Jump to content
IGNORED

HQPlayer Linux Desktop and HQplayer embedded


ted_b

Recommended Posts

On 6/30/2021 at 10:22 AM, Miska said:

 

OK, HQPlayer OS went back to the generic build that doesn't work nicely on AMD, to be compatible with those smaller embedded CPUs.

 

For big AMD CPUs it is better to install Ubuntu Server, and either Ubuntu low-latency kernel or my custom kernel there and use AMD optimized build on it.

 

After I believe being the one to first report the white noise issue and thankfully it being resolved in 4.24.1 I've just ordered parts to build a Ryzen 7 5800X machine.

 

I've not kept track of the AMD builds nor the Ubuntu custom kernel. The machine will be dedicated to HQPe. I'm fine running Ubuntu 20.04 LTS but where do I find the custom kernel and AMD optimized build?

 

https://www.signalyst.eu/bins/hqplayerd/focal/ Here?

Link to comment
24 minutes ago, lpost said:

I've not kept track of the AMD builds nor the Ubuntu custom kernel. The machine will be dedicated to HQPe. I'm fine running Ubuntu 20.04 LTS but where do I find the custom kernel and AMD optimized build?

 

https://www.signalyst.eu/bins/hqplayerd/focal/ Here?

 

That's the place for both generic/Intel and AMD optimized builds of HQPlayer Embedded.

 

Kernel builds, and needed libgmpris, as they are sort of unsupported/unofficial thing I do for myself and as a hobby, are on my private page:

https://www.sonarnerd.net/src/focal/

And corresponding Debian Buster ones:

https://www.sonarnerd.net/src/buster/

 

But since these are subject to relocation, I recommend going to download pages through my main web pages. These links may become invalid over time.

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
3 hours ago, Miska said:

 

That's the place for both generic/Intel and AMD optimized builds of HQPlayer Embedded.

 

Kernel builds, and needed libgmpris, as they are sort of unsupported/unofficial thing I do for myself and as a hobby, are on my private page:

https://www.sonarnerd.net/src/focal/

And corresponding Debian Buster ones:

https://www.sonarnerd.net/src/buster/

 

But since these are subject to relocation, I recommend going to download pages through my main web pages. These links may become invalid over time.

Thanks, noted.

 

If one was to consider adding CUDA hardware with Ubuntu how many CUDA cores would be 'required' to make it worthwhile, for say a 10%-20% boost or enough to run 7EC at 512 (if this is possible with current hw) in performance?

 

Would an old available GPU with 500 CUDA cores be worth the power to run it or does it require 5000+ cores to be a worthwhile gain in performance?

 

 

Link to comment
3 hours ago, lpost said:

If one was to consider adding CUDA hardware with Ubuntu how many CUDA cores would be 'required' to make it worthwhile, for say a 10%-20% boost or enough to run 7EC at 512 (if this is possible with current hw) in performance?

Jussi has said before CUDA cores are kind of irrelevant, depends on the generation of the card and the specific Nvidia architecture. I wanted to ask the same thing a few days back and read his posts that's why I'm replying to yours, assuming you could get the cards I think the 30 series 3070 / 3080 were good

Link to comment
33 minutes ago, luisma said:

Jussi has said before CUDA cores are kind of irrelevant, depends on the generation of the card and the specific Nvidia architecture. I wanted to ask the same thing a few days back and read his posts that's why I'm replying to yours, assuming you could get the cards I think the 30 series 3070 / 3080 were good

Thanks. I don't believe those will be available for a good while yet. If not CUDA cores what does it use? At the office we have a high Tensor core need but I know this is not what most games use...

Link to comment
3 hours ago, lpost said:

If one was to consider adding CUDA hardware with Ubuntu how many CUDA cores would be 'required' to make it worthwhile, for say a 10%-20% boost or enough to run 7EC at 512 (if this is possible with current hw) in performance?

 

Would an old available GPU with 500 CUDA cores be worth the power to run it or does it require 5000+ cores to be a worthwhile gain in performance?

 

Number of CUDA cores is pretty pointless figure. It is only possibly useful for comparing relative performance of two different models of GPUs of same generation/family.

 

Old GPU will likely just slow down things. Anything before 10-series (Pascal) is practically useless. Already Volta/Turing generations notably increased performance over Pascal. At the moment I would certainly go with minimum RTX 3060. But this also depends on your current CPU. In order for GPU to accelerate something, it needs to be faster on performing given task than the CPU.

 

Sometimes the offload helps because it lowers load on non-modulator cores, allowing higher turbo clocks to be reached on the modulator cores.

 

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
16 minutes ago, lpost said:

Thanks. I don't believe those will be available for a good while yet. If not CUDA cores what does it use? At the office we have a high Tensor core need but I know this is not what most games use...

 

There is much more to the performance than just number of CUDA cores. For example Ampere generation CUDA core is much faster than Pascal generation CUDA core. In addition, things like multitasking on the GPU has been significantly improved over recent generations, and in addition the way memory I/O between CPU and GPU is managed.

