[Yeang]

Wow that is really amazing @Energy . I am in awe that there was NO need for throttlestop, C-states or energy settings. That is mindblowing. I am very excited as I am rooting for AMD (I own lots of Intel processors and even stock in company but I feel they are just abusing monopoly).

Question: Did you try upsampling to DSD512 across rate families? I.e. upsampling 24/96 file to 22Mhz instead of 24Mhz? That is super super processor intensive, Jussi's old 6950X could NOT do it, but could do it with auto rate family. Just curious, no practical reason.

I sent a message to Jussi to look at this thread as you are the first person I think who has used Threadripper for this purpose, and I also think first to test 7900X. I hope you do not mind.

Yeang

ISO REGEN ISSUE
- ISO REGEN stopped working after Windows update. I uninstalled all the audio drivers and updated to the Windows Fall Creators Build 1709, then I reinstalled the Holo Audio Spring 3.40 driver and the ISO REGEN still did not work, Singxer SU-1 would not be recognized by the computer. Because of that I thought the ISO REGEN was broken. For some strange reason today I decided to it throw back into the equipment chain and it detected the Singxer SU-1 and plays nicely now without pop noises or detection issues so I'm glad I don't need to send it back.

RAM ISSUE
- The G.Skill Trident Z 32GB Quad Channel I bought did not run stable with it's rated X.M.P profile for 3000MHz. Right now I have it at 2133MHz but it's still good enough for xtr filters.

PC Settings


- No ThrottleStop, Disable C-States, or Energy Settings is needed.
All you need is the CPU at stock speeds with at least dual channel ram at 2133MHz. USB Streaming mode default "Reliable" is okay. No need for Extra Safe.

- ISO REGEN problem did NOT affect why HQPlayer still stuttered on the Intel 7900X. The CPU simply does not have enough L3 cache. HQPlayer is a program that utilizes more L3 cache which is why the Intel 6950X worked and not the 7900X. I think the cheaper AMD 1920X should be good enough for the job with the extra 2 cores and more L3 cache.

MODULATOR
AMSDM7 512+fs sounds more weighty to me than the ASDM7 so I prefer it. ASDM7 was a little more thin sounding. It makes sense since AMSDM7 512+fs takes more CPU power so I assume it's the better performing modulator. You can experiment with it and let me know your results.

 

[elan120]

ISO REGEN ISSUE
- ISO REGEN stopped working after Windows update. I uninstalled all the audio drivers and updated to the Windows Fall Creators Build 1709, then I reinstalled the Holo Audio Spring 3.40 driver and the ISO REGEN still did not work, Singxer SU-1 would not be recognized by the computer. Because of that I thought the ISO REGEN was broken. For some strange reason today I decided to it throw back into the equipment chain and it detected the Singxer SU-1 and plays nicely now without pop noises or detection issues so I'm glad I don't need to send it back.

RAM ISSUE
- The G.Skill Trident Z 32GB Quad Channel I bought did not run stable with it's rated X.M.P profile for 3000MHz. Right now I have it at 2133MHz but it's still good enough for xtr filters.

PC Settings


- No ThrottleStop, Disable C-States, or Energy Settings is needed.
All you need is the CPU at stock speeds with at least dual channel ram at 2133MHz. USB Streaming mode default "Reliable" is okay. No need for Extra Safe.

- ISO REGEN problem did NOT affect why HQPlayer still stuttered on the Intel 7900X. The CPU simply does not have enough L3 cache. HQPlayer is a program that utilizes more L3 cache which is why the Intel 6950X worked and not the 7900X. I think the cheaper AMD 1920X should be good enough for the job with the extra 2 cores and more L3 cache.

MODULATOR
AMSDM7 512+fs sounds more weighty to me than the ASDM7 so I prefer it. ASDM7 was a little more thin sounding. It makes sense since AMSDM7 512+fs takes more CPU power so I assume it's the better performing modulator. You can experiment with it and let me know your results.

Hi @Energy ,

Thank you for sharing your findings as I am also looking into my next server build presently, and was wondering if you could share which Mobo you are using for your server?

 

[Energy]

Wow that is really amazing @Energy . I am in awe that there was NO need for throttlestop, C-states or energy settings. That is mindblowing. I am very excited as I am rooting for AMD (I own lots of Intel processors and even stock in company but I feel they are just abusing monopoly).

