Best CPU for hqplayer

[Miska]

1 minute ago, kelvinwsy said:

I remember now it was stuttering a bit at 4.1 ghz 1st 4 cores b4. With AV512 ON 3.9 ghz runs .. does that it shows executionof HQplayer instruction sets is now more 

efficient?

 

Yes, similar to my Xeon W-2245, at it's base clock it can do same Core CPUs can do at higher clocks. But it would be even better if it could do the same at higher clocks, meaning it could do more work at same clock speed.

 

[kelvinwsy]

Ok different aim.. Tweaking for me is how I try each tweak for SQ for every little audible improvemt. 😂 I dont even use DSP filters for my HPs anymore! I did b4 as my HD800s needed Bass boost! My Abyss Ab1266 phi/cc just needs Amplifier juice a lot! . so I hv loads of Computing power I dont even need😂

I will chk if loading Optimized Settings changes my AO3 settings .. I remember it does.. will cost me to biy more AO3 keys so maybe I will pass....

 

[Miska]

Yeah, but the entire point of HQPlayer is to be DSP engine for improving DAC performance and as result improving reconstruction accuracy of the output analog signal. HQPlayer is just big very optimized DSP engine wrapped into a software package. Sound quality is result of the DSP algorithms. All objectively verified and quality controlled.

 

More DSP capacity you can build, more accurate results HQPlayer DSP can compute.

 

[kelvinwsy]

Yes I can try but Normal Turbo boost the mobo and 11900k out of the box goes to 5.0ghz.. Xmp goes to 3200 profile 1

load optimal defaults puts more load on the cpu and higher voltages .. and Heat..

Does this make DSD 256/EC7 xtr-lp sound BETTER? If so How and Why?

[kelvinwsy]

I think others hv found lowering to minimum cpu clock speed and lowest voltages give quieter sound.. ??

[Miska]

2 hours ago, kelvinwsy said:

Yes I can try but Normal Turbo boost the mobo and 11900k out of the box goes to 5.0ghz.. Xmp goes to 3200 profile 1

load optimal defaults puts more load on the cpu and higher voltages .. and Heat..

Does this make DSD 256/EC7 xtr-lp sound BETTER? If so How and Why?

 

i9-11900K has base clock of 3.5 GHz and maximum turbo boost of 5.3 GHz:

https://ark.intel.com/content/www/us/en/ark/products/212325/intel-core-i911900k-processor-16m-cache-up-to-5-30-ghz.html

 

OS will clock the CPU then as high as necessary for a given load. It makes things sound better by allowing you to run more DSP things. IOW, get better data out to the DAC.

 

2 hours ago, kelvinwsy said:

I think others hv found lowering to minimum cpu clock speed and lowest voltages give quieter sound.. ??

 

It doesn't affect sound as such, computation results are the same regardless of clock speed and voltages and thus it sounds the same. If you are worried about noise, use a NAA which moves the actual playback out of HQPlayer computer and leaves HQPlayer only as DSP data processor.

 

[pis99]

43 minutes ago, Miska said:

 

i9-11900K has base clock of 3.5 GHz and maximum turbo boost of 5.3 GHz:

https://ark.intel.com/content/www/us/en/ark/products/212325/intel-core-i911900k-processor-16m-cache-up-to-5-30-ghz.html

 

OS will clock the CPU then as high as necessary for a given load. It makes things sound better by allowing you to run more DSP things. IOW, get better data out to the DAC.

 

 

It doesn't affect sound as such, computation results are the same regardless of clock speed and voltages and thus it sounds the same. If you are worried about noise, use a NAA which moves the actual playback out of HQPlayer computer and leaves HQPlayer only as DSP data processor.

 

Yes, I can hear the difference between my Mac mini M1 and my son's gaming computer 11900K/RTX3090. DAC is a new Holo 3 L3 just arrived from China. I already felt M1 has a decent performance with Hqplayer, but after purchasing another license of Window desktop HQ. The difference is quite obvious with powerful computer. And there is no coming back, at least for me. Next step I want to try Linux version of Hqplyer. Many said the performance will be even better. If this case, performance equals to better sound.

