Building a DIY Music Server

[seeteeyou]

truenas with new intel p5800x optane

https://forums.servethehome.com/index.php?threads/truenas-with-new-intel-p5800x-optane.32586/

 

That's posted in April and therefore someone did manage to purchase P5800X somewhere, and then these guys in Hamburg should be "claiming" to have a whopping 21 units of 800GB P5800X in stock

 

https://cgatesolutions.de/en/home/226-intel-ssd-p5800x-800go-25p-pci-e-.html

 

How and where on Earth did they find so many drives if that "magical" number were true? Even Intel couldn't provide any straight answers back in January

 

P5800X how to actually buy it?

https://community.intel.com/t5/Solid-State-Drives/P5800X-how-to-actually-buy-it/m-p/1245256

 

---

 

Processors that Support Intel® Optane™ Persistent Memory

https://www.intel.com/content/www/us/en/support/articles/000055996/memory-and-storage/data-center-persistent-memory.html

 

 

In addition to SSDs, there's also Optane DC Persistent Memory with MUCH better performance but they do require 2nd / 3rd Gen Xeon Scalable processors with specific motherboards

 

https://www.youtube.com/watch?v=mm31q64lif4

 

https://www.youtube.com/watch?v=BQXcuZj2Axw

 

https://www.youtube.com/watch?v=pnNR7GHCJoc

 

https://www.youtube.com/watch?v=ZeR8xHNq1gg

 

https://www.youtube.com/watch?v=UsQ-UAu6SBs

 

https://www.youtube.com/watch?v=dOV3gGncGU8

 

[Dev]

So the Taiko DC-ATX has been running continuously over a day and it has made a substantial improvement in my system. The level of realism has increased many folds.

 

 

My sole purpose of getting the Taiko was to power the CPU and based on what I have read so far, it seemed that it would really shine with high power devices but the motherboard was indeed an eye opener. Never thought powering the motherboard (Asus ROG) to be as critical as powering the CPU. I suspect that clean and low ripple noise power to the chipset, mainboard clocks and vrm is helping immensely. I had a similar experience when decoupling the CPU power with an exclusive 12V high current rail from the motherboard ATX rail.

 

The Asus motherboard consumes around 32watts and currently the Taiko is been powered with a  SJ dc3 diy 19v/5A rail. My hunch is the better your motherboard, the better it's outcome. I don't have a 16-48v high current rail to feed the Taiko and power the CPU alongside and hence can't wait to build a high current unregulated LPS - that would real exciting. Anyway, it will continue break-in and will report back if things change.

 

 

There are very few occasions when we buy stuff and feel that its totally worth the admission cost - this is probably one of them, at least in my system. A step towards the right direction.

 

 

 

 

 

 

 

 

 

[exeric2]

One thing I'd like to know, on motherboards that have separate EPS +12 volt and ATX +12 volt connections do you have to disconnect the +12 v connection at the output of the ATX power supply if you want to power the CPU separate from the motherboard? Sorry if that's a newb question. I guess I'm a newb.

[Dev]

8 minutes ago, exeric2 said:

One thing I'd like to know, on motherboards that have separate EPS +12 volt and ATX +12 volt connections do you have to disconnect the +12 v connection at the output of the ATX power supply if you want to power the CPU separate from the motherboard? Sorry if that's a newb question. I guess I'm a newb.

 

Depends on the motherboard design. In most of the Supermicro server motherboards and also Asus C621, you would have to - they are just wired in parallel. Most, if not all, gaming motherboards, they are independent rails for a reason and it would not boot if you don't connect the EPS since there is no power to CPU.

[exeric2]

7 minutes ago, Dev said:

 

Depends on the motherboard design. In most of the Supermicro motherboards and also Asus C621, you would have to - they are just wired in parallel. Most, if not all, gaming motherboards, they are independent rails for a reason and it would not boot if you don't connect the EPS since there is no power to CPU.