Question: Did you try upsampling to DSD512 across rate families? I.e. upsampling 24/96 file to 22Mhz instead of 24Mhz? That is super super processor intensive, Jussi's old 6950X could NOT do it, but could do it with auto rate family. Just curious, no practical reason.

I sent a message to Jussi to look at this thread as you are the first person I think who has used Threadripper for this purpose, and I also think first to test 7900X. I hope you do not mind.

Yeang

It's amazing. I can play music and do whatever else, even Photoshop or video edit and the music doesn't stutter at all! No problem at all. Feel free to link anyone.

I tested 24/96 and 24/192 and both play at 22MHz smoothly. Picture provided.

Hi @Energy ,

Thank you for sharing your findings as I am also looking into my next server build presently, and was wondering if you could share which Mobo you are using for your server?

I built a ultra quiet gaming computer to handle DSD512.
Fan's are only 15dB. With the side panel on, 10dB, so fan noise is not an issue.

The motherboard used was an ASUS STRIX X399-E.







A server board, board graphics, with standards memory would be much cheaper.

 

[gr8soundz]

Very nice build @Energy . Glad to hear your Threadripper investment paid off.

It'll be a while before I can build another machine and I'm a fanless kind of guy (built a dozen or so fanless HTPCs and other machines for music over the years). I doubt the insane core counts and 180W tdp of Threadripper could be passively cooled but the upcoming Calyos NSG-S0 (http://fanlessfan.com/) may be able to handle it (assuming my imaginary future wallet could handle the parts list).

Anyway, I've been wondering how DSD512 on your Holo Spring compares to Chord's DACs (assuming you've heard them)? Technically, I know they're two different approaches (plus it's hard to overlook the Holo's NOS ability) but the Hugo / Mojo oversample at 2048x and I can't help thinking which oversampling method sounds better?

 

[elan120]

I built a ultra quiet gaming computer to handle DSD512.
Fan's are only 15dB. With the side panel on, 10dB, so fan noise is not an issue.

The motherboard used was an ASUS STRIX X399-E.

A server board, board graphics, with standards memory would be much cheaper.[/QUOTE]
@Energy , Excellent looking server! I will check out this mobo and see if I can search for fans and case that will be as quite as yours.

 

[Energy]

Very nice build @Energy . Glad to hear your Threadripper investment paid off.

It'll be a while before I can build another machine and I'm a fanless kind of guy (built a dozen or so fanless HTPCs and other machines for music over the years). I doubt the insane core counts and 180W tdp of Threadripper could be passively cooled but the upcoming Calyos NSG-S0 (http://fanlessfan.com/) may be able to handle it (assuming my imaginary future wallet could handle the parts list).

Anyway, I've been wondering how DSD512 on your Holo Spring compares to Chord's DACs (assuming you've heard them)? Technically, I know they're two different approaches (plus it's hard to overlook the Holo's NOS ability) but the Hugo / Mojo oversample at 2048x and I can't help thinking which oversampling method sounds better?

Thank you and much appreciated. The threadripper has been a huge help towards pursuing high end audio. To be able to play these insane filters without fearing computer lag when doing day to day activities is incredibly soothing.

It would be hard to build this machine without some form of fan. Are you more eccentric towards fans for their actual noise or the electrical noise they generate?
A fan running at 500RPM will only produce about 10dB of noise and if you use a good enclosure such as the ones provided nowadays with fan noise in mind, they all have internal dampening through the use of closed cell foam. It does a very good job in eliminating internal part vibration energy as well as the listenable noise you hear. I use a fancier tempered glass computer case (Corsair 570X) where there are larger vents for air but even in a room where studio monitors are used, the fans cannot be heard when the side panels of the case are on. It does not disrupt music listening whatsoever for me. Nowadays with all the transport technologies coming out, there is very little to gain from using linear power supplies on motherboards, SSD's, etc or having heavy urge to eliminate all fans that cause electrical noise such as current ripple.

