Jump to content
IGNORED

Building a DIY Music Server


Recommended Posts

On 4/15/2020 at 1:56 AM, Nenon said:

 

My new build is:

- ASUS ROG Crosshair VII Hero full size ATX motherboard 

- AMD Ryzen 7 3700X CPU

- Apacer ECC RAM

- Upgraded motherboard clock with PinkFaun ultraOCXO clock

- The new JCAT XE USB card

- JCAT NET Femto card

- PinkFaun I2S Bridge with ultraOCXO clock

- Optane card for the OS

- HDPlex H5 Case

- HDPlex 800W DC to DC ATX Convertor

- Gaia feet

- Mundorf Silver Gold DC wire all around

- Euphony OS

 

IMG_2982.thumb.jpg.149c965ce588dff540555e5569529fb9.jpg

 

I made that carbon fiber plate with two connectors that I quite like. One connector goes to the CPU / EPS and the other to the HDPlex.

IMG_2986.jpg.e51f18a6cfd93ee388cf6f68b2f0ac56.jpg

 

On the other side I just drilled a hole on the plate and installed a connector to power the OCXO clock. I will be adding a grounding post this weekend as well. 

IMG_2989.jpg.22fe70172a4c85c066e1fba4062dd313.jpg

 

And of course vibration isolation with Gaia feet:

IMG_2995.jpg.fc4113b7a00f3778ee41399bd6bfd469.jpg

 

 

There were three big questions I wanted to answer with this build.

 

I will be honest with you and tell you upfront that I did not spend enough time to apply the typical systematic approach of testing. Here is why - I am doing this as a hobby and when I hear that something I did convincingly sounds better, I am not interested to go back and forth and study how much better it sounds. I wish I spent a little more time with each small change, but I had too many thing going on. 

 

Let's start with the motherboard. Does it sound better than the mini-ITX motherboard? Yes, it does. But I don't think we should make the conclusion that every full size ATX motherboard sounds better than its mini-ITX sibling. That's not true, and I have seen the opposite. But this particular ASUS ROG Crosshair VII Hero sounds a little better than the ASUS ROG Strix X470-I Gaming mini-ITX motherboard. Not much better, just a little. 

 

After adding the PinkFaun ultraOCXO clock, this motherboard sounded much better than the mini-ITX model. Not all OCXO clock upgrades are up to my taste. They all seem to help with details, but some of them make the sound too sterile, too clinical, even lean in some cases. That's kind of what I expected by replacing this motherboard clock. But my expectations were wrong, and I was pleasantly surprised. The effect was quite different than what I thought. The first thing I noticed was a lot of air between the instruments. There was also more layers, more depth, and the bass became tighter and more real. It was immediately noticeable, but after a week I decided to go back to the stock motherboard clock just to double check. I did not listen for long with the stock clock. It was all confirmed immediately and I switched back to the ultraOCXO clock. Another thing I noticed was that the PinkFaun ultraOCXO clock improved a lot the USB output of the motherboard. Especially the USB ports that are connected to the chipset sounded a lot better with the clock. I think for the first time I actually preferred the USB ports that go through the chipset than the ones that don't. But ultimately I preferred the new JCAT USB XE card. More on that later. One little detail - this clock requires really good and clean power. The cleaner, the better. And good DC wires. My Mundorf silver/gold DC cables came to rescue again. 

 

Having said all that, it's hard for me to answer the second question. 

I don't know. It's a different motherboard with ultraOCXO clock. It definitely sounds better, but how much of that has to do with the fact that I did not have to use a PCIe extension cable, and how much was due to the different hardware? I have no idea. 

 

I was also wondering how the passive cooling on the HDPlex H5 chassis compares with the Streacom. I was thinking to install two identical motherboards and run some tests to see how efficient in heat dissipation each case was. Well, I will leave that test to someone else. A lot of IT youtubers out there do those experiments. I just don't have the time. 

