Jump to content
IGNORED

Building a DIY Music Server


Recommended Posts

First, I just want to thank everyone that has contributed to this thread. I've learned a ton and I'm not even finished reading yet lol. 

 

Second, I wanted to solicit opinions to help finish my hybrid DIY taiko'ish build. The problem I'm running into is signal conversion from 1/10 gig LC single mode/multimode fiber to 3-pin AES/EBU XLR. Ethernet to AES is super easy, but fiber to AES is proving to be much more difficult. I would prefer to go the AES route as it traditionally carries far less noise and in almost every instance (save for the USB implementation on the legit Taiko Extreme), AES tends to sounds better to my ears than USB. 

 

Does anyone know of a good quality fiber to balanced 3-pin AES/EBU XLR converter or native PCIE to 3-pin AES/EBU XLR (not the 25 pin)? Pink Faun sells a PCIE "bridge" with AES/EBU, but I can't find any pictures to confirm it is the 3-pin version. I reached out to Jord at Pink Faun for clarification and I'm waiting to hear back. 

 

This is the only solution I could find, but it doesn't support more than 48kHz: https://www.transwan.com.cn/collections/balanced-audio-to-fiber-converter-line-level-audio-over-fiber-extender/products/4-channels-balanced-audio-to-fiber-converter-over-20-kilometres-single-mode-fiber?variant=31416469323838

 

My other options are to ditch the 10G intel fiber card and use my Adnacos instead (pcie fiber card to usb converter) or ditch my AES dream and go with USB.

 

Any help you folks could provide would be hugely appreciated.

Link to comment
1 hour ago, MarcelNL said:

RAM is random access memory, aka the sticks that provide volatile memory in your computer.

 

SsD is Solid State Disc, which is memory that does not suffer from Alzheimer....

Pseudo SLC is a mode in which you use another SSD storage format way less efficuent but with a sonic advantage

Understood. In that case, I can't confirm TLC can indeed be used in pseudo SLC mode. Also not sure about the penalties you speak of, but I haven't noticed any restrictions using my 16G RAM sticks in a 10/10/12 config (10G dedicated to RAM, 10G to ZRAM/a.k.a. ramroot/a.k.a swap, 12G free). 
 

I assume the fake Taiko (Faiko?!?) server I'm building will be more of the same. Obviously, 48G RAM will be supreme overkill, so I'll play around with the configs and see what sounds best. Given the direct PCIE bus access for both storage and boot drives, I do wonder if loading to RAM will actually sound better in this hardware configuration. 

Link to comment
  • 2 months later...
On 3/20/2023 at 4:07 PM, drjimwillie said:

The SK Hynix P31 2TB is selling for $119 on Amazon, that is about $100 off the regular selling price.
 

070EE4C5-5028-431E-A650-3A8C4D302CD8.thumb.png.c991aaa0fb8e94f4110d865b4f3d8980.png

 

I picked up two of these during a similar sale a while back. The main reason I did so is they have significantly lower power consumption compared to nearly every other similarly spec'd M.2 drive on the market. However, this may have been a poor choice as one of the key design aspects of the C621E build is not restricting power the same way an ultra low power build would.

 

Guess I'll find out soon enough. 

 

 

 

Link to comment
  • 1 month later...

Gents,

 

I finally finished building my DIY "Faiko" server based on all the outstanding recommendations and resources in this thread. Unfortunately, I seem to be having trouble with the CPUs, motherboard, or taiko DC-ATX. I can't seem to get both CPUs running at the same time. I've swapped cables, CPUs between MB slots, ATX ports, etc., etc. but no matter what I do I can't seem to get both CPUs working at the same time.
 

I've been through BIOS settings, but no luck thus far. I'm now combining through the MB documentation and the next step seems to be updating the firmware (if one is available). In any case, I thought it best to see if anyone has had similar issues. Relevant components are: 

 

1 x Asus 621E motherboard 

2 x Intel 4109T

1 x Taiko DC-ATX

12 x APACER 4G RAM (78.B1GN0.4000B)

 

Thanks, 

 

- lj

Link to comment

You guys are geniuses. I re-seated the CPUs again and it worked. I guess 5th times' the charm. 