Thanks.  I have an Asus ROG motherboard so I guess that means I don't have to cut it.  Great. If not it would mean a lot of fiddling with connecting and disconnecting that ATX 12 volt power while experimenting with different schemes.

[MarcelNL]

@Dev Nice tube amp, 211 SET? 

[Dev]

2 minutes ago, MarcelNL said:

@Dev Nice tube amp, 211 SET? 

 

this one is 805. The 211 is the one I have been lurking around and would probably be my next SET :-)

[MarcelNL]

6 minutes ago, Dev said:

 

this one is 805. The 211 is the one I have been lurking around and would probably be my next SET :-)

 

 

aha, the top cap fell of the picture ;-)

 

Nice Graphite anodes! I'm into mesh tubes lately, with a monster build around a PL519 in G2 Triode ongoing for quite a while now.. 400H chokes etc...should be finished when the other monster project is done, my Trionor front loaded open baffle speakers,

 

 

[LTG2010]

Got my Taiko DC-ATX today here in Brexit UK after spending 2 days in Customs and arrived with a large rip to the box, what did they expect to find inside?

Big improvement in every sense over my HDPlex 800W, mostly clarity and lack of grain and distortion, warmer and more natural througout the frequency range. No brightness basically more transparent to the source after only a few hours warmup. Cymbals are cleaner, voices more natural, you can distinguish the differing deep bass notes.

Using it to power the dual sage C621 board, with an unregulated power supply, Mundorf Caps including silver in oil bypass caps. 2 chokes and 600 Watt torroidy audio grade @ 36V ( the metal cased supreme didnt fit my case).

I might change the transformer and rectifier bridge when @Nenon releases the specs but won't touch the caps, they sound good after 200 hours but mine took 9 months to lose a slight mechanical edge and let the music  flow.

Only niggle is lack of PCIE output so I just powered the 2 CPU inputs on the board, its a great upgrade.

[RickyV]

49 minutes ago, LTG2010 said:

Got my Taiko DC-ATX today here in Brexit UK after spending 2 days in Customs and arrived with a large rip to the box, what did they expect to find inside?

Big improvement in every sense over my HDPlex 800W, mostly clarity and lack of grain and distortion, warmer and more natural througout the frequency range. No brightness basically more transparent to the source after only a few hours warmup. Cymbals are cleaner, voices more natural, you can distinguish the differing deep bass notes.

Using it to power the dual sage C621 board, with an unregulated power supply, Mundorf Caps including silver in oil bypass caps. 2 chokes and 600 Watt torroidy audio grade @ 36V ( the metal cased supreme didnt fit my case).

I might change the transformer and rectifier bridge when @Nenon releases the specs but won't touch the caps, they sound good after 200 hours but mine took 9 months to lose a slight mechanical edge and let the music  flow.

Only niggle is lack of PCIE output so I just powered the 2 CPU inputs on the board, its a great upgrade.

Very Nice but we do want pictures. 😇

[Downtheline]

Also breaking in my taiko dc/atx into a single 4210 xeon with 621 chipset, power by 2 rail Paul Hynes SR7t. 

 

More relaxing. Sound stage is similar, but no longer feels artifactual. Sweetness or realness of the decay on notes. Reduction in glare, though there was little previously. 

 

Looking forward to making and  utilizing a new eps cable to replace my ghent cable that uses adapters. 

[LTG2010]

19 minutes ago, RickyV said:

Very Nice but we do want pictures. 😇

You had to ask Ricky :)

Sorry, its a messy work in progress, (which as usual will never get finished)

 

[Exocer]

I've been testing the 19V SR7T rail again to power the entire server (Today) and I wish I could redact some observations from my previous post. This sounds great

 

Vs 2 rails:

1. There is absolutely no harshness or sibilance (I have been listening for 1.5 hours now)

2. Dynamics are where they need to be

3. No thinness to report

4. Soundstage is about the same size

 

So for those who do not want to go with the unregulated LPS and already have a PH SR7T/DC4 LPS I highly recommend the Taiko DC-ATX to power both ATX and the CPU.