But I think it's obvious that the HDPlex H5 passive cooling is much better. Streacom on the left, HDPlex on the right:

IMG_2999.thumb.jpg.d48388dbf495079aa8c8a59934fac6db.jpg

 

Streacom has 4 copper pipes. HDPlex has 8 copper pipes. Streacom has a tiny aluminum mounting plate. HDPlex has a massive copper piece with a big aluminum heatsink going on top. Streacom's pipes are flattened and they lose performance this way. HDPlex are not and go inside groves in the heatsink. And the HDPlex heatsinks on the chassis are bigger, at least compared to the Streacom FC9. Plus the HDPlex supports full size ATX motherboard with vertical PCIe cards and no risers.

The only things I hate about the HDPlex is that you need to remove the entire backplate to install PCIe cards. Maybe it's convenient for one card, but try doing 4 PCIe cards at the same time when you also have 3 DC connectors with soldered wires on the plate... not fun! But some people think that's the best thing about this case. I won't repeat the conversation that happened previously in this thread regarding that. Between the two cases, my preference is the HDPlex case now, but that's mainly because of the full size ATX support. The HDPlex has thinner bottom and top covers - that's not good for vibration treatment. 

 

Okay, what's next? JCAT USB XE Card. It's really good and keeps getting better and better over time. I really like it, and it's an easy recommendation. I can't tell how it compares with the PinkFaun USB bridge with ultraOCXO clock, but (besides the PF) I can tell you that it's better than any other USB card I have tried. It would be interesting to do an A/B comparison between the PinkFaun with ultraOCXO and the JCAT card.  

 

Let me touch on CPU choices a little bit. Emile from Taiko has done an extensive research on CPUs and shared some of his results on another forum. My understanding is that he has tried pretty much every CPU he thought might have a chance no matter the price. And interestingly enough he did not like the sound coming out of the AMD CPUs he tried. Romaz also built a computer with an AMD Ryzen 9 that has 105W TDP in a Streacom FC9 chassis. I don't think the Streacom chassis is up to the task to cool down a 105W TDP CPU. At that point you need to start reducing the voltage / speed on the CPU just to keep it cool enough. That in my opinion diminishes the sound quality. And even when my chassis get somewhat hot, I can hear degraded sound. The HDPlex is a better choice, but as far as AMD CPUs go I don't think we can go higher that the 65W TDP of the AMD Ryzen 7 3700X CPU I am using. 

 

Emile is a genius. We all know that. And it seems like he has resources we can only dream about. But I feel like it's difficult to compare CPUs and have a definitive conclusion about the sound of each CPU. Is it possible that the same CPU sounds very different in his system with his power supply and his choice of motherboard with his tweaked operating system than it sounds in my system with my power supply and my choice of motherboard and completely different operating system? I think the answer to that question is yes. Not only it is very possible, but it's also very likely. I really like the way this new server turned out. It's better than my previous AMD build, which I quite liked. It might be better them my previous Intel builds. It certainly sounds different and probably different people would have different preferences. 

But regardless of that, Emile's test sharing is very useful. Ultimately, there is only one way to find out, so here it is:

IMG_2974.thumb.jpg.8fed614781831df9d46c1883846bd03b.jpg

 

This is the ASUS WS C621E Sage Extreme motherboard with two Intel XEON SILVER 4210 CPUs that the Taiko Extreme uses inside an HDPlex case. It would be interesting to complete this build and compare with my other builds. Granted, I won't have the million small tweaks Emile has put in the Extreme, but I will apply all the stuff I have learned during my builds. Not trying to reach the Extreme but just to explore another route. I have a custom order Apacer RDIMM RAM coming in several weeks. Between now and then there is a lot of challenges I need to overcome. Don't ask me how I am going to passively cool down those CPUs for example. I don't know. There is no adaptor for the fclga3647 socket in the HDPlex or Streacom kits. I will have to build my own. I have several different Dynatron coolers coming my way as well as various cooling pipes and a pipe bending tool. Ideally, the CPU that is closer to the front of the chassis would be cooled by the left heatsink and the second CPU would be cooled by the right heatsink. That would be a heck of a project. 