I have had experience listening to the Chord Dave. The POLY-SINC-XTR filter within HQPlayer utilizes very similar algorithm to what the large FPGA (or supporting components) within the Dave does. In classic oversampling DAC's there are usually two or more low pass filters used so the availability of NOS mode on the Spring is a good thing to have, however like with what Alex from UpTone Audio says, I feel there is still some low pass prior to the R2R network. Regardless of which, 512, 1024, or 2048, I still would say a modern day computer is better for oversampling due to the quality it brings in terms of noise shaping, better filters, and higher processing and algorithms used that directly translate to better sound quality.

I've also have had owned a Hugo and Mojo quite a number of times.

Excellent looking server! I will check out this mobo and see if I can search for fans and case that will be as quite as yours.

It's more of a gaming motherboard than a server board but should service nicely! Be on the lookout for 500-1500RPM fans! Maybe find some below 500RPM if you want super duper quiet.

 

[elan120]

It's more of a gaming motherboard than a server board but should service nicely! Be on the lookout for 500-1500RPM fans! Maybe find some below 500RPM if you want super duper quiet.

I am still searching for a motherboard that isn't necessary can serve as gaming purpose, since I will be using this server primarily for music, but your success in building this server adds great encouragements using this motherboard. I had been checking in linear power supply area until your comments about possible diminishing return, and I hope you could elaborate further about this?

 

[gr8soundz]

It would be hard to build this machine without some form of fan. Are you more eccentric towards fans for their actual noise or the electrical noise they generate?

Maybe a little of both. No moving parts equals less interference and (more importantly) no dust. The first machine I fully built used heat pipes and after 10 years never had anything more than surface dust.

To anyone who hasn't tried fanless: It's hard to go back after using 0dB machines. You'd think you wouldn't notice the sound fans make (even super quiet ones) but I always notice it.

 

[louisxiawei]

@Energy Hi, thanks for the informative report. But could you please test the 44.1 redbook file upsampling → 48K *512 (24.576) instead of 22.579 with XTR non-2s filter? I believe this will be the most intensive setting to test out. Thanks in advance.

 

[Energy]

I am still searching for a motherboard that isn't necessary can serve as gaming purpose, since I will be using this server primarily for music, but your success in building this server adds great encouragements using this motherboard. I had been checking in linear power supply area until your comments about possible diminishing return, and I hope you could elabAorate further about this?

I understand, however building a computer or server is fairly easy. Any server or gaming motherboard that has the TR4 socket for the AMD ThreadRipper 1950X should work fine. For music purposes you don't even need a graphics card as long as your motherboard comes with some from of graphics output in either VGA, DVI, HDMI, or Display Port..

For the price I just felt it was a lot of money wasted on a CPU and so I made it into a gaming rig. With DSD512 music cranked up with the heaviest modulator and filter the CPU hits no more than 25% load. I think you can get away with the 12-core 1920X. As for computers fed with linear power supplies, that would be more current ripple and generated power noise on the 12V/5V rails. If one were to use something like the ISO REGEN, iGalvanic, or SOtM equivalent then it severs the computer connection through the use of galvanic isolation and the data stream sent from the computer to the transport gets retrieved, counted for (1's and 0's), it's USB signal get's regenerated from scratch using cleaner external power (or regenerated internal power) and a reclocked with a betterr performing crystal oscillator within the device before sending it onto a DAC or DDC. Through the use of these superior upstream transports, digital music sources and their power become less relevant. For example, if computer generated power noise affected my sound quality at all I would not have built a computer with 6 fans, a water pump, and graphics card. But the truth is, this newly made computer sounds identical to a fanless music server I built a year or two ago. Personally I believe that DDC (digital to digital converters) are the way to go. As an audiophile the recommended thing to pursue is galvanic isolation of your DAC from it's music source followed by picking the most optimal method of transmission for it in order to further reduce jitter (digital reflections that causes timing delays that can be heard). In order to do both, two separate DDC's are needed. I use the ISO REGEN and Singer SU-1. The first offers galvanic isolation from computer generated noise and recreates an ideal USB signal with cleaner 5V power. It feeds the second DDC (Singxer SU-1) and USB is then converted to a superior balanced digital output called I²S. I²S is superior due to it being a native language of the DAC so there is no need reconvert USB or S/PDIF to I²S when hitting the DAC.

I see you on the Computer Audiophile forums so I'm sure what I just said you should have already known but maybe someone else out there reading this may pick up on it. Cheers.

Maybe a little of both. No moving parts equals less interference and (more importantly) no dust. The first machine I fully built used heat pipes and after 10 years never had anything more than surface dust.