 

No, I have not switched out the EPS cable yet.

 

Some changes in my system since the first test:

 

1. CPU now runs at 4.5ghz (up from 3.8).

2. Tests were conducted with music files and OS on separate drives.

3. I've downgraded my Tidal subscription and listen mostly to Qobuz.

4. Perhaps the Taiko DC-ATX is appropriately broken in after several weeks of use.

5. Upgraded to a newer better sounding version of Euphony Stylus

 

I think i'll be keeping it this far for a while. The issue was elsewhere in my chain.

It'll be nice to have an extra 12v rail :)

 

Cheers,

-Rob

[Nenon]

Just a quick note to make here (still away on vacation with limited Internet time)...

 

What feeds the Taiko ATX needs to be a very low impedance supply. This is more important than regulated vs. non-regulated. Emile does not recommend longer leads than 4'' of thick gauge cable. I understand 4'' might be difficult but keep that in mind. Some of the DC cables we've used in the past (with great success) might be 50-100mOhm just for the cables. An impedance higher than 20mOhm could make the Taiko ATX sound lean. The unregulated LPS I am currently using has about 1.4mOhm output impedance (it also has about 5uV ripple). Something to keep in mind. I rarely say that, but in this case short and lower gauge wires work better than fancy 15-16 AWG wires (although having both is a good idea). With that in mind perhaps those powering with PH SR7T may try a thicker and shorter wire.

And there is of course the burn-in period. It takes a 7-10 days in my experience for the Taiko ATX to go through the peaks/valleys process. As @Exocer just reported, his finding have changed after some burn-in time.

@Exocer - try a thicker and shorter wire between the SR7T and the Taiko ATX if you can next. Pure silver wire like the one supplied by PH may not be the best match given how transparent the Taiko ATX is. Feel free to experiment... I am currently using copper wire (although I still need to try the Mundorf silver/gold wire).

Those of us who would be building the unregulated LPS inside the chassis from Taiko would be able to achieve very low output impedance. But there are ways we can do that with an external supply as well. 

[dctom]

Hi asking for some very basic help.

 

I have got my Taiko ATX however not sure how to connect the SATA power lead. My SATA lead has a 6 pin connector (as used on the HDPLX 800) but the Taiko SATA out has four pins?? 

[Nenon]

20 minutes ago, dctom said:

Hi asking for some very basic help.

 

I have got my Taiko ATX however not sure how to connect the SATA power lead. My SATA lead has a 6 pin connector (as used on the HDPLX 800) but the Taiko SATA out has four pins?? 

 

Sounds like you need a new cable. Here are some Molex connectors that should work with the 4-pin connector:

 

4 PIN – 5V / 12V
Part number: 39012040 (white)
Part number: 39012045 (white)
Part number: 39013042 (white)
Part number: 39013045 (black)

 

PINs
18 AWG Gold plated: 457501212
16 AWG Gold plated: 457503212

 

Here is what I have discovered for the other connectors:

6 PIN – External Power:
Part number: 39012060 (white)
Part number: 39012065 (white)
Part number: 39039062 (black)

 

8 PIN – EPS Connector:
Part number: 39012080 (white)
Part number: 39012085 (white)
Part number: 39013085 (black)

 

24 PIN – ATX Connector:
Part number: 39012240
Part number: 39012245
Black part that I think should work: 46992-2410
 

Note: The list above was not provided by Taiko. It was compiled by my own research. 

[dctom]

Thanks Nenon

 

I need a cable like this?

 

[Nenon]

51 minutes ago, dctom said:

Thanks Nenon

 

I need a cable like this?