 

Thanks so much for your help!

 

17 hours ago, ciccio1112 said:

HI. Check the perfect tightening of the bolts of the heatsinks that connect with the CPUs. It takes very little for all the contacts of the CPU not to touch correctly with the motherboard socket, and therefore not seen correctly.

 

10 hours ago, Nenon said:

 

+1 

This is a common issue with this CPU socket. 

 

Link to comment

I just acquired a license for Win 10 Ent IoT 2021 (thanks for the guide to obtaining a legit license through Avantech btw) to compare with Euphony 4 on my DIY Faiko. I wanted see if there are any proven recommendations for software optimization tools. I'm not asking for specific optimizations mind you, as I remember reading that Nenon & Co. will post a guide at some point in the future. I'm only asking for consensus recommendations for tools. I'm used to gutting and trimming Windows manually, which is a supreme PITA, so I'm all about investing in something that can make my life significantly easier (and hopefully survive Windows updates).

 

Based on my reading thus far, Process Lasso seems like the de facto standard and an absolute no-brainer for this use case. But I also see other options like Audiophile Optimizer (which may or may not support Win 10 IoT?!?), Fidelizer, XXHighEnd, etc.. Some of these seem additive to PL while others appear to be semi drop-in replacements. From an IT perspective, all of these add-ons (save for PL) make me nervous. Most seem like they are run by lone wolf devs that slap a GUI over a series of glorified PowerShell scripts, which makes me instantly wary of their ability to support said fancy pants scripts over term.

 

Does anyone use any of these, or other, tools with their DIY Taiko that can share any feedback or experiences?

 

Link to comment
6 hours ago, c-w said:

Yes, If you leave your Euphony machine running and you lost internet connection it will revert back to unregistered status after some time but you don't have to open ticket to "re-register" - just restart the application once you establish internet connection - it should find your registration online and set "registered" status again (in V4, V3 will need you to enter registration code again).

Only if your fingerprint changed you have to contact support to update it.

 

This was my understanding as well, but my hardware fingerprint somehow reset while I was away without anyone or anything touching my rig. I therefore concluded that it had to be the online requirement. If not, then this raises even more troubling questions as to what could have caused the hardware fingerprint to reset in the first place. It had to be some effectual combination of losing power, losing internet and/or time as the hardware never changed and no components failed. 

 

Regardless of the cause, the end result was the same - I couldn't use my server until I registered it again. I must say that Euphony support is/was excellent and very responsive. But in the end, my server and my ability to listen to music were held hostage by my operating system software. Euphony is excellent and I assume that my situation is an exceptionally rare occurrence. But hopefully the dev(s) can tweak the licensing enforcement a bit to ensure this sort of thing can no longer happen as it can cause tremendous frustration. 

Link to comment
2 hours ago, Exocer said:

I regret missing my opportunity to own a Taiko chassis, but I am holding out hope for when/if HDplex release a full-size Fanless chassis. Also, I may ultimately stick with two separate chassis long-term. It has served me very well.

 

 

The DIY chassis is high quality to be sure, but some of the tolerances were maddening. I never assumed I would have to mod a DIY chassis to get PCIE cards to fit properly.

 

Truth is, I went with an external power supply design and don't even need the Taiko DIY chassis. I only purchased one because I wanted to support Taiko supporting the DIY community. I guess the juice isn't worth the squeeze for them anymore, if it ever was in the first place.

 

Hit me up if you really want a Taiko DIY chassis. I wouldn't mind swapping it out with someone that will actually leverage all the space and design considerations it provides. 

