Best CPU for hqplayer

[AudioDoctor]

1 hour ago, mikicasellas said:

Hi Guys, 

 

I have been using a RYZEN 7 3700X & ASUS ROG STRIX X470-I which got toasted, and would like to go full ATX, i'm putting my eye on these:

 

ASRock B550 PG Velocita Support 3rd Gen AMD AM4 

GIGABYTE AORUS ATX B550 ELITE

ASUS ATX ROG STRIX B550-F

 

Any Suggestions ?

 

Thanks !!

 

I recently built a "CAPS" following Chris's guide here on AS and I find the 10900 intel processor to be far more than just adequate. Coupled with the Nvidia GPU I have zero complaints with regards to running HQPlayer.

[lmitche]

1 hour ago, AudioDoctor said:

 

I recently built a "CAPS" following Chris's guide here on AS and I find the 10900 intel processor to be far more than just adequate. Coupled with the Nvidia GPU I have zero complaints with regards to running HQPlayer.

I'd look at the B450 boards as well. Some have two direct to CPU M.2 slots. I've never seen the same in a B550 board, unless I missed it.

[AudioDoctor]

28 minutes ago, lmitche said:

I'd look at the B450 boards as well. Some have two direct to CPU M.2 slots. I've never seen the same in a B550 board, unless I missed it.

 

Is that a Ryzen board? I am new to all this computer building stuff which is why I followed Chris's guide so closely. And then I went and built my own Ryzen 9 powered computer I am writing this on.

 

[salaryman]

I have both intel and AMD system

intel i7-9700K + Z390

AMD Ryzen pro 4750G + B550

 

general speaking AMD system provide better performance and value, but for Hqplayer, using intel system with high clock is no brainer.

 

And if you look for AMD, they have just announced Zen3 series with 8 core per CCX, which should reduce the latency issue that affect HQPlayer performance.

I would recommend to wait for Zen 3 which is available in early Nov.

 

By the way, B450 chipset will support Zen 3 but may be on selected board only, this rely on whether the board manufacturers will provide the required BIOS updates.

[AudioDoctor]

If you're going AMD AM4 socket, all of the X570 boards will accept Zen 3 processors.

[evalon]

@Miska:

Just a quick question - I noticed that earlier in this thread (I think it was) you mentioned that it likely would take some time before there would be a computer capable of running your new modulators at DSD512 speed. To this end: Do you have any estimate of the clock speed required to run these DSD512 modulators?

 

Cheers,

 

Jesper 

[Miska]

2 hours ago, evalon said:

Just a quick question - I noticed that earlier in this thread (I think it was) you mentioned that it likely would take some time before there would be a computer capable of running your new modulators at DSD512 speed. To this end: Do you have any estimate of the clock speed required to run these DSD512 modulators?

 

Yes, that's the case for EC modulators. Of course regular ones run fine even on low power CPUs.

 

If CPUs clock per instruction efficiency or cross-core communication performance doesn't increase (they do), it would take about 8 GHz CPU for EC modulators. It is hard to estimate when these things will meet the required level.

 

[giordy60]

a question Jussi .....
does it make sense to design a modulator (EC) if to date no pc is able to reproduce it correctly? ....
8 ghz of cpu is obtained only in experimental form and with a cooling system impossible to use continuously .....
in my opinion today HQP has reached a very high level, adding non-usable options penalizes it .... my opinion.

[evalon]

2 hours ago, Miska said:

 

Yes, that's the case for EC modulators. Of course regular ones run fine even on low power CPUs.

 

If CPUs clock per instruction efficiency or cross-core communication performance doesn't increase (they do), it would take about 8 GHz CPU for EC modulators. It is hard to estimate when these things will meet the required level.

 

Thanks Jussi for this feedback ... So likely there is a bit of time before the new DSD512 modulator will find its computer mate ... Good to know. 

