Building a DIY Music Server

[dminches]

41 minutes ago, Dev said:

I am noticing that many are looking to build an Extreme as DIY with all sort of requirements - best parts, best chassis, best power supply, TAS, Taiko USB card, etc. If this is going to mimic the Taiko Extreme as close as possible, I wonder what's left of Extreme ? I am sure Taiko is not going to give away their IP but its good to understand how far apart or how close this build will be from it so that some can decide if it makes sense just to buy the Extreme instead ?

 

Some do DIY for the love of it, irrespective of the cost, some do it for the savings, while many are in both the categories. I belong to both and for me its critical to understand the price to performance ratio. If I need to spend $15k for this build, I would probably have double thoughts on spending couple of grand more and getting a commercial server instead. What do others think ?

 

No matter how much Emile contributes to this effort I don’t think what people end up with will be the equivalent of an Extreme.  That doesn’t mean the final product won’t be great, but the only way to get an Extreme (hardware, customer service, some upgrade options, etc) is to buy an Extreme. Buying one costs over $25k so spending $15k is quite a bit less, albeit still expensive.  I certainly understand how people would consider a commercial product if they were laying out $15k.

 

[Nenon]

Glad we got the $10K out of the way. It's just a number I am trying to stick to. It could be a little bit less or more. Otherwise, the cost would end up uncontrollably high. I am not good at making compromises. And I don't think Emile likes to make a lot of compromises either. So, this would be a very high-end DIY server. The best I know. 

 

Most of the people who have been chasing DIY music server perfection have realized how quickly things add up. When you add power supplies, clocks, reclockers, servers/streamers, USB cards, network cards, cables, vibration treatment, network switches, software, etc. etc. you end up with a big spaghetty mix and spend a lot of money. It's actually quite easy to reach $10K and still be unhappy with the digital source. Most of these people will appreciate what we are trying to do here and realize that $10K is not that much. 

 

If you are using an Intel NUC and a TeraDAK power supply (not picking on either of these, just the first products that came up to my mind), trying to do DIY on a budget, that project is probably NOT for you. There are many paths people can take and none of them is wrong. There are many proven more affordable DIY recipes, some of them posted on this same thread. We don't need Taiko's help for those. 

 

3 hours ago, Dev said:

Will there be any option for the chassis only ?

Yes. You would be able to buy just the chassis.

Taiko is partially sponsoring the DIY community... this is not a profit-driven initiative.  The chassis would be great for other projects too. I see it as an end game top quality DIY chassis. 

 

2 hours ago, MarcelNL said:

It would be great to offer alternatives for each major component, as in; 1st choice CPU, 2nd choice CPU 3rd choice CPU, same for PSU etc...

Well, this is DIY. Feel free to get just the parts you need. Experiment with different choices and report back.

My focus would be to provide a complete DIY recipe for people to start with. Giving many options is difficult, complicated, and confusing. Plus, change one major component and you have to re-evaluate everything. I would try to stay on track and provide whatever I can to build a DIY server tweaked for the Asus Sage motherboard with dual Xeon CPUs.

And hopefully everyone who builds it would experiment and contribute something back to the community. 

Also, the $10K target is for the basic build. If people want to unleash their budget there might be some more exotic parts to recommend.  

 

2 hours ago, MarcelNL said:

Having TAS available would be great , especially if they could add the undressed W version they use

Taiko would not provide Windows LTSC. That is one thing I am pretty sure about. But there will be a guide how to obtain a legal copy of Windows LTSC. I was hoping @dminches could do that part. 

 

1 hour ago, Dev said:

I am noticing that many are looking to build an Extreme as DIY with all sort of requirements - best parts, best chassis, best power supply, TAS, Taiko USB card, etc. If this is going to mimic the Taiko Extreme as close as possible, I wonder what's left of Extreme ? I am sure Taiko is not going to give away their IP but its good to understand how far apart or how close this build will be from it so that some can decide if it makes sense just to buy the Extreme instead ?