 

For example old generations couldn't hold multitasking data on the GPU, but instead at every time a workload was launched on the GPU the entire thing had to be loaded there, executed and then unloaded, and no parallel tasks. This had a huge overhead, of which even most recent AMD GPUs suffer from, even though they have very good performance on paper.

 

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
17 hours ago, Miska said:

 

Latest release relocates home from /var/hqplayer to /var/lib/hqplayer, in order for this to complete properly, hqplayerd must be stopped before installing the package (see note in the release announcement on my web page).

 

thank miska.

i'm remove old file, scan my libary. i hope hqpe will be make my libary hapy.

Link to comment

I have an odd issue today.

 

I came to use the desktop version in Linux Mint and I can open the software (4.12.2-47 I think) but then I cannot actually click on anything, it doesnt respond. I have tried killing the app and also reinstalling. Everything else seems to be working on the PC. Any ideas?

Link to comment
On 7/8/2021 at 4:30 PM, rossco said:

I came to use the desktop version in Linux Mint and I can open the software (4.12.2-47 I think) but then I cannot actually click on anything, it doesnt respond. I have tried killing the app and also reinstalling. Everything else seems to be working on the PC. Any ideas?

 

Network issue accessing NAA or the previously used content? Or some other problem accessing the DAC?

 

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
13 hours ago, perpetualapprentice said:

CUDA question. Instructions on setting up a key to securely download CUDA are on the CUDA toolkit download site. (No key = error message when trying to add as a repository to 'apt update').

 

Are specific commands available to only download the runtime parts required for HQPlayer? Or perhaps HQP requirements with a minimum extra?

 

You only need the "nvidia" kernel device driver, nothing else.

 

Nvidia has meta package called "cuda-drivers", but I think it pulls in a lot of extra. But is still better option than pulling in entire thing. I think smallest thing you can install is "nvidia-dkms-470", but since they don't have a meta package, the downside is that you need to manually track for updates (from 470 branch onwards).

 

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
1 hour ago, Miska said:

 

You only need the "nvidia" kernel device driver, nothing else.

 

Nvidia has meta package called "cuda-drivers", but I think it pulls in a lot of extra. But is still better option than pulling in entire thing. I think smallest thing you can install is "nvidia-dkms-470", but since they don't have a meta package, the downside is that you need to manually track for updates (from 470 branch onwards).

 

The device is used exclusively for HQPlayer purpose.

 

If running 20.04 Focal LTS server headless, would either the “nvidia-headless-470” or “nvidia-headless-no-dkms-470” be valid options to pull in less? Nvidia instructions say to download the NVIDIA CUDA toolkit as a “pre-installation action” but this I will skip.

Link to comment
2 minutes ago, perpetualapprentice said:

The device is used exclusively for HQPlayer purpose.

 

If running 20.04 Focal LTS server headless, would either the “nvidia-headless-470” or “nvidia-headless-no-dkms-470” be valid options to pull in less? Nvidia instructions say to download the NVIDIA CUDA toolkit as a “pre-installation action” but this I will skip.

 

You can also install nvidia-headless-470. But you need the dkms (drivers) at minimum.

 

nvidia-headless-470 pulls in:

nvidia-headless-no-dkms-470, nvidia-dkms-470

 

nvidia-headless-no-dkms-470 pulls in:

nvidia-kernel-source-470, libnvidia-compute-470, nvidia-compute-utils-470, libnvidia-cfg1-470

 

Of which the first one is useful with custom kernels while the other ones shouldn't be strictly needed but shouldn't hurt.

 

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment
On 7/8/2021 at 8:30 AM, rossco said:

I have an odd issue today.

 

I came to use the desktop version in Linux Mint and I can open the software (4.12.2-47 I think) but then I cannot actually click on anything, it doesnt respond. I have tried killing the app and also reinstalling. Everything else seems to be working on the PC. Any ideas?

Just a FWIW, different setup but similar behavior

Only a early version of 4x works, beyond that it will not.  This would seemingly indicate something may have changed in the application not meshing in some environments like it used to

My rig

 

Link to comment
On 7/9/2021 at 9:18 AM, Miska said:

 

You can also install nvidia-headless-470. But you need the dkms (drivers) at minimum.

 

nvidia-headless-470 pulls in:

nvidia-headless-no-dkms-470, nvidia-dkms-470

 

nvidia-headless-no-dkms-470 pulls in:

nvidia-kernel-source-470, libnvidia-compute-470, nvidia-compute-utils-470, libnvidia-cfg1-470

 

Of which the first one is useful with custom kernels while the other ones shouldn't be strictly needed but shouldn't hurt.

 

Thank you, Miska. As a Linux newbie I'd been seeking out step-by-step instruction online but this method was so much easier than all others I've used. Embedded version now on day 4 of continuously play as I put hours on a new DAC. Very large looping shuffle of DSD & PCM material with a wide variety of sample rates working perfectly.

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...