 

Cheers - Jesper 

[kohmelo]

1 hour ago, giordy60 said:

a question Jussi .....
does it make sense to design a modulator (EC) if to date no pc is able to reproduce it correctly? ....
8 ghz of cpu is obtained only in experimental form and with a cooling system impossible to use continuously .....
in my opinion today HQP has reached a very high level, adding non-usable options penalizes it .... my opinion.

 

I'm not Jussi, but EC modulators are obviously not "non-usable" as many users prefer EC modulators with DSD256 over anything possible to-date on DSD512 with "original" modulators (myself included).

[Miska]

DSD256 gives you totally flat noise floor up to at least 100 kHz, so it is good for a lot of things. DSD512 extends that to 200 kHz and DSD1024 up to 400 kHz.

 

Rates above DSD256 allow optimizing analog filters and conversion sections, but also make it harder to create low noise switching.

 

With offline conversion one can of course create DSD512 and DSD1024 with EC modulators, but it takes a lot of time and the files get really big. So it is not necessarily very practical for large libraries. And of course doesn't work for something like streaming Tidal or Qobuz.

 

[ted_b]

4 hours ago, giordy60 said:

a question Jussi .....
does it make sense to design a modulator (EC) if to date no pc is able to reproduce it correctly? ....
8 ghz of cpu is obtained only in experimental form and with a cooling system impossible to use continuously .....
in my opinion today HQP has reached a very high level, adding non-usable options penalizes it .... my opinion.

You are only thinking of realtime, on-the-fly processing.  Keep in mind Jussi is developing software for all his HQPlayer variations, one of which is HQPlayer Pro (which our NativeDSD uses, btw).  In that version the EC modulators for DSD512 work wonderfully (offline and slowly).

[AudioDoctor]

Maybe this is a dumb question from someone that hasnt been following all the conversation closely enough, but...

 

Why is something like a threadripper and a powerful nVidia GPU not capable of the more power hungry processing?

[AnotherSpin]

I would like to add to the discussion a bit. The best is not necessarily the most powerful. What kind of computer "ready from the shelf' might be optimal for HQP, if no ultra-processing in the SDM part is required (only PCM) and budget is sensible? 

[jcbenten]

On 10/16/2020 at 11:17 PM, AudioDoctor said:

Maybe this is a dumb question from someone that hasnt been following all the conversation closely enough, but...

 

Why is something like a threadripper and a powerful nVidia GPU not capable of the more power hungry processing?

Software has to be written to take advantage of the cores; it doesn't "just happen".  I am not a software guy but the way I understand it is the algorithms written prior to multiple core systems are linear and not easily broken into chunks to split and reassemble.  

[AudioDoctor]

On 10/16/2020 at 11:49 PM, AnotherSpin said:

I would like to add to the discussion a bit. The best is not necessarily the most powerful. What kind of computer "ready from the shelf' might be optimal for HQP, if no ultra-processing in the SDM part is required (only PCM) and budget is sensible? 

 

Before I built my CAPS, I was doing just that on a 2011 i5 Mac mini dual core with 16GB of ram and an SSD.

[AnotherSpin]

41 minutes ago, AudioDoctor said:

 

Before I built my CAPS, I was doing just that on a 2011 i5 Mac mini dual core with 16GB of ram and an SSD.

 

I also still use the 2011 mac mini. Sooner or later the time will come to replace it with a newer device. I don't understand computers very well, but based on what I have been reading lately, there are doubts that the new mac mini is the best option. Hence the question asked above. 

[AudioDoctor]

1 hour ago, AnotherSpin said:

 

I also still use the 2011 mac mini. Sooner or later the time will come to replace it with a newer device. I don't understand computers very well, but based on what I have been reading lately, there are doubts that the new mac mini is the best option. Hence the question asked above. 

 

Yep, I've been planning on switching away from it for a while now and was doing my research into a new music server when Chris posted the latest CAPS article. It inspired me to build one.

[AnotherSpin]

43 minutes ago, AudioDoctor said:

 

Yep, I've been planning on switching away from it for a while now and was doing my research into a new music server when Chris posted the latest CAPS article. It inspired me to build one.