 

 

I am not a fan of SATA drives, and haven't looked at them in a while, but generally speaking the SATA connector provides 3 voltages (3.3V, 5V, and 12V) and different drives use different voltages. Some spinning hard drives use the 12V for the spinning and 5V for electronics. Some solid state drives need only 5V, others need 3.3V, etc. You would need to check the specs of your drive and what voltages are needed.

And then you need to find a cable with the correct pinout or get one made for you. Here is the pinout on the Taiko ATX:

 

 

You can use any of these connectors:

Part number: 39012040 (white)
Part number: 39012045 (white)
Part number: 39013042 (white)
Part number: 39013045 (black)

 

With these pins:

18 AWG Gold plated: 457501212
16 AWG Gold plated: 457503212

 

On the other side, it would be a standard SATA connector.

You can ask the seller of the cable you showed on the picture for the connector type used and the pinout, and it if matches the above (and your SATA drive does not need 3.3V, which is very rare anyway), it should work.

Or you can ask Ghent to make you one.

 

Perhaps other people who use SATA drives would have some better ideas, but that's all I can offer. 

[dctom]

Thank you for all that information Nenon

 

I only use the SATA ssd to boot the OS into RAM and then switch the ssd out. It only requires 5v.

 

The HDplex 800w's SATA output has 6 pins but the two central ones seem to be superfluous.

 

 

[NewOldman]

I can strongly recommend this for those how are unable to make cables by them self

 

are made by multi strande 16 and 14 awg

 

 

https://www.corsair.com/ww/en/search/?q=Premium+Individually+Sleeved%3Arelevance%3AfirstLevelCategorySearch%3ACor_Products_Storage%3AfirstLevelCategorySearch%3ACor_Products_Accessories_Parts&text=Premium+Individually+Sleeved&pageSize=0#rotatingText

[Gavin1977]

I promised to provide some feedback on moving from Asus Gene XI to Asus Apex XI... took me a while to find an Apex XI as they're no longer in production, and I ended up importing!

 

Technical changes:

• iGPU on my 8086k is now disabled (I now use a Nvidia 1030, from the lowest 4x speed PCI slot, which goes through the z390 chipset as the performance of video is not a priority).
• M2 disk is now CPU direct (via Asus DIMM.2)
• Both my Solarflare network card and JCAT USB XE are now CPU direct in the top two bifurcated PCI slots (8x each)
• Apex XI has 1 extra phase in it's VRM.

Acoustical changes:

• Largest change is the size of the soundstage - obvious straight off the bat that it's improved.
• Lower noise floor (less hash) - I'm guessing this is primarily because of the disabled iGPU.
• Sounds more 'natural', there is no digital glare... to be fair, there wasn't with the Gene, but transients were softened on the Gene because I had to run my JCAT USB XE through the z390 chipset (this mATX board only has a single CPU direct PCI slot).

So if you can only do mATX then Gene XI is a solid choice, but this ATX motherboard is better because there is more opportunity to go CPU direct with PCI cards and M.2 disks.  So the theory stacks up - go CPU direct everything and try to avoid using motherboard chipsets.

 

No iGPU on Xeons, so must be a similar effect for 621 Sage users / SGM Extreme.

 

One year on, still waiting for my HDPlex 500w linear to arrive...

[OAudio]

14 hours ago, Nenon said:

The unregulated LPS I am currently using has about 1.4mOhm output impedance (it also has about 5uV ripple).

 

@Nenon 

 

Wow, 5uV = ~-135db of rejection ( assuming a ~30v rail ) ! 

 

I have been playing along in the background for a while using simulations of an unregulated supply based on the component choices in your earlier posts. With components listed and tweaking components parameters to near ideal performance levels its hard to get <3mv pk - pk @ 1 amp constant load.    

 

5uV is exceptional. Can I ask how you measuring the ripple and and under what test conditions ?

 

Thanks. 

[Nenon]

I had several questions about how I “voiced” the unregulated LPS, some of which with a hint of “should I trust you?”, “what is your sound preference?”.