To anyone who hasn't tried fanless: It's hard to go back after using 0dB machines. You'd think you wouldn't notice the sound fans make (even super quiet ones) but I always notice it.

You're right. My pump started making slight crackling noises which means it's time to RMA and get a new one. These things happen sometimes when it comes to PC building unfortunately. Moments like this I miss the 0dB fanless solution.

@Energy Hi, thanks for the informative report. But could you please test the 44.1 redbook file upsampling → 48K *512 (24.576) instead of 22.579 with XTR non-2s filter? I believe this will be the most intensive setting to test out. Thanks in advance.

It's not possible to upsample 44.1KHz to 48KHz with any settings on HQPlayer no matter what modulator or filter is used.
There's also no point in doing so as it would take the sampling rate away from it's native resolution and that in theory will not sound better.

In the pictures I posted above, it's possible to convert 24/96KHz and 24/192KHz to 44.1 x DSD512, DSD256, DSD128, and DSD64 but nothing else.

 

[louisxiawei]

It's not possible to upsample 44.1KHz to 48KHz with any settings on HQPlayer no matter what modulator or filter is used.

Just quote Miska (aka Jussi)'s reply to me on CA here: This is because on Windows, some ASIO drivers (Thesycon as an example) don't include 48k-base DSD rates, even though the DAC supports such. This can be solved using Linux or alternatively a Linux-based NAA for the Windows machine.

Guess you need to add something like microrendu/ultrarendu or SMS-200 from SOTM to do the 48K upsampling.

There's also no point in doing so as it would take the sampling rate away from it's native resolution and that in theory will not sound better.

If you really think it will take the sampling rate away from its "native" resolution, you are already doing it. Because you are upsampling 48K file to 44.1*DSD512.

I've also discussed this with Miska long time ago. This is a common misunderstanding about upsampling. 44.1 base to 48 base is still rational ratio (44.1 -> 96 = 320/147, 48 -> 96 = 2/1) which means that it is possible to be equally good to 48-to-96 case and technically different only by one parameter.

Plus, at DSD512 the rates are 22.5792 MHz and 24.576 MHz - that is 1.9968 MHz difference, more than half of the rate of DSD64, if you do some research about DSD/SDM dac, you will know the importance of the sampling rate for the performance. Nevertheless, YMMV, but these explanation can be proved in my listening experience, I can feel obvious improvement when doing upsampling to 48K*DSD512 from 44.1K base file compared to 44.1K*DSD512. More spacial, accurate image, smoother and more refined sounding. I'm using T+A DAC8 DSD.

A final tip from me: for HQplayer upsampling, the best file will be DXD or the DSD files at higher sampling rate (DSF64,DSF128, DSF256), since red book 44.1/16 are really the worst lossless files and make your PC have a hard time to do the upsampling that's why from 44.1 → DSD512 needs the most intensive computing power. For less computing power and much better listening result, try higher resolution PCM files or DSD files, I believe DXD files will give you the best result since it has much more information compared to 44.1 or 48/96 files while it still can take advantage of SDM oversamping filter. DSF files only use the modulator settings, oversampling filter will not be used when you do DSDxxx → DSDyyy sampling, and I believe there are not too many DSD256 recordings on the market now, but there are still some, check them on nativeDSD, I'm sure these recordings will make you smile.

 

[Energy]

Just quote Miska (aka Jussi)'s reply to me on CA here: This is because on Windows, some ASIO drivers (Thesycon as an example) don't include 48k-base DSD rates, even though the DAC supports such. This can be solved using Linux or alternatively a Linux-based NAA for the Windows machine.

Guess you need to add something like microrendu/ultrarendu or SMS-200 from SOTM to do the 48K upsampling.

If you really think it will take the sampling rate away from its "native" resolution, you are already doing it. Because you are upsampling 48K file to 44.1*DSD512.

I've also discussed this with Miska long time ago. This is a common misunderstanding about upsampling. 44.1 base to 48 base is still rational ratio (44.1 -> 96 = 320/147, 48 -> 96 = 2/1) which means that it is possible to be equally good to 48-to-96 case and technically different only by one parameter.