[jamesg11]

/& we wait to see what the mooted M1x will bring ...

[kelvinwsy]

For whatbit is worth I run the 11900k and activating AV512 allows for a lower overclocking freq and DSD256/EC7 will still run smoothly wo a GPU.. I go it down to 3.9 ghz all 7 cores with polysinc xtr lp .. 

Temps of CPU REALLY really low

This was after 1 hour of DSD 256/EC7/polysinc xtr lp.. 

I dont think I am imagining it but the music sounds calmer esp in the HF

[kelvinwsy]

Ok @Miska I hv done the tests as U hv suggested:

1. Load Optimal Default Settings

2. Load XMP Profile 1 / 3200 mhz in my setup

3. The Turbo was autonset at 4800 mhz CPU

4. Manually disabled the Disabled for AVX512

Hqplayer 4 plays ALL Filters except Sinc L WITHOUT any stutter.. Sinc L plays but a stutter every 7-8 secs.

5. A short 10 min test running thru all the tough filters sees the Average CPU temps at 43-45deg Celsius

I hv 4x8 = 32 gb Dram4 in all memory slots

 

This is a v good result ??

[Miska]

3 hours ago, kelvinwsy said:

Hqplayer 4 plays ALL Filters except Sinc L WITHOUT any stutter.. Sinc L plays but a stutter every 7-8 secs.

5. A short 10 min test running thru all the tough filters sees the Average CPU temps at 43-45deg Celsius

I hv 4x8 = 32 gb Dram4 in all memory slots

 

This is a v good result ??

 

Yes, pretty much expected result, performs as it should. For sinc-L you'd likely need a Nvidia GPU.

 

Not worth buying GPU for that unless you'd for some reason specifically want that filter.

 

[kelvinwsy]

Thx Jussi for your helpful guidance

 

I just like to share MY personal preferences..

1. XMP does introduce a bit of treble "fizz" when a female soprano hits and holds a high note.. That is just what my ears hear

2. I like the 11900k vs my other 10700k PC where at 3.9/4 ghz I can play up to Closed Form filter very smoothly.. Everything else does not play

3. I discover that higher filters say gauss-xla has better resolution than eg..  polysinc xtr lp and I hear a bit more of softer/ ambient sounds.. Sub Bass(< 20hz ) rumbles away nicely( my Abyss 1266 goes down to 7 hz) more clear than the xtr-lp

 

Over the next few weeks, I will compare the filters more intently since I could not play these with the 10700k and NOW with the 11900k I can!!

 

You are right .. more PC computing power, more DSP sound improvemt that you can hear

Thx 

 

[kelvinwsy]

Hi Jussi -- what is your developmt pkan on HQplayer 4 Desktop using the 2 types of Cores of the 12900k?

[Miska]

3 minutes ago, kelvinwsy said:

1. XMP does introduce a bit of treble "fizz" when a female soprano hits and holds a high note.. That is just what my ears hear

 

It has nothing to do with sound. If you have such effects, something is wrong electrically in the system. You could try using a NAA to isolate processing and playback side. This way processing and playback are isolated from each other. This also allows you to move your HQPlayer server out of the listening room to basement or somewhere else.

 

All the DSP calculation results are the same, just memory speed is higher and thus calculations finish faster.

 

[jacky5555]

On 9/27/2021 at 4:31 PM, kelvinwsy said:

For whatbit is worth I run the 11900k and activating AV512 allows for a lower overclocking freq and DSD256/EC7 will still run smoothly wo a GPU.. I go it down to 3.9 ghz all 7 cores with polysinc xtr lp .. 

Temps of CPU REALLY really low

This was after 1 hour of DSD 256/EC7/polysinc xtr lp.. 