This will not be a Taiko Extreme equivalent. The Extreme is packed with technology and tweaks that will not be shared with our DIY community. Most of them are not visible on the pictures, but some are. I just called this chassis end game for DIY, but it is nothing compared to the Extreme chassis. We won't have the same power supply as the Extreme. We won't have the panzerholz vibration treatment. We won't have the custom Lundahl choke. We won't have the custom RAM that Taiko developed with reduced refresh rates and other custom specs. We won't have the copper heatsink and copper plates. Forget about capacitor in panzerholz. Those are just some of the obvious things that you see.

Take this post for example. It shows the level of attention to details to just pick an IEC inlet for the Extreme. There was similar posts about the fuse selection. And people who tried many audiophile fuses could not necessarily beat the stock fuse on the Extreme. We won't get that. But hopefully we would perfect the DIY recipe with the help of our community. 

There are dozens or even hundreds of tweaks on the Extreme that are not visible to the eye. Then we have all the software tweaks. But most importantly we won't have Emile's or Taiko's support. The Extreme is a platform, not just a server. If you follow the Taiko thread on WBF, you can see how the Extreme evolved over the past two years and people were getting major sound quality improvements after a TeamViewer support session with a member of the Taiko team. And now they even developed their own software player. 

We would get a few parts, some hints and tips, and we are on our own from there. The final product would be as good as the overall community effort. 

 

3 hours ago, Dev said:

I own’t ask for details but is it from Taiko or SJ ?

The truth is I don't know yet. I am testing various things including parts from SJ and Taiko. For example, I tried 6 different transformers, and I prefered one of the six by a big margin. There are quite a few unknowns about the power supply yet (although the main technology has been decided on). That's why I have been avoiding the subject. I know it's one of the most important subjects. We'll get to that part. 

 

4 hours ago, Darryl R said:

I'm sure you'll fire up some of your group buys for off-the-shelf components, to keep those costs down, right?

I wasn't planning to do group buys, but we can consider. It takes a lot of time to do that.

[MarcelNL]

@Dev: I could not agree more, if this is an all out attempt to build an 'extreme class' DIY server; cost no object but for labor cost, VAT and profit margin we are likely missing the point of DIY for most folks. As much as I'm able to buy an Extreme today if I wanted I won't do that for the very reason I love tinkering and discovering what makes audio tick, especially when this is done together with others.

Much of the high End customer scene for me is too close to the Little Britain character Andy 'I want that one'  Pipkin, if you know what I mean.

 

 

[Exocer]

1 hour ago, Dev said:

I am noticing that many are looking to build an Extreme as DIY with all sort of requirements - best parts, best chassis, best power supply, TAS, Taiko USB card, etc. If this is going to mimic the Taiko Extreme as close as possible, I wonder what's left of Extreme ? I am sure Taiko is not going to give away their IP but its good to understand how far apart or how close this build will be from it so that some can decide if it makes sense just to buy the Extreme instead ?

 

Some do DIY for the love of it, irrespective of the cost, some do it for the savings, while many are in both the categories. I belong to both and for me its critical to understand the price to performance ratio. If I need to spend $15k for this build, I would probably have double thoughts on spending couple of grand more and getting a commercial server instead. What do others think ?

 

@DevVery good question! A question many will find tough to answer. 

 

@NenonWhile I am aware we are in the early stages of development, I would greatly appreciate if you conducted the following comparison in the future:

 

1. External DC4 19v into HDPlex 800W ATX-DC for minor voltages + Dedicated 12V EPS rail.

2. SOTA LPSU with dedicated rails (I assume this is the direction things are headed, but you can correct me if I'm wrong here)

 

Yes, the first scenario is a compromise, but a compromise I am and many others may be willing to take if it means saving thousands and having a system within 10% of a system with dedicated rails. Some food for thought.

[Gavin1977]

I think it should be affordable... in reality not many will go for a dual CPU option and all those banks of expensive RAM. The budget necessary would stretch into thousands. Those people who can do this might as well buy an Extreme.