Nice work. One of the reasons why server motherboards likely yield better sonics than what I'd call client motherboards is due to better quality components (resistors, capacitors and regulators in particular). In my day job I work with ODMs who design computers and in order to meet the higher server specs for constant uptime to manage critical data these components are much better quality than the 1-2 year longevity requirement for your run of the mill Acer, ASUS, Gigabyte etc. client motherboards. Then there are trace length and layout requirements that, for client MBs the ODMs will cut corners to save $ while again on the server motherboard they won't. There are other factors too, too much to type but you get the picture. 

4x20A circuits | Shunyata Triton + Typhon | Source 1: HDPlex HD100 PSU -> OCX clock + EtherRegen -> Paul Hynes SR4T + HDPlex HD500 PSUs -> Music PC w/JCAT XE nic, HQPlayer, Roon, Tidal / Qobuz |  TotalDAC D1 Twelve DAC + Mk II Streamer | Source 2: Acoustic Signature Ascona with Kuzma 4 Point tonearm | Koetsu Rosewood Signature cart | Pass XP-15 phono pre | Audionet Pre G2 preamp | Audionet Max mono blocks | Vivid Audio Giya Spirits |  4 JL Audio Fathom subs | Echole ICs /SCs / Siltech King jumpers, Shunyata/Audioquest PCs / Eth. Cs | Critical Mass CS2s -> components, Isoacoustics -> speakers + subs | Adona Eris II rack w/ Herbie's titanium footers | Fully treated and dedicated 2 channel room

Link to comment
On 11/12/2020 at 10:03 PM, sandyk said:

Marcin

Unless it uses separate transformers for both sections then it will never be quite as good as 2 separate hi-end linear power supplies assuming that they both  share a common earth reference. There is always some minor degradation when using separate secondary windings on the same transformer as I found when making 2 low noise regulated PSUs for USB use.

Kind Regards

Alex

Alex, just  to add 2 cents: the power transformer in optimo 3 duo has lower inductance which makes it much larger compared to other transformers of similar power. And that results in a more stable magnetic stream. There is almost no audible difference between using 2 separate optimo 3 duo psus and one with 2 outputs. 

 

Regards, 
Marcin

JPLAY & JCAT Founder

Link to comment
1 hour ago, Nenon said:

I have. Both are really good and really a matter of personal taste. The question is which nuance of good do you prefer. It's also dependant on the synergy with the rest of your server components and the entire system. Sorry I don't have a clear answer saying X is better than Y by this much, but that's how many things in audio are. There is however a significant price difference that needs considering too... The PinkFaun USB ultraOCXO is almost double the price of the JCAT XE.  

 

I have done some of these experiments too. I replaced the chipset clock with an ultraOCXO clock. There are USB ports that are directly connected to the CPU and USB ports that are connected to the chipset. Before the clock upgrade, the CPU USB ports sounded better. After the clock upgrade, the chipset USB ports sounded better. Much better. But adding a good external card with good power was still better. The onboard NIC on this motherboard had a different clock. I did not try replacing it, but I am sure it would be an improvement. This is what Innuos did in the Innuos Statement.

 

I have similar experience with Sean Jacobs DC4 LPS, which is using very similar but larger transformers (6 times larger I believe). 

1370025136_ScreenShot2020-11-17at12_16_13PM.png.ace4855f29d81c082fef7e916dbe7d1b.png

 

1788379799_ScreenShot2020-11-17at12_20_38PM.png.3d39355f3da5785eaf2813184948beb3.png

I guess it still depends on what you are feeding with the power supply. Feeding a JCAT Net Femto and a JCAT USB XE cards with two rails from the same transformer vs. two different transformers would be very difficult to hear. But feeding something like the Chord Electronics Hugo M Scaler, which is a noisy device, would make an audible difference. 

A good quality transformer is certainly playing a big role in the power supply and that shiny transformer used by Marcin and Sean is the best I have heard so far. I am a big fan of good quality components, and it makes me sick when I see 5-figure audio components using the cheapest parts possible and making false claims about how they were specifically "voiced" with them. Actually that's the main reason I got into DIY audio. 

 

@Marcin_gps When is the JCAT XE Network card coming up? I think a lot of the readers on this thread are interested. 

