Best CPU for hqplayer

[bibo01]

37 minutes ago, Ralf11 said:

will it run on a GPU?

NAMD makes use of GPU. It's not a video card benchmark as such. Other tests can do that.

[Miska]

6 hours ago, bibo01 said:

I looked at NAMD site extensively, but I was not able to find a list of benchmark results, or at least what I found was referring to the number of nodes in parallel computing, not to a list of CPUs which is what we want.

 

Problem is that people would need to buy CPUs to figure out how they perform. While, IMO, idea with benchmarks is that you would know before buying by just looking at published listings... So I'd look for some well established benchmark that gets as close as possible to same CPU order.

 

Looking at these, I guess both POV-Ray and CFD tests are good and maybe HQPlayer is somewhere between the two:

https://hothardware.com/reviews/amd-2nd-gen-ryzen-threadripper-2920x-and-2970wx-review?page=4

https://hothardware.com/reviews/intel-core-i9-9900k-9th-gen-cpu-review?page=5

 

CFD puts somewhat similar stress on RAM and cache, although maybe a bit more. While POV-Ray is otherwise computationally heavy.

 

[Ralf11]

someone familiar with the software design could make a good estimate as to whether raw clock speed, # cores, use of a GPU, etc. mattered the most...

[Miska]

3 hours ago, Ralf11 said:

someone familiar with the software design could make a good estimate as to whether raw clock speed, # cores, use of a GPU, etc. mattered the most...

 

It is just that there is no any single defining variable, but combination of. And for example CPU cache performance is sort of black magic which you don't know until you run a certain usage pattern on it... And different CPU architectures perform differently with same code, even if they have same clock speed, number of cores and amount of cache.

 

[Miska]

Some news related to the topic:

https://venturebeat.com/2018/11/06/amd-reveals-zen-2-processor-architecture-in-bid-to-stay-ahead-of-intel/

 

"It also has better 256-bit load/store floating point processing, or double the previous generation."

[Jking]

Audirvana (Win) can use 1 core per channel to upsample.  If they are offloading to Cuda I am wondering if HQPlayer is a single threaded app.  

[Miska]

2 minutes ago, Jking said:

Audirvana (Win) can use 1 core per channel to upsample.  If they are offloading to Cuda I am wondering if AudioHQ is a single threaded app.  

 

What is AudioHQ? And how is this relevant?

 

[Jking]

Edited....

[Miska]

HQPlayer can use any number of CPU/CUDA cores per channel.

 

If you look for example this screen shot, you can see number of threads (77) in use from the resource monitor in top right corner:

 

Or this from Mac (54 threads):

 

(and no, they are not all doing DSP work, most of them are actually sleeping)

[Jking]

Thanks for posting that.  It is definitely not single threaded.  It looks like 4 cores.  With a I3-3225 I am at 47% usage upsampling to DSD 256.  Would be interested in your HQPlayer settings if you are still using Botic. 

[Ralf11]

5 hours ago, Miska said:

 

It is just that there is no any single defining variable, but combination of. And for example CPU cache performance is sort of black magic which you don't know until you run a certain usage pattern on it... And different CPU architectures perform differently with same code, even if they have same clock speed, number of cores and amount of cache.

 

 

sure, but the question is the shape of the response surface in multi-factorial space - I understand it is non-trivial

[HumanMedia]

The new Mac Minis were mentioned a few pages back, has anyone tried them out with HQPlayer?  What are their processing limitations with HQPlayer?

 

I was going to buy one as a media server, but if it could also be used for HQPlayer as well maybe I should go for the higher end model.

 

Anyone have experience with them yet?

[Jens_G]

Today I have completed the built of my new Audio PC. I’m very happy so far and I can confirm that the Xeon E5-2689v4 can upsample 44.1kHz to DSD512 using poly-sinc-xtr (non -2s) without hickups. CPU load is at approx. 44%.

[asdf1000]

8 minutes ago, Jens_G said:

Today I have completed the built of my new Audio PC. I’m very happy so far and I can confirm that the Xeon E5-2689v4 can upsample 44.1kHz to DSD512 using poly-sinc-xtr (non -2s) without hickups. CPU load is at approx. 44%.

 

Thanks for sharing. Can it also do 44.1 to DSD512x48 ?

 

If your DAC supports DSD512x48, what is that CPU load approx.?

[Jens_G]

Just testing 44.1kHz to DSD512x48. Works fine too. CPU load is at around the same level as before, e.g. 43-44%. Bit I’m not sure whether my DAC can really do it? It‘s a iDSD micro BL... Have to look into the spec sheet.

[asdf1000]

1 minute ago, Jens_G said:

Just testing 44.1kHz to DSD512x48. Works fine too. CPU load is at around the same level as before, e.g. 43-44%.

 

Thanks again. Which DAC, if you don't mind me asking?

 

[Jens_G]

It‘s an iDSD micro Black Label.

[Miska]

14 hours ago, Jens_G said:

Today I have completed the built of my new Audio PC. I’m very happy so far and I can confirm that the Xeon E5-2689v4 can upsample 44.1kHz to DSD512 using poly-sinc-xtr (non -2s) without hickups. CPU load is at approx. 44%.

 

Do you have a single or dual socket system? So do you have two of those CPUs? Just asking because it supports dual-socket...

 

[Jens_G]

It‘s a single socket system. The motherboard is an ASUS x99 Deluxe ll with 32 GB RAM DDR4 2400MHz.

[PieterP]

Is anyone planning to buy and test ryzen 3700x with HQPlayer in a short term? 

[teera_ash]

I plan to build Ryzen 9 3700x probably next week using Streacom fanless case DB4.  It will run on Audiolinux with Roon + HQplayer.  Will let you know whether it can do dsd512 with xtr variations.

[salaryman]

On 10/11/2019 at 10:46 PM, teera_ash said:

I plan to build Ryzen 9 3700x probably next week using Streacom fanless case DB4.  It will run on Audiolinux with Roon + HQplayer.  Will let you know whether it can do dsd512 with xtr variations.

 

I have made this test setup (no case, naked testbed):

Ryzen 3700x + Stock HSF

Giga X570 ITX

Corsair 8G x2 3200C16

Old Samsung XP941 128G M.2

Seasonic SFX 650W

old fanless display card

 

WinSvr 2016 + AO + HqPlayer Desktop 3.5.4

iFi micro DSD BL + HD800

 

Result is very positive:

ploy-sinc-xtr-mp - ASDM7 - 44.1k x512 - SDM(DSD)

No glitch, very smooth playback,

After 10 min, HSF is just a bit warm only, it should be possible to do it in fanless case

Have to sell my 1950X system ASAP.

 

 

[teera_ash]

Very encouraging to hear that.  I'll have my Asrock B450 motherboard and Streacom DB4 case in a day or two.  Cannot wait to see how Ryzen 7 3700x upsampling everything to dsd512-xtr with absolute silence fanless case in tropical weather (I live in Bangkok).  

[rickca]

7 hours ago, salaryman said:

ploy-sinc-xtr-mp - ASDM7 - 44.1k x512 - SDM(DSD)

No glitch, very smooth playback,

Have you tried the EC modulators at DSD256?

[salaryman]

3 hours ago, rickca said:

Have you tried the EC modulators at DSD256?

 

Oh, I just search and realize what EC stands for, coz I am still with HQP 3.x.

May try later.