Plus, at DSD512 the rates are 22.5792 MHz and 24.576 MHz - that is 1.9968 MHz difference, more than half of the rate of DSD64, if you do some research about DSD/SDM dac, you will know the importance of the sampling rate for the performance. Nevertheless, YMMV, but these explanation can be proved in my listening experience, I can feel obvious improvement when doing upsampling to 48K*DSD512 from 44.1K base file compared to 44.1K*DSD512. More spacial, accurate image, smoother and more refined sounding. I'm using T+A DAC8 DSD.

A final tip from me: for HQplayer upsampling, the best file will be DXD or the DSD files at higher sampling rate (DSF64,DSF128, DSF256), since red book 44.1/16 are really the worst lossless files and make your PC have a hard time to do the upsampling that's why from 44.1 → DSD512 needs the most intensive computing power. For less computing power and much better listening result, try higher resolution PCM files or DSD files, I believe DXD files will give you the best result since it has much more information compared to 44.1 or 48/96 files while it still can take advantage of SDM oversamping filter. DSF files only use the modulator settings, oversampling filter will not be used when you do DSDxxx → DSDyyy sampling, and I believe there are not too many DSD256 recordings on the market now, but there are still some, check them on nativeDSD, I'm sure these recordings will make you smile.

No wonder my 48KHz doesn't show up.

I don't use ROON as an end point so no poiunt in using microRendu (or equivalent).

I see what you mean. When you put it like that it makes perfect sense. The extra rate should be beneficial. Thank you for the advice however it's a little saddening for me since the majority of my music consists of KPOP and EDM which I can only find in rare trace amounts in 16/44.1 lossless thus why I'm trying my best in terms of improving the PCM to DSD upsampling chain. I've given DSD native music files a listen and they are incredible indeed.

Do you think it's possible to install a virtual Linux of some sort on the Windows machine for the 48KHz? I really want to give it a try but I cannot function without Windows OS.

 

[louisxiawei]

I don't use ROON as an end point so no poiunt in using microRendu (or equivalent).

Not really. If I were you, I would pick microrendu/ ultra-rendu without a second thought. It's not about Roon, it's about HQplayer NAA (network audio adapter). Have a look at the HQplayer site or join the CA community, lots of sensible nerds discussing such there.

In short, by having NAA, you will literally galvanically isolate your HQplayer DSP PC from your DAC since ethernet transmission is inherently galvanic isolated. So you will be less worried about the RFI/EMI your PC generates during intensive HQplayer processing transmitting to your DAC/interface via usb cable. You can literally place your PC and microrendu at two difference places. Disadvantage of it is: your cannot stream your youtube or movie through your system, but these things have crappy SQ, so I never bother streaming them from PC to my HIFI system.

NAA as a second setter, only do the bit-perfect transmission without doing heavy processing. So it can be a compact, small, low-power consumption, software optimized NUC (I regard microrendu as a linux-based NUC as well), you can have better signal integrity USB output and less noise problem. It's also easy for you to upgrade lots of accessories such as LPSU for your NUC (since they are low power consumption, find a good LPS-1 is much cheaper and easier than big linear power supply in your PC).

You can go for ultra-capacitor-based LPS like LPS-1 from Uptone or mini-dc-4ever from VR and your microrendu powered by these things are basically off the gird. You can also put a optical ethernet switch in between, and install a NIC pice card on your PC, so you transmit your HQplayer DSD512 data through your PC via optical Ethernet cable which is a step further of isolation compared to normal copper Ethernet cable. Thinking about these up-gradable and flexible options.

Only one thing you need to be careful is some DAC supports DSD512 (44.1k or 48k) only on windows OS like my T+A DAC and your DAC supports DSD512 at 48K base only on linux. I prefer a linux-based NUC since it save me the trouble to optimize OS and product like microrendu has optimised its USB output. But unfortunately, i cannot use them since my DAC's firmware does not support it.

I've given DSD native music files a listen and they are incredible indeed.

You can also upsample DSD64 file to DSD512 one without PCM conversion, this is also one of the unique selling points of HQplayer, which also yield wonderful result. Just untick the DirectSDM in the DSF/DSDIFF settings.


Hope it helps.

 

[wazzupi]

How does one go about purchasing a holo audio cyan...

 

[ahmadfaizadnan]

How does one go about purchasing a holo audio cyan...

I have been searching for them but I guess they are not out yet.