 

A chassis with an integral multi-rail linear PSU capable of accepting and powering single CPU's upto 120w, with dual molex 5v outputs to power a pair of JCAT cards would better fit my bill.  Price £1500?

 

Perhaps it might be better to view the product as a multi-rail linear power supply, but with the bonus of it including a chassis?

 

I'm also not against some limited zero noise / vibration isolated fan assistance if it cuts down costs & size.

 

Go over £1500 and I'd rather start looking at commercial offerings to be honest.

 

I agree with comments about the enclosure size - 480mm width is a bit too much for most.

[MarcelNL]

@Nenon; Thanks for clarifying, the way I see it we should not underestimate the power of DIY....as example, and not talking down what Taiko is doing, capacitors in Panzerholz is nice but in a whole the effect is not even close to 'undressing' a capacitor and sealing it in non metal case. Folks are trying exotic things, there are more means to an end etc... 

 

This is not bragging or trying to diminish anyones work in any way, just trying to clarify there is a wealth of knowledge out there that, when combined,  could work wonders and would make up for an interesting journey (also for Taiko). 

 

If this is a  'Mahlen nach Zahlen'  (paint by numbers) exercise it's more like building a kit, do some soldering, stick in some parts and screw a few things together, count me out.

 

[Gavin1977]

+1 @MarcelNL I bought a load of Salas L-adaptor pcbs... but you know what I just can't be bothered to solder them all up... it just takes too much time.  Good quality connection wire is also expensive, as are the right crimping tools.  Modular completed components is better for me and less likely for me to end up swearing at them.  I love the HD Plex chassis I have, I would certainly consider paying more to have a 'deluxe' version of this with an integral psu.  Love Paul Hynes and Sean Jacobs stuff, the prices are fine if you just want to power a single rail NUC, but count me out on based on multi-rail prices.  Perhaps I'm HDPlex's target audience, price of the H5 chassis and their new 500w linear is what I'm comfortable with.

[Nenon]

16 minutes ago, Exocer said:

@NenonWhile I am aware we are in the early stages of development, I would greatly appreciate if you conducted the following comparison in the future:

 

1. External DC4 19v into HDPlex 800W ATX-DC for minor voltages + Dedicated 12V EPS rail.

2. SOTA LPSU with dedicated rails (I assume this is the direction things are headed, but you can correct me if I'm wrong here)

I think I've answered that before. #2 sounds better than #1. But #1 gives you a real ATX controller. It's a balancing act. 

 

8 minutes ago, Gavin1977 said:

I think it should be affordable... in reality not many will go for a dual CPU option and all those banks of expensive RAM. The budget necessary would stretch into thousands. Those people who can do this might as well buy an Extreme.

 

A chassis with an integral multi-rail linear PSU capable of accepting and powering single CPU's upto 120w, with dual molex 5v outputs to power a pair of JCAT cards would better fit my bill.  Price £1500?

 

Perhaps it might be better to view the product as a multi-rail linear power supply, but with the bonus of it including a chassis?

 

I'm also not against some limited zero noise / vibration isolated fan assistance if it cuts down costs & size.

 

Go over £1500 and I'd rather start looking at commercial offerings to be honest.

 

I agree with comments about the enclosure size - 480mm width is a bit too much for most.

Appreciate your feedback @Gavin1977. This project is certainly for a very limited group of people. That's pretty clear to me. 

For a budget of £1500, we don't need Taiko's help. 

 

8 minutes ago, MarcelNL said:

If this is a  'Mahlen nach Zahlen'  (paint by numbers) exercise it's more like building a kit, do some soldering, stick in some parts and screw a few things together, count me out.

This won't be a kit. Taiko is not interested in selling a kit. I am not interested in selling a kit either. If you buy a kit, you would expect to get commercial support. Taiko would not be providing commercial support for this project. Actually, I am going to quote Emile:

Quote

The idea here is for this to be a non commercial community project, we supply a few quality building blocks which are currently not available in the DIY market. Support and further development would have to come from the DIY community, our input would be limited to general building guidelines and some tips & tricks.

