know the system is working/cooled properly, and more importantly, recognizing heavy loads and increasing clock speed.  Is there any reason why Windows does not "see" or "sense" the demand of HQPlayer and therefore does not increase the CPU clock speed?  The music stutters with the XTR filters so clearly the CPU is not keeping up, but it's also not running as fast as it can...Any insights would be much appreciated.  Thank you!

[Outlaw]

2 hours ago, Hammer said:

Hi, I'm running HQPlayer 4.2.3 on Windows Server 2016 with a 16 core AMD Threadripper 1950X CPU.  I've read in some postings that HQPlayer runs much better on Intel CPU's due to higher clock speeds.  However, I'm wonder why when I run HQPlayer say using the non -2s XTR filters that the clock speed of my system does not go up, but stays at a relatively low 2.2 GHz.  When I run Prime95 to stress the CPU, the clock speed jumps up to 3.6 GHz, so I know the system is working/cooled properly, and more importantly, recognizing heavy loads and increasing clock speed.  Is there any reason why Windows does not "see" or "sense" the demand of HQPlayer and therefore does not increase the CPU clock speed?  The music stutters with the XTR filters so clearly the CPU is not keeping up, but it's also not running as fast as it can...Any insights would be much appreciated.  Thank you!

Under power options what power plan are you using ?

[Hammer]

On 8/7/2019 at 8:11 PM, k6davis said:

Core i7-9700K. It seems to like EC modulators. 😀

 

It will easily upsample PCM 24-192 => EC DSD256 x 48 with poly-sync-ext2. With any OS. Without a GPU. Without high CPU utilization. No overclocking or special settings or special cooling required. 

It's set to Performance

[Hammer]

58 minutes ago, Outlaw said:

Under power options what power plan are you using ?

 

58 minutes ago, Outlaw said:

Under power options what power plan are you using ?

It's set to performance

 

[ronnie.sengupta]

On 12/31/2019 at 5:58 PM, Miska said:

 

OK, I can check the load figures for my three GPUs (GTX1080, RTX2080 and RTX2080Ti) when I get back home.

 

 

I tried Hqplayer pro to generate DSF files upsampling redbook to dsd512 xtr asdm7 512. Strangely, the dsf file when played Direct DSD also stutters. 

 

 

[randytsuch]

Deleted, sorry wrong place.

[Miska]

14 hours ago, ronnie.sengupta said:

I tried Hqplayer pro to generate DSF files upsampling redbook to dsd512 xtr asdm7 512. Strangely, the dsf file when played Direct DSD also stutters. 

 

Are you playing with Direct SDM enabled or disabled?

 

[Miska]

18 hours ago, Hammer said:

When I run Prime95 to stress the CPU, the clock speed jumps up to 3.6 GHz, so I know the system is working/cooled properly, and more importantly, recognizing heavy loads and increasing clock speed.

 

Just curious, if you look on Resource Monitor the per-core CPU load figures, how does it look like with Prime95?

 

18 hours ago, Hammer said:

Is there any reason why Windows does not "see" or "sense" the demand of HQPlayer and therefore does not increase the CPU clock speed?

 

Since the load is not constant full load like in case of Prime95, Windows may sometimes decide that the load is not enough to warrant increasing CPU clocks. Although HQPlayer tells Windows that it is an audio application. Somehow Windows still sometimes decides that it is not time critical, meaning that in Windows opinion it doesn't matter how quickly a processing job is completed.

 

And I'm not sure if Windows Server editions even support audio applications as such.

 

18 hours ago, Hammer said:

Windows Server 2016 with a 16 core AMD Threadripper 1950X CPU

 

This is one possible reason. Windows Server versions are optimized for maximum network data throughput instead of maximum responsiveness. These have opposite requirements for the OS and one reason why to have two different versions in first place.

 

In addition, Windows Server 2016 may be too old to recognize special properties of more recent CPUs. For example for latest Ryzen CPUs, latest feature updates of Windows 10 are needed to get proper support for scheduling on these CPUs. Windows Server editions more likley follow AMD EPYC release cycles.

 

 

If you can have for example spare SSD (doesn't need to be large), you could alternatively try for example Fedora Workstation 31 and HQPlayer there, for comparison.

 

[Hammer]

2 hours ago, Miska said:

 

Just curious, if you look on Resource Monitor the per-core CPU load figures, how does it look like with Prime95?

 

 

Since the load is not constant full load like in case of Prime95, Windows may sometimes decide that the load is not enough to warrant increasing CPU clocks. Although HQPlayer tells Windows that it is an audio application. Somehow Windows still sometimes decides that it is not time critical, meaning that in Windows opinion it doesn't matter how quickly a processing job is completed.

 

And I'm not sure if Windows Server editions even support audio applications as such.

 

 

This is one possible reason. Windows Server versions are optimized for maximum network data throughput instead of maximum responsiveness. These have opposite requirements for the OS and one reason why to have two different versions in first place.

 

In addition, Windows Server 2016 may be too old to recognize special properties of more recent CPUs. For example for latest Ryzen CPUs, latest feature updates of Windows 10 are needed to get proper support for scheduling on these CPUs. Windows Server editions more likley follow AMD EPYC release cycles.

 

 

If you can have for example spare SSD (doesn't need to be large), you could alternatively try for example Fedora Workstation 31 and HQPlayer there, for comparison.

 

Hi, thanks for your help.  When I run Prime95, looking at Performance monitor, all 32 virtual cores are at 100%.

 

Yes, I can install fedora or Win 10 on the same hardware to see if it’s an OS issue or Threadripper issue.

 

Thank you.

[ronnie.sengupta]

4 hours ago, Miska said:

 

Are you playing with Direct SDM enabled or disabled?

 

 

Direct SDM Enabled. 

 

I managed to solve the problem. I had Matrix Pipeline enabled to swap Left-Right channels. Disabling that fixed it.

 

The T+A Firmware I installed for Linux flips channels for DSD.

 

@miska any thoughts on which way to go - CPU or GPU upgrade for 512 xtr.

 

 

 

[ronnie.sengupta]

22 hours ago, Hammer said:

Hi, I'm running HQPlayer 4.2.3 on Windows Server 2016 with a 16 core AMD Threadripper 1950X CPU.  I've read in some postings that HQPlayer runs much better on Intel CPU's due to higher clock speeds.  However, I'm wonder why when I run HQPlayer say using the non -2s XTR filters that the clock speed of my system does not go up, but stays at a relatively low 2.2 GHz.  When I run Prime95 to stress the CPU, the clock speed jumps up to 3.6 GHz, so I know the system is working/cooled properly, and more importantly, recognizing heavy loads and increasing clock speed.  Is there any reason why Windows does not "see" or "sense" the demand of HQPlayer and therefore does not increase the CPU clock speed?  The music stutters with the XTR filters so clearly the CPU is not keeping up, but it's also not running as fast as it can...Any insights would be much appreciated.  Thank you!

 

 

On my Intel i7 7700k without Cuda - running XTR ASDM7 512 from redbook - I noticed that if I turn on Multicore DSP in settings, the CPU hits close to 100% on all 4 cores (I have Hyper threading turned off). If I disable Multicore DSP the CPU utilization comes down to around 58%.

 

With Multicore enabled and Cuda enabled the CPU utilization comes down to 48% with GPU at 80 to 85%

 

The audio stutters in all three cases.