 

In its simplest form a power supply is a voltage source that needs to remain stable regardless of the current draw. The higher the current draw and the more dynamic it is, the more difficult the job of the power supply is. I used the Asus SAGE motherboard here with two Intel Xeon Silver 4210 processors. These are 85W TDP processors, but the TDP specs are quite useless these days. They can draw more than 170W. Given that what I ended up with works really well on this dual CPU hardware, I suspect it would be a piece of cake for this power supply to handle a single CPU that is rated at 35-95W TDP, which is what a lot of people here use. You can think of it as an overbuilt (overkill) power supply. But who does not like overbuilt components in high-end audio :)?

 

It’s relatively easy to put together an unregulated LPS. You need a transformer, a rectifier, and optionally some components to smoothen out the DC / filter out some noise. Every part has its own sound signature.

When you compare two parts, in many cases one part is better in one area, and the other part is better in another area. But in some cases you have parts that are overall better in all areas. Typically, those are more expensive too.

 

There are two areas with the unregulated LPS - schematics and parts used. A few experiments with different schematics made it obvious what is important. Higher voltage sounded better. The lower the impedance, the better it sounded. There was benefit from better ripple rejection. That narrowed down the schematics selection to CLC and CLCLC. C is capacitor, L is inductor.

 

Inductor / choke selection:

For less than $30 it’s really hard to beat the 10mH / 5A Hammond 159ZJ inductor. It’s cheap, easy to implement, and sounds really good. Another option is to ask Lundahl to make us a custom choke, but Emile advised me that they are much more difficult to implement, so I scratched this idea for now. One of the problems when people like me (I.e. the OCD type) design something is that you can never put a border line and stop improving things… thus the design is never finished. I have promised to people here to provide an easy to build and great sounding unregulated LPS to go with the Taiko ATX. With that in mind, I decided to settle on the 159ZJ choke. It’s widely available, relatively inexpensive, and great for this application.

 

Transformer selection:

The next part to pick was the transformer. While transformers do sound different, the difference is not so much of a “voicing” tool. In my experience, well made toroidal transformers don’t sound brighter, leaner, deeper, etc. What I hear with different transformers is lower/higher noise and less/more distortion. I tried different voltages, different core sizes, and different brand models. Higher voltage was better sounding, so that was a no brainer. The VA value is an interesting one, because it interacts with the rectifier. Using a heavily oversized transformer caused ringing, which was an audible high frequency noise. A 750VA Toroidy Supreme transformer I tried definitely had that. I managed to reduce the ringing by using a very long 16 AWG wire between the transformer and the rectifier. But then what’s the point to use an oversized transformer and throttle its output by highly resistant wires. I found the 350-400VA to be the sweet spot. My first pick was the Toroidy Supreme. The guys from Toroidy are great. Their transformers are exceptionally good. And they are a very reliable vendor. But then Emile sent me a few more transformers to try. One of them was a custom Noratel transformer that is difficult to get. It felt like the Noratel was a little bit lower distortion. How much lower? If the Noratel is 100 points, I would put the Toroidy at 95+ points. Very little difference. Given that the Noratel is hard to obtain, and you have to place an order in the hundreds, I was going to go with the Toroidy. But then Emile offered to place an order with Noratel, and I gladly accepted that. But probably the main reason I picked the Noratel was that it has a built-in soft start. The Taiko ATX prototype I got had no protection and I fried one of the GaN FETs on the first day. Emile added protection on the production units. But I would rather be on the safe side still. Not only I picked the transformer with built-in soft start, but I also added an extra soft start before the transformer. This prolongs the life of the electrolytic capacitors, and the power supply does not need days to recover from the capacitors charge surge on startup.