 

It would be a DIY recipe with recommendations for parts, software, etc. You can use that as a baseline or use whatever parts you need to build whatever you want. Or not participate at all.

What I am hoping to see happening is people building the baseline and improving from there. Think of this as an open source project where everyone can contribute to. No secrets, it's all an open book, and every idea for improvement counts. 

[MarcelNL]

I just want to clarify that I DO appreciate the initiative of both you Nenon, and Taiko! 

[matthias]

@Nenon

 

Hi Nenon,

 

as TAS is based on LMS it could work on different platforms.

 

Do you know if Taiko Audio have plans to offer TAS as software player (without hardware) like Audirvana etc. on different platforms?

 

Thanks

 

Matt

 

[Nenon]

15 minutes ago, matthias said:

Do you know if Taiko Audio have plans to offer TAS as software player (without hardware) like Audirvana etc. on different platforms?

No idea. Given that it is still in alpha, my guess is they probably don't know either at this stage. 

[Exocer]

1 hour ago, Nenon said:

I think I've answered that before. #2 sounds better than #1. But #1 gives you a real ATX controller. It's a balancing act. 

 

Yes, 2 definitely would sound better. Thank you.

My question was more about how close one can get with two rails and an DC-ATX converter vs dedicated rails in a dual Xeon build.

 

If building dual a CPU RIG only brings you slightly above a well executed i9 rig when you're limited to two rails of high quality power and an HDPlex DC-ATX converter, the entry price may not be worth it for many.

 

If building a dual CPU rig (again limited to two rails) can substantially improve upon a well executed i7/i9 rig with the same power, this journey becomes a lot more attractive (to me).

 

Does a build like this require dedicated rails to truly sing, is the question.

 

Edit: I am still very much on board either way.

 

[Darryl R]

6 hours ago, Nenon said:

I use HQP and Roon and will share what I do. That can be used as a starting point. But I am hoping the DIY community would test other software and come up with better options. 

 

Interesting, so you came up with something you like better than Euphony.  I always liked HQP, and have been meaning to try HQPE (embedded), but I never experienced the SQ problems others claim with Roon so I've stayed with it, and it's by far the best UI (though I have no big problem with the HQP UI options).  The combo is interesting.  You don't have to get into that now while you're focused on hardware.

[Nenon]

17 minutes ago, Exocer said:

Yes, 2 definitely would sound better. Thank you.

My question was more about how close one can get with two rails and an DC-ATX converter vs dedicated rails in a dual Xeon build.

 

If building dual a CPU RIG only brings you slightly above a well executed i9 rig when you're limited to two rails of high quality power and an HDPlex DC-ATX converter, the entry price may not be worth it for many.

 

If building a dual CPU rig (again limited to two rails) can substantially improve upon a well executed i7/i9 rig with the same power, this journey becomes a lot more attractive (to me).

 

Does a build like this require dedicated rails to truly sing, is the question.

The dual Xeon server I built for myself with the mockup HDplex passive cooling with a 3-rail SJ LPS (one for the HDplex 800W DC to DC ATX, one for the 12V EPS connectors, and one for the JCAT XE card) sounded significantly better (after a lot of tweaking) than my previous builds. It's worth building it with your existing power supplies @Exocer. Can't comment about other people, but I know a bit about your system from our private chats, and I feel confident you would like it. 

And there will be some other options to try as this project unveils. 

 

10 minutes ago, Darryl R said:

 

Interesting, so you came up with something you like better than Euphony.  I always liked HQP, and have been meaning to try HQPE (embedded), but I never experienced the SQ problems others claim with Roon so I've stayed with it, and it's by far the best UI (though I have no big problem with the HQP UI options).  The combo is interesting.  You don't have to get into that now while you're focused on hardware.

I liked Euphony on other servers. On this dual CPU hardware I prefer Windows LTSC. I use HQP+NAA for bitperfect local files playback (I don't upsample) and Roon for streaming Qobuz. I don't use both together... some people do that to upsample the Roon output, but that's not my thing. 