I dont think I am imagining it but the music sounds calmer esp in the HF

I’m interested in z590 + 11900k for my new server. It seems you use active cooler (fan) to keep cpu temperature that low?

[kelvinwsy]

CoolerMaster ML360R v1..

6 fans in 3x Push (Noctua NF-A12) snd 3 Orig fans in Pull ..

Only the Aio pump is on the mobo fan header..See photo below

 

1. 2x 14 cm exhaust at 50% speed outside the upper side loose flexible mounted with copper mesh shielding it from the mobo below

2. 3x 12 cm intake at 50% flexibly mounted under AND OUTSIDE my my LIAN LI O11Dynamic case lower side

3, the 3 noctuas at 50% on Aio radiator inside case in a shielded and earthed Mumetal compartmt ( my DIY effort) inside the Lian Li O11D case

 

All powered by a 2nd ATX PSU powering a fan controller 

 

All the fans are shielded away from the mobo by external mounting or a shielded compartment!!

The PC is so quiet I sit 6 feet away on my Abyss Hp.. between tracks (no music) it is silence

 

 I just cranked the overclocking bk fown to 4.6 1st x 4 cores and 4.4 the next 4 cores.. 36deg C after 2 hr with DSD 256/ASDM7EC with pokydinc Long and Av512 Enabled and memory XMP Disabled..

CRAZY DIY PC setup but works for me

 

Very very cool CPU🤣Temps and v low fan EMI/RFI and vibration impact on SQ

[kelvinwsy]

Took me Almost 1 year to figure how to bring my fan / AIO setup to here.. SQ really does improve a lot all ad constant

[kelvinwsy]

Oops wrong photo abv.. this us 4.8 ghz all 8 cores XMP Profile 1 - 3200ghz.. AV512 enabled playing Sinc Mx at DSD256/EC7

[kelvinwsy]

Correct photo for 4.6 4 cores/4.4 other 4 cores playing Gauss Long XMP disabled AV512 enabled 12 min playtime.. Also 

AC on Econ Quiet mode is quite a bit louder than my PC 

[emailtim]

On 10/31/2019 at 9:37 AM, Miska said:

Most of the performance hit is related to SMT (HyperThreading) which is not really used much by HQPlayer, except on dual-core machines.

 

BIOS update is related to CPU microcode update. Assuming older CPU models (BIOS) are also updated to use the latest microcode security updates, the performance implications should be similar on both K and KS.

 

I have gone through several CPU microcode updates on my servers, but I have not noticed any performance changes for HQPlayer due to such. I have noticed that latest Linux kernels are again faster, due to improvements and optimizations on security fixes related to the CPU flaws. Many of the flaws are also as much related to AMD CPUs. Actually fixing those flaws in CPU hardware itself instead of microcode and kernel code, should make the performance hit much smaller than it would be with all the firmware and software workarounds. However, this requires also that the operating system detects that the CPU has fixes in place and doesn't therefore anymore apply also those software workarounds. I would assume this to happen at least for Linux kernel 5.4 release and likely that gets also propagated to 4.9/4.14/4.19 LTS kernels. Not sure if for example Windows has already been updated to recognize that newest CPU models have got fixes for some of the hardware flaws and thus don't anymore need costly workarounds.

 

 

Hi Miska,

 

Do you have any feedback on the lastest CPU Vendor microcode updates (from 2021) for Linux distros with respect to performance ?

e.g. Debian Package: intel-microcode (3.20210608.2) [non-free] - I have looked through the docs on this one and it appears to include mostly threat mitigations.

 

Also, FWIW, the newer Linux Kernel 5.14 supposedly has lower latency USB music driver enhancements

 

https://www.phoronix.com/scan.php?page=news_item&px=Linux-5.14-USB-Audio-Latency

 

and Linux Kernel 5.15 incorporates complete NTFS support from a previously proprietary paid package from Paragon Software.