 

 

Extremely helpful - many thanks

Link to comment
1 hour ago, Nenon said:

A good quality transformer is certainly playing a big role in the power supply and that shiny transformer used by Marcin and Sean is the best I have heard so far.

 Did you ever try bypassing the fuse in series at the output of Sean's PSU to see if it made any difference ?  

( I may have forgotten your reply)

Obviously, for normal use it still needs a fuse somewhere, but the fuse at this location MUST increase the PSU's output impedance, although a low value cap straight across the output terminals will help the HF area a bit.

 

How a Digital Audio file sounds, or a Digital Video file looks, is governed to a large extent by the Power Supply area. All that Identical Checksums gives is the possibility of REGENERATING the file to close to that of the original file.

PROFILE UPDATED 13-11-2020

Link to comment
1 hour ago, sandyk said:

 Did you ever try bypassing the fuse in series at the output of Sean's PSU to see if it made any difference ?  

( I may have forgotten your reply)

Obviously, for normal use it still needs a fuse somewhere, but the fuse at this location MUST increase the PSU's output impedance, although a low value cap straight across the output terminals will help the HF area a bit.

 

Alex - the standard DC4 does NOT have a fuse in series at the output! It's easy for someone to take this out of context, so I felt it's good to clarify it. Your comment was regarding a DIY power supply @elan120 built. He was nice enough to share his step by step building process and listening impressions in the following excellent thread:

Removing that output fuse on the high current does make a difference. I suspect moving it to another location to reduce the output impedance (while still having it as a safety measure) coould also be somewhat beneficial. I have sent my feedback to elan120, including some other things he can try, and it's up to him if he wants to pursue them or not. 

 

BTW, I did respond a few pages earlier. Maybe you have missed it. All good. Ping me on PM again if you have other questions.

Cheers.

Industry disclosure:
https://chicagohifi.com

Dealer for: Taiko Audio, Conrad Johnson, Audio Mirror, and Sean Jacobs

Link to comment
27 minutes ago, Nenon said:

BTW, I did respond a few pages earlier. Maybe you have missed it. All good. Ping me on PM again if you have other questions.

Cheers.

Nenon

You did not respond to my inexpensive suggestion about a method to bypass that output fuse in the high current supply rail.

PM sent.

Alex

 

How a Digital Audio file sounds, or a Digital Video file looks, is governed to a large extent by the Power Supply area. All that Identical Checksums gives is the possibility of REGENERATING the file to close to that of the original file.

PROFILE UPDATED 13-11-2020

Link to comment

Has anyone tried the i7-9700t?

 

It is more or less the same price as i7-8700t, but has 2 more cores (vs. 8700t).

However, the number of threads is only 8 (vs. 12 of 8700t).

The 9700t goes up to 4.3 Ghz,while 8700t to 4.0 Ghz.

 

Which CPU would be your choice, in case of very similar prices?

 

I would be using Audiolinux, Roon and HQPlayer.

Link to comment
5 hours ago, Nenon said:

I am cleaning up some of my older parts and servers and have one of my previous AMD builds that is collecting dust. 

 

Specs are:

AMD Ryzen 7 3700X
ASUS ROG STRIX Z390-I Gaming mini-ITX motherboard
2 x 4GB Apacer D31.23185S.001
32GB Optane MEMPEK1J032GA01
Streacom FC9 ALPHA HTPC Chassis
Gaia feet

IMG_4604.thumb.jpg.c3d16d0e80c161b1ef403bd0bf0b50b0.jpg

As you can see in the pictures I have installed a heavy duty Jaeger connector (really the best I could find) for the CPU/EPS connector. I can provide the matching connector on the other side. 

There is also a m.2 to PCIe extender installed in case you need to use two PCIe card. I had the USB card installed on the motherboard PCIe slot and the network card installed on the extender part.

There is a m.2 Optane card on the bottom m.2 slot with Euphony installed (license is not included). 