When you align the logical motherboard diagram with the hardware placement and careful configuration of process affinities, the magic happens.

For example, I have

- The motherboard onboard NIC, all Windows processes, and HQP all assigned to one CPU.

- The Optane card for the OS is physically placed on a PCIe slot connected to the CPU above. 

- I have my USB card physically placed on a PCIe slot directly connected to the other CPU and assign NAA and any other audio related processes with affinities to that CPU.

It makes a difference! 

 

[Darryl R]

3 hours ago, Nenon said:

This will not be a Taiko Extreme equivalent. The Extreme is packed with technology and tweaks that will not be shared with our DIY community.

 

On the other hand, you mentioned Taiko was willing to share some tech which was not in the Extreme.  That should be interesting, but either way, it's hard to truly critique the pricing estimates until we see the power supply, which as we all know is a big cost variable (I was hoping somebody has a breakthrough prototype in SMPS for this project, which could shrink chassis size requirements ;)

[ted_b]

I gotta assume power supplies are half the budget.  Given what the likes of Sean, Paul Hynes, etc cost (although Nenon can maybe help here in a group buy :)  ) it will be a good goal to stay around $10k all in.  

[MarcelNL]

So adding all up am I correct in thinking that this route will be leading to a dual Xeon based server in a seriously nice case, operating under some Windows OS ? And potentially having TAS available.

 

I see the potential advantage of a dual CPU setup, and even when I'm using Daphile I have nothing against some other OS if it sounds as good or better. Daphile does not (or I am wrong mainly due to a lack of documentation) allow for simple task control at the core level. I am reading the Daphile thread on this board and see some have been tinkering with assiging IRQ's to a core, yet how the current Daphile version handles this is not very clear.

 

With an abundance of cheap used Xeon MB's and CPU's it might be a feasible build even when on a budget, the big question is; is the TDP of older CPUs too large for the case?

 

[seeteeyou]

Regarding TDP, that really depends on whether we're going for a single or dual Xeon setup

 

https://en.wikipedia.org/wiki/List_of_Intel_Skylake-based_Xeon_microprocessors

https://en.wikipedia.org/wiki/List_of_Intel_Broadwell-based_Xeon_microprocessors

https://en.wikipedia.org/wiki/List_of_Intel_Haswell-based_Xeon_microprocessors

https://en.wikipedia.org/wiki/List_of_Intel_Ivy_Bridge-based_Xeon_microprocessors

https://en.wikipedia.org/wiki/List_of_Intel_Sandy_Bridge-based_Xeon_microprocessors

https://en.wikipedia.org/wiki/List_of_Intel_Nehalem-based_Xeon_microprocessors

https://en.wikipedia.org/wiki/List_of_Intel_Core-based_Xeon_microprocessors

 

For a config with single Xeon, the range is all the way from Xeon E3-1220L v3 @ 13W TDP to Xeon E5-2699P v4 @ 300W TDP.

 

When it comes to dual Xeon, the only choices (preceding Xeon Scalable) are listed below

• Xeon 5000-series
• Xeon 7000-series
• Xeon E5-24xx
• Xeon E5-26xx
• Xeon E5-46xx
• Xeon E7-28xx
• Xeon E7-48xx
• Xeon E7-88xx

E5 and E7 actually went from v1 to v4 but we don't have to be Captain Obvious here, they're ranging from Xeon L5508 @ 38W TDP to Xeon E5-2699P v4 @ 300W TDP.

 

---

 

However, let's take a look at a little something below and then we'll realize that investing in anything less than Asus WS C621E SAGE might be considered a "waste" of some sort

 

https://www.whatsbestforum.com/threads/taiko-audio-sgm-extreme-the-crème-de-la-crème.27433/page-88#post-613373

Quote

As for the stock server-grade ASUS motherboard used in the Extreme, I have often wondered why Emile didn't design his own motherboard like Antipodes and Innuos have and it turns out Emile has. As I have a background heavily entrenched in DIY, I have explored motherboard design myself and if one has the capital, any of these motherboard manufacturers will gladly build you one. What Emile found, however, was the motherboard he designed didn't sound better. In fact, he tested dozens of boards and I was able to witness for myself some of these boards that failed to make the cut including several that were in active prototype form.