 

https://www.techrepublic.com/article/ntfs-support-gets-a-significant-boost-in-linux-kernel-5-15/

 

Thanks much.

 

 

[Miska]

13 hours ago, emailtim said:

Do you have any feedback on the lastest CPU Vendor microcode updates (from 2021) for Linux distros with respect to performance ?

e.g. Debian Package: intel-microcode (3.20210608.2) [non-free] - I have looked through the docs on this one and it appears to include mostly threat mitigations.

 

Some of those threat mitigations also impact performance (technically) to some extent, but I have not seen any noticeable performance difference in HQPlayer algorithms.

 

BIOS updates are best way to keep CPU microcode up to date. This applies also to AMD CPUs.

 

13 hours ago, emailtim said:

Also, FWIW, the newer Linux Kernel 5.14 supposedly has lower latency USB music driver enhancements

 

https://www.phoronix.com/scan.php?page=news_item&px=Linux-5.14-USB-Audio-Latency

 

I got one report that the feature breaks HQPlayer. I will look into it at some point.

 

13 hours ago, emailtim said:

and Linux Kernel 5.15 incorporates complete NTFS support from a previously proprietary paid package from Paragon Software.

 

https://www.techrepublic.com/article/ntfs-support-gets-a-significant-boost-in-linux-kernel-5-15/

 

I have already added read-only NTFS and bunch of other filesystems to next HQPlayer OS release. It will take some time before 5.15+ finds it's way to HQPlayer OS.

 

I'm relying on Longterm kernel releases:

https://www.kernel.org/category/releases.html

 

 

[emailtim]

On 10/10/2021 at 3:58 AM, Miska said:

 

Some of those threat mitigations also impact performance (technically) to some extent, but I have not seen any noticeable performance difference in HQPlayer algorithms.

 

BIOS updates are best way to keep CPU microcode up to date. This applies also to AMD CPUs.

...

 

 

 

Miska,

 

Thanks for the information.

 

FWIW, I tried the Intel-Microcode patch on my older system yesterday.  The last beta BIOS update for my Gigabyte Mobo was in 2017 and it appears the 2021 Intel-Microcode patch had some Intel updates from 2019 for my specific CPU.  I went with the beta Gigabyte BIOS because it had undocumented support for PCIe/NVMe boot drives specific to my old IvyBridge MoBo model.  I was able get a 4X NVMe speed boost over SATA/SSDs until I get a newer system that officially supports NVMe.  The microcode patch doesn't appear to have broken anything so far and it looks like it picked up 2 more years of support from Intel.

 

 

[Miska]

On 10/10/2021 at 12:05 AM, emailtim said:

Also, FWIW, the newer Linux Kernel 5.14 supposedly has lower latency USB music driver enhancements

 

I have tested this again on Fedora 34 which is currently using kernel "5.14.9-200.fc34.x86_64" and seems to work fine with HQPlayer Embedded. I don't notice any functional difference to earlier versions. For DAC manufacturers it may make some difference though.

 

I would also expect 5.15 kernel to land on Fedora pretty quickly, so if you want to stay on bleeding edge with OS officially supported for HQPlayer Embedded (and semi-officially for HQPlayer Desktop), Fedora is good baseline.

 

[Miska]

People with AMD CPUs may want to hold upgrades to Windows 11 some time further:

https://tech.slashdot.org/story/21/10/13/1550231/windows-11s-first-update-makes-amd-cpu-performance-even-worse

 

[Khordo]

On 10/16/2021 at 10:18 AM, Miska said:

People with AMD CPUs may want to hold upgrades to Windows 11 some time further:

https://tech.slashdot.org/story/21/10/13/1550231/windows-11s-first-update-makes-amd-cpu-performance-even-worse

 

No issue for me for the moment with a 3900x. Just weird GPU behavior with CUDA fixed when i unchecked a flag regarding the latency in the 'Default graphic card' options.

Can't see / hear differences in w11 vs w10.(Note that i did a clean install, not an upgrade)