 

Good for you. Perhaps one day I will dare to do something like that. Is there a reference to the detailed assembly instructions for CAPS?

[AudioDoctor]

3 hours ago, AnotherSpin said:

 

Good for you. Perhaps one day I will dare to do something like that. Is there a reference to the detailed assembly instructions for CAPS?

 

Not on this website. Just the parts he used. However, there are a bunch of good guides on the internet and even video guides on youtube. I had never built a computer before and I found it super easy. I believe anyone here could get it done if they followed the instructions and read the manual...  ha. The documentation that comes with the equipment is pretty much foolproof for anyone of reasonable intelligence which I think most here probably exceed. Its nothing more complicated than sliding Tab A into Slot B and plugging in a few cables, all with clear instructions and they only go in one way.

[Miska]

On 10/19/2020 at 4:34 PM, jcbenten said:

Software has to be written to take advantage of the cores; it doesn't "just happen".  I am not a software guy but the way I understand it is the algorithms written prior to multiple core systems are linear and not easily broken into chunks to split and reassemble.  

 

There are some algorithms that are mathematically or logically bound such way that there's no reasonable way to split them for parallel processing.

 

[ted_b]

Is anyone doing (or thinking of doing) this with their server build?  In order of importance (and not at the same time, of course :) )

*  HQPlayer server for DSD256 2 channel upsampling using EC modulators

*  HQPlayer server for some level of DSD (likely 256) upsampling for multichannel music

*  Kodi serving multichannel audio (and some video) for 5.1.4 file-based movie or 5.1.4 Atmos-music

 

I am open to either using NAA or one-box (NAA seems the right choice here, but I've been on the sidelines for awhile :) ).  My Holo, exaSound and AVR (Atmos and movie only) are up to the task at the far end. 

 

Net/net, wondering if the multichannel and Kodi stuff adds any issues to simply building for DSD256 EC modulator work.

[elan120]

42 minutes ago, ted_b said:

Is anyone doing (or thinking of doing) this with their server build?  In order of importance (and not at the same time, of course :) )

*  HQPlayer server for DSD256 2 channel upsampling using EC modulators

*  HQPlayer server for some level of DSD (likely 256) upsampling for multichannel music

*  Kodi serving multichannel audio (and some video) for 5.1.4 file-based movie or 5.1.4 Atmos-music

 

I am open to either using NAA or one-box (NAA seems the right choice here, but I've been on the sidelines for awhile :) ).  My Holo, exaSound and AVR (Atmos and movie only) are up to the task at the far end. 

 

Net/net, wondering if the multichannel and Kodi stuff adds any issues to simply building for DSD256 EC modulator work.

 

I build mine based on the first option, and my recent comparison between one and two-box configuration, the one-box setup turned out to be a better sounding configuration.

[novaca]

Can anyone guess what the Xeon W-1290T can handle with HQPlayer? It is a 10 core processor with TDP 35 W. The base frequency is low 1.9 GHz, turbo 3.6 / 3.7 GHz. I assume that resampling PCM should not be a problem, but what about DSD... I'm also thinking about not the "T" variant, W-1290 (TDP 80W, base 3.2 GHz, turbo 5.2 GHz), but a good power supply is a problem ...
Thanks

[firedog]

3 hours ago, novaca said:

Can anyone guess what the Xeon W-1290T can handle with HQPlayer? It is a 10 core processor with TDP 35 W. The base frequency is low 1.9 GHz, turbo 3.6 / 3.7 GHz. I assume that resampling PCM should not be a problem, but what about DSD... I'm also thinking about not the "T" variant, W-1290 (TDP 80W, base 3.2 GHz, turbo 5.2 GHz), but a good power supply is a problem ...
Thanks

I had a similar setup. Did most things (not all) with DSD 128.  With some more intensive actions, there could be hiccups with DSD 128. PCM is no issue.