 

Simply put, even best of the best couldn't design a better one on his own. Besides, other manufacturers might have to spend at least quite a few years to catch up with what SGM Extreme could offer right now.

 

---

 

BTW, talking about TDP alone might not be all that meaningful since the actual power consumption would always depend on what we're doing with the CPU in the first place. A pair of Xeon Silver 4210 @ 85W TDP should be doing fine without any upsampling with HQ Player, though ASDM7EC would be a very different story.

[Summit]

18 hours ago, austinpop said:

I love the direction this is going! I anticipate this being one of the most important threads on AS in 2021.

 

You’re probably right, but don’t forget inspiring threads like MQA in Vaporware and Fas42’s Stereo ‘Magic’ 😁.

[MarcelNL]

To me it's great to have some of the experience from Taiko available as loose map of the scenery but I'd like to avoid tunnel vision of focussing too much on their route, their solution apparently works great yet there likely are some other solutions that work too.

 

We will not likely build a DIY Extreme, as mentioned; if I wanted one and thought the 25k or so are justified I'd buy one and not bother with tinkering.

 

Whatever dual (or single CPU) TDP is IMO something to take into account from scratch when deciding on a cabinet as sufficient cooling is important. I acknowledge that a CPU likely dissipates way less than the advertized TDP (which is usually underestimated)during normal use (sans heavy DSD upsampling or active filtering etc).

If it were for me and if it would be possible I'd hardwire the whole thing from CPU to output without a MB or a circuiiot board in between, stick a couple of whopping cooling profiles to the CUP and put it all in a dampened and EMI treated wooden box rather than any metal. Alas, if it was that simple (or feasible)... I'll keep that for my tube amp(s).

 

in the extreme thread I read that having more than 10 cores total is of sonic benefit, I have little reason to doubt that. Do we know if it matters and how much it matters WHAT makes the total of 10 cores? dual CPU over a single, which CPU?

 

 

 

 

 

 

[Nenon]

1 hour ago, MarcelNL said:

If it were for me and if it would be possible I'd hardwire the whole thing from CPU to output without a MB or a circuiiot board in between, stick a couple of whopping cooling profiles to the CUP and put it all in a dampened and EMI treated wooden box rather than any metal. Alas, if it was that simple (or feasible)

You probably know this is not possible... But not because it's hard to solder tiny wires - that part, as difficult as it is, is actually possible. It's the high speed circuit and the interference there. We are talking about Gigahertz frequencies. Take a working motherboard and just move some of the PCB traces by a couple millimeters without changing anything to the circuit. Chances are it would stop working or become so unstable that it would be unusable. The interference in those circuits is HUGE. Hardwiring it is simply IMPOSSIBLE! This is not a tube amplifier that only needs to handle frequencies less than 100,000 KHz. 

[Nenon]

2 hours ago, MarcelNL said:

in the extreme thread I read that having more than 10 cores total is of sonic benefit, I have little reason to doubt that. Do we know if it matters and how much it matters WHAT makes the total of 10 cores? dual CPU over a single, which CPU?

 

I wish it was that simple to explain...

Every CPU sounds different and has its own character. 

The 10-core Intel Xeon Silver 4114 has the best and most transparent midrange I've heard from any CPU I've tried.

The 12-core Intel Xeon Silver 4214 is an overall step down. You would think that as a newer model with more cache and more cores, it just has to sound better. But it doesn't.

The AMD Ryzen 7 3700x has super nice highs.

The Intel i9s I have used have really nice body. 

And so on...

 

Then every motherboard sounds different too. Take a look at the beginning of this thread. I started with an Asrock motherboard and switched to Asus, because it sounded better to me with everything else being equal. 

 

How many cores is best - that depends on the OS, scheduling, the hardware architecture, the software you are using, the way you tweak the OS processes, the way you assign affinities, etc. etc. 