Link to comment
2 hours ago, JJSim said:

 

Thanks for reporting your findings. It's long been a topic of debate over low core/power CPU's vs high core/power CPU's. Looking at the market of music servers available today, you can see the division between the two ideologies. The question always remains, however–how much is too much? Even a 12-core CPU is left twiddling its thumbs under a normal music playback workload (save for upsampling in HQPlayer). You can argue that even with extreme core counts and TDP, with such a small workload placed on the CPU it will draw less power and generate less heat than it's rated for. And having more cores allows for the workload to be distributed more evenly across the cores. But at what point is there diminishing returns?

 

There are also opponents and proponents of active cooling. On one hand active cooling can create vibrations and noise (both audible and electrical) that are detrimental to sound quality. On the other hand high temperatures can put greater stress on components, and keeping components at a lower temperature can improve sound quality. The best approach is a well implemented passive cooling system that is capable of dissipating the total amount of heat produced by the server. This is also the most difficult to develop. Taiko put much effort into developing the heatpipe system in the Extreme. But sometimes additional cooling is still needed. Taiko had to redesign their network card to include a copper heatsink as the cards were shutting down due to high temperatures. This buildup of heat inside the chassis can have a parasitic affect on the entire server. Active cooling, even in the form of an externally powered fan blowing air over the components, can have a positive result on sound quality.

 

Excellent info, thank you very much! I also hope I didn't inadvertently kick a hornets nest with the high power/vs low power argument. I only wanted to report what I personally tested as many of the results were not subtle. As a result, I can't help but stand firmly in the "more cores = more better" camp, but my sample size/e-peen is small and inexperienced.  As for the question of, "how much is too much?", increasing core counts seems to scale linearly with user insanity - haha! Seems like striving to answer the "why" behind core counts with lower cost components is more a sane approach than spending between 4-6k for the highest core count CPUs that will run under the maximum wattage supported by the C621e motherboard. I wish I could personally investigate this further, but unfortunately I lack the time, tools, talent, and inclination to do so. 

 

Specifically regarding power, you hit the nail on the head - it's all about real world utilization and desired outcomes for this specific use case. The highest core CPUs I tested were 20 core Xeon Gold 6148s, which are rated at 150W under constant load. My rig momentarily peaked at approx. 170W during startup, then settled in at approx. 112W during normal operation (approx. 105W when undervolted the CPUs to 1.47V). Taking into account Intel's notional CPU power ratings and my recorded start up draw, I set a theoretical 150W-per-CPU maximum for testing other CPUs specifically with this ULPS (officially rated max around 200W). Using said estimated theoretical formula, the highest core CPU currently available that could safely pair with this ULPS is the 26 core Xeon Platinum 8164. This CPU also happens to be rated at 150W, discontinued, cheap, and readily available. A have a pair on the way for testing as we speak. The startup draw seems to be the biggest hurdle to using higher core CPUs with this ULPS design and as you stated - it's anyone's guess if doing so will result in a tangible increase in performance.  

 

Cooling is the other fly in the ointment, as you stated. Running tests with active air cooling was a total PITA due to fan noise, though this did have the added benefit of cooling other components in the chassis. However, removing heat directly from the CPUs and quickly removing it from the chassis should help prevent components from overheating as opposed to dumping it back and locking it in with passive cooling. This is why I am considering using a 100% external water cooling solution. All pumps, controls, gauges, radiators, fans, etc. are contained within an external chassis, so coolant flowing across the CPUs should be the only noise introduced...which unfortunately still could be significant. I'm almost certain you are right about passive cooling being the optimal solution. But to support high core CPUs kicking off this much heat I would have to run heat pipes through the chassis directly to a giant external copper brick or design a different chassis cooling solution all together. This is why the Calyos solutions are so intriguing. I submitted an inquiry, though I'm almost certain it will be prohibitively expensive as you stated...and may or may not completely solve the problem in the first place.

 

Do you, or anyone else, have any other ideas?

Link to comment
15 minutes ago, BOBO said:

do you guys do any tweaks on RAM timings, or you just run on auto ?