 

I have some additional connectors (DC connectors and a grounding post) installed on the back side:

IMG_4606.thumb.jpg.4f63bc4a510bf88a3e0c56af3eb02666.jpg

They currently go to 4 x Sean Jacobs regulators. Unfortunately, I would have to remove those. Sean has an agreement with Innuos not to sell regulators inside a computer. Now that I am affiliated with Sean, I would rather not cause troubles. Although I made this way before I had any affiliation with Sean. Too bad, because that was an excellent power supply, and a lot of Mundorf silver/gold wire will get wasted. I can help the new owner with some advices on how to power it up. 

 

When I remove the regulators, I estimate there are about $1500 in parts inside. I haven't decided how much to sell it for. When I do, I will list it. But if you have any questions feel free to PM me. 

IMG_4607.thumb.jpg.cb477d2b88d895e21502cbfd21c04870.jpg

 

IMG_4605.thumb.jpg.7a5bcdce7186eab854187364cc8bedba.jpg

 

If you add USB and NET cards and a good power supply this would be a very competitive server/streamer!  Easy starting project for someone who wants to get involved with DIY server building. 

Solid build.

 

Still contemplating whether to go i9 9900k or to upgrade to full ATX z490/w480 based board with top tier 10 gen CPU.

 

Leaning towards the latter for the built in remote management functionality. It would be great to not have to drag my server to another room and connect my desktop monitor to play with RAM settings :).

Link to comment
15 hours ago, Csaba said:

Has anyone tried the i7-9700t?

 

It is more or less the same price as i7-8700t, but has 2 more cores (vs. 8700t).

However, the number of threads is only 8 (vs. 12 of 8700t).

The 9700t goes up to 4.3 Ghz,while 8700t to 4.0 Ghz.

 

Which CPU would be your choice, in case of very similar prices?

 

I would be using Audiolinux, Roon and HQPlayer.

 

I have been using an i7 9700t in a mini itx board, AL, and HQp HDplex running turbo on in AL. Sounded very good no problem with over heating unfortunately the board malfunctioned, under warranty but am still waiting for replacement. 

In the meantime purchased an msi msg on offer and i9700k. This combination sounds better than the original uspsamples to DSDx2 with ASDM7EC and still runs slightly cooler than the T version on turbo. My advice would be to go with the i9700k as long as you have a decent cooling implementation (95w).

Link to comment
2 hours ago, dctom said:

 

I have been using an i7 9700t in a mini itx board, AL, and HQp HDplex running turbo on in AL. Sounded very good no problem with over heating unfortunately the board malfunctioned, under warranty but am still waiting for replacement. 

In the meantime purchased an msi msg on offer and i9700k. This combination sounds better than the original uspsamples to DSDx2 with ASDM7EC and still runs slightly cooler than the T version on turbo. My advice would be to go with the i9700k as long as you have a decent cooling implementation (95w).

 

Sounds very promising, however my Streacom FC-10 case is specified for max. 65W TDP processors.

Therefore, maybe I should stay with i7-9700T.

 

Is there a big difference in sound between "K" and "T" versions or only subtle?

Link to comment

Happy Thanksgiving!

 

Did anyone build a 10th gen Intel i9-10900K or any of the Comet lake series ? I am curious how you are managing the higher TDP of 120Watts ?  Will the HDplex H5 be upto the task ? And what motherboards are you using with it ? Any comparison to the 9th gen will be nice ?

 

 

Link to comment
1 hour ago, bit01 said:

@Dev Happy Thaanksgiving! FYI

 

 

 

 

@bit01 awesome. Thanks you. How do you like the sound ? yes, my application will be basic Euphony/Roon playback without any upsampling. So I am assuming I can keep the temp within 50deg C. My 8-core Xeon E is able to keep within 50deg C with normal Roon playback. Do you hear any noise out of the CPU if you put your ear close to it ?

 

Have anyone compared between Gigabyte and Asus mobo for the 10th gen Intel ?

 

I am thinking Aus ROG MAXIMUS XII HERO but most Asus mobo seems to have shared 12V rail between the CPU EPS and 24-pin ATX. Not sure how to deal with this if I want to power the EPS separately. Anyone know ?

 

 

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...