 

Dual CPU over single CPU. Well, that also depends on everything else. Running stock Euphony on a dual Xeon / Sage server did not sound much better than my AMD Ryzen 7 3700x with OCXO clock motherboard. I actually preferred the AMD back then. But the open and transparent midrange of the Intel Xeon Silver 4114 was apparent right away. But then:

16 hours ago, Nenon said:

When you align the logical motherboard diagram with the hardware placement and careful configuration of process affinities, the magic happens.

For example, I have

- The motherboard onboard NIC, all Windows processes, and HQP all assigned to one CPU.

- The Optane card for the OS is physically placed on a PCIe slot connected to the CPU above. 

- I have my USB card physically placed on a PCIe slot directly connected to the other CPU and assign NAA and any other audio related processes with affinities to that CPU.

It makes a difference! 

And the dual CPU was quite better. 

 

Combine all these variations of different motherboards, chipsets, different CPUs, number of cores, number of CPUs, CPU cache, OS, affinities, software player, etc. and you get so many combinations that one can spend an entire lifetime to figure out which combination works best. 

 

You want to put a meaningful number to each variable? That's also not possible. It's a moving target. You can make the same change on a computer using noisy power supply and the impact may be 5%. That same change on a computer with a transparent power supply could be 90%. And that's without even taking under consideration the rest of the system - your DAC, analog amplification, speakers, power, room, etc. If the power supply does not mask a lot of the music under noise, it can be your room that impacts it. 

 

After reading this, it may start to make more sense what we are trying to do with this DIY recipe. We want more people to participate in this, and the hope is that it would lead to some new discoveries. New discoveries is the main driving factor for Emile.

First, the high cost of parts and components would make this less appealing to people who don't have a quite resolving system already. Someone using a PS Audio Sprout 100 with a pair of affordable bookshelf speakers would likely have no interest in building a $10K server. And I would indeed advise them to spend their money elsewhere. The assumption is that those who are willing to spend $10K on a server have very resolving systems already.

Second, this is DIY. You would need to be able to put a computer together, install Windows, configure everything. That eliminates a huge group of people and leaves those who really enjoy tinkering with that stuff. Yes, some people may decide to build some services around this (i.e. build this for someone else for a fee), and that's perfectly fine. I've built computers for other people in the past and still have learned a lot from their feedback. 

Third, the power supply recommendation for this build would be as transparent as it gets. In fact, I've never heard more transparent source myself. 

That means we would have a relatively small group of people with quite resolving systems, building one of the most transparent sources that exists today as a baseline. We all enjoy tinkering with that stuff and as soon as we finish the build, we would be on a mission how to further improve it. And the real hope is that we would succeed and provide incremental improvements. That's the type of discoveries and feedback that would be extremely useful to the DIY community and to the whole industry. 

[MarcelNL]

53 minutes ago, Nenon said:

You probably know this is not possible... But not because it's hard to solder tiny wires - that part, as difficult as it is, is actually possible. It's the high speed circuit and the interference there. We are talking about Gigahertz frequencies. Take a working motherboard and just move some of the PCB traces by a couple millimeters without changing anything to the circuit. Chances are it would stop working or become so unstable that it would be unusable. The interference in those circuits is HUGE. Hardwiring it is simply IMPOSSIBLE! This is not a tube amplifier that only needs to handle frequencies less than 100,000 KHz. 

absolutely clear that hard wiring is no option....

[AME]

Hello,

 

I'm  following this thread with great interest, I'm fairly new on the forum, but hope to contribute more in the future. 

I'm also in the process of building a music server and almost have finished it, I allready started building, before I learned of the Taiko DIY project, otherwise I probably would have waited. 