 

Rectifier selection:

I have written a bit about this previously, but I tried different rectifiers with different transformers combinations. The most expensive Cree diodes I tried sounded the worst. The lower voltage drop Schottky diodes sounded pretty good. The active MOSFET rectifier Emile designed was on par with the best Schottky diodes. They sounded a little bit different, but I had a really hard time deciding which one I prefer. I’ve done many A/B comparisons and got lost at some point. Didn’t take notes, so don’t ask me to explain what I heard, because I don’t remember, but the difference was subtle. At the end I picked the active MOSFET rectifier for its efficiency and lower temperature. Keep the temperature lower in our music servers usually pays out in better sound quality. It’s also easier as we don’t need to to design a PCB board for the Schottky diodes.

 

Capacitors selection:

This is the part where you can apply the most personal preferences and be more creative and “voice” things to your liking. There are several aspects to consider - capacitor specs (capacitance, ESR, etc), capacitor brand, model, etc. Having had a lot of DIY experience with capacitors, I know how different capacitors sound. I have tried many - Nichicon, Panasonic, Mundorf, Audio Note, Elma, Kemet, Black gate to name a few. Without any doubt, Mundorf is my go to when it gets to high capacitance capacitors. I did try a few other capacitor brands I had on hand, but I quickly narrowed down the choice between Mundorf AG and Mundorf HC. The Mundorf AG had better / smoother highs. The Mundorf HC had deeper and better bass but was a little more rough in the higher frequencies. How do you get the best of both worlds? It’s not easy for the Mundorf AG caps in this application to get the deep bass of the Mundorf HC (although there are ways to get close). But there is a trick to significantly improve the highs on the Mundorf HC - using film bypass capacitors. After trying a dozen different film caps, I liked the VCAP ODAM as bypass caps the best. In my opinion the Mundorf HC with VCAP ODAM film caps sounds better in every level than the Mundorf AG in this particular application. That makes the unregulated LPS huge and more expensive, but that was my choice. It’s an overkill, and perhaps people with smaller speakers don’t need that. Mundorf AG caps would work perfectly fine here. But my highly efficient 18’’ paper cone woofers driven by dedicated SET mono blocks love the HC+ODAM combo. As far as capacitance, I have tried quite a few option and have settled on the two options I liked the best.

 

It’s still to be determined if we can use a full set of Mundorf HC capacitors with the Taiko chassis. I am not sure we can fit all that, especially in the CLCLC configuration, but once I get the chassis, I can do some quick experiments and pick perhaps a combination between Mundorf AG/HC/Film caps. There is also one more rectifier I have on order to test…

 

That’s all for now from me. More posts when my vacation is over... I know I have several other unanswered questions... will get to them eventually. 

[dctom]

Struggling to get the Taiko ATX up and running. I cannot get the system to boot up beyond bios screen, the  sata OS ssd and my nvme m.2 storage drives do not show up in bios screen. 

If I substitute my hdplex 800 it boots up with no problem??

[Nenon]

2 hours ago, OAudio said:

 

@Nenon 

 

Wow, 5uV = ~-135db of rejection ( assuming a ~30v rail ) ! 

 

I have been playing along in the background for a while using simulations of an unregulated supply based on the component choices in your earlier posts. With components listed and tweaking components parameters to near ideal performance levels its hard to get <3mv pk - pk @ 1 amp constant load.    

 

5uV is exceptional. Can I ask how you measuring the ripple and and under what test conditions ?

 

Thanks. 

Have you tried simulating the dual choke CLCLC configuration? 

 

19 minutes ago, dctom said:

Struggling to get the Taiko ATX up and running. I cannot get the system to boot up beyond bios screen, the  sata OS ssd and my nvme m.2 storage drives do not show up in bios screen. 

If I substitute my hdplex 800 it boots up with no problem??

Try booting from a USB stick (disconnect the SATA and NVME for the test). This should narrow down the problem between:

a. not having enough power to boot; or

b. something related to powering the SATA / NVME

If you can boot from a USB drive, it is likely "b". Sorry, I won't be around later to help troubleshooting, but other people here can advice.