 

Parts used:

- HDPlex case

- Asus Sage C621 single processor mainboard

- Intel Xeon Silver 10 Core

- 6 RDIMM memory modules with Samsung IC's

- Custom copper cooling for Xeon

- Intel P900 Optane 280Gb for OS

- M2 drives for music files on M2 PCI Express adapter

- Windows LTSC 2019 and Process Lasso

 

I choose to use a single processor, mainly because of the added extra cost for a dual Xeon system, the complexity with the passive cooling and the power supply requirements. The Power supply I'm planning to use can simply not handle two CPU's.

 

I choose to use a Xeon, because of the 48 PCI Lanes direct to CPU and the sound signature according to Emile from Taiko.

The Asus board has 4 PCI slots directly attached to the CPU.

 

One thing which was a challenge, was to find a proper passive cooling solution for the Xeon. I looked initially at the solution Nenon used for his build, but decided to take another route. So together with a friend of mine, who also wanted to build a new music server with the same components, we started to make our own cooling solution.

It all started with a copper block of a few KG. It was a lot of work, but the result is great.

 

I did a burn in test for a couple of days and the CPU never exceeded 49 degrees celsius. Under normal load it stays around 29 degrees celsius.

 

 

 

 

I have used a HDPlex 400W ATX PSU to test the server and it sounds great, a lot better then my former Intel 7700 build.

 

I also compared a SoTm USB Card with external clock from my former build with the motherboard onboard USB and couldn't believe that the onboard USB sounded better. Probably the new USB controller chip is a lot better.

 

Next step would be a very good power supply. I actually had a Farad Full ATX prototype PSU on loan, not cheap, but it sounded fantastic. I will write a seperate post on this.

 

I'm very interested in the Taiko DIY project and probably can use a lot of parts from my current build.

 

@Nenon any idea when the Taiko DIY components wil be available?

 

 

 

 

 

 

[Nenon]

3 minutes ago, AME said:

Hello,

 

I'm  following this thread with great interest, I'm fairly new on the forum, but hope to contribute more in the future. 

I'm also in the process of building a music server and almost have finished it, I allready started building, before I learned of the Taiko DIY project, otherwise I probably would have waited. 

 

Parts used:

- HDPlex case

- Asus Sage C621 single processor mainboard

- Intel Xeon Silver 10 Core

- 6 RDIMM memory modules with Samsung IC's

- Custom copper cooling for Xeon

- Intel P900 Optane 280Gb for OS

- M2 drives for music files on M2 PCI Express adapter

- Windows LTSC 2019 and Process Lasso

 

I choose to use a single processor, mainly because of the added extra cost for a dual Xeon system, the complexity with the passive cooling and the power supply requirements. The Power supply I'm planning to use can simply not handle two CPU's.

 

I choose to use a Xeon, because of the 48 PCI Lanes direct to CPU and the sound signature according to Emile from Taiko.

The Asus board has 4 PCI slots directly attached to the CPU.

 

One thing which was a challenge, was to find a proper passive cooling solution for the Xeon. I looked initially at the solution Nenon used for his build, but decided to take another route. So together with a friend of mine, who also wanted to build a new music server with the same components, we started to make our own cooling solution.

It all started with a copper block of a few KG. It was a lot of work, but the result is great.

 

I did a burn in test for a couple of days and the CPU never exceeded 49 degrees celsius. Under normal load it stays around 29 degrees celsius.

 

 

 

 

I have used a HDPlex 400W ATX PSU to test the server and it sounds great, a lot better then my former Intel 7700 build.

 

I also compared a SoTm USB Card with external clock from my former build with the motherboard onboard USB and couldn't believe that the onboard USB sounded better. Probably the new USB controller chip is a lot better.

 

Next step would be a very good power supply. I actually had a Farad Full ATX prototype PSU on loan, not cheap, but it sounded fantastic. I will write a seperate post on this.

 

I'm very interested in the Taiko DIY project and probably can use a lot of parts from my current build.

 

@Nenon any idea when the Taiko DIY components wil be available?

 

 

 

 

 

 

 

Nice work @AME. I am sure people here would be interested in your cooling solution. In fact Hdplex might be interested too. 

I am expecting the Taiko stuff to be available sometime in the second quarter of this year. 

My advice to you - hold off on the power supply decision!