At a minimum, manually set your RAM speed to match the recommended speed with your CPUs (i.e. 2400MHz, 2600MHz, etc.). For example, if you are using Xeon 4210s your RAM should be set to 2400Mhz. Also, depending on your RAM type and capability, you can try playing with (under) voltage, CAS latency, etc. but only if you know exactly what you are doing.

Link to comment
13 hours ago, flkin said:


Have you considered external, active cooling through the sucking action of a fan? Not touching the streamer nor electrical in the same circuit. 
 

I float a suction fan on the outside and above my streamer’s CPU and the hot air extracted is significant. 

 

This is an interesting idea...and no, I hadn't considered it! Fan noise is my concern with this approach, but that could be minimized with the right configuration. Then it becomes a question of how much air flow it will take to keep the CPUs at a reasonable temperature. My test setup is using four 120mm noctuas (two of which are industrial grade running at 2k rpm) and it is LOUD. 

 

12 hours ago, MarcelNL said:

you can design your own passive cooling system, but you'll have to bend your own heatpipes and route your heatsink....two large -ish 300W heatsinks should be perfectly capable of taking care of around 125W TDP CPU's

 

What TDP are you looking at? 

 

150W minimum. I would love to go for 195W to support max wattage for the MB, but that's a pipe dream as I would need to upgrade my power supply.

 

The problem is cooling, specifically the compounding heat effect as @JJSim mentioned. Passive systems (at least this one) dump heat back into the chassis and if the chassis can't dissipate it fast enough, it keeps building on itself. I did a couple of quick heat propagation tests this morning:

 

  • It took just over 4 minutes for two 150W CPUs to exceed 70C from room temperature (approx. 18C)
  • Approx. 17 minutes for one CPU to exceed 70C from room temperature. 

 

Unfortunately, the amount of passive cooling required to support these higher TDP CPUs likely exceeds both the available space and dissipation capabilities of in the Taiko DIY chassis. Best idea I can think of is replacing the top chassis panel with an enormous, finned copper slab. I assume the biggest challenges with this approach would be:

  • Ensuring proper heat transfer between the existing grooved side panel of the Taiko chassis and the proposed top panel copper heat sink. In theory, a handful of copper "L" brackets adhered (screwed?) to both the inside of the Taiko chassis and top panel should do the trick. 
    • It would make far more sense to land the CPU heat pipes directly on the top copper panel, but then taking it apart becomes a supreme PITA.
  • Pray this doesn't lock even more heat and cook literally everything inside chassis. Perhaps include some venting holes and a priest? 
  • Given the cost, I would likely be better off designing a completely new chassis all together. 

 

The more I think about it, the more an external water cooling solution makes sense. I just worry water flowing across the CPUs will generate unwanted vibrations. Perhaps it's the lesser of two evils as opposed to running everything at higher temps than they should be. FYI, this is likely what I would use: https://shop.alphacool.com/en/shop/water-cooling-sets/external-sets/11932-alphacool-es-orbiter-360-ts

 

Was just launched a couple of months ago. 

Link to comment
6 minutes ago, baconbrain said:

 

Perhaps cooling one CPU per heatsink / side of the Taiko chassis?

 

Indeed. That's why I tested a single CPU from ambient to max temp, but had to shut it down after it exceeded 70C after approx. 17 minutes. 

 

The cold truth is, I should ditch the high cores, keep using the 4210s, be happy and stfu. Otherwise, I'll have to use water cooling or a different chassis all together. The TDPs are just too high. 

 

Link to comment
  • 2 months later...
1 hour ago, baconbrain said:

Wow, thanks for the very detailed response!

 

How loud is the orbiter?

You are quite welcome!

 

The orbiter is very quiet. The pump and fans are barely audible unless you are overclocking, using HQPlayer, PGGB, etc. To be clear, the orbiter is way, way overkill for this use cases and that was by design. During normal operation w/undervolted CPUs, the pump and fans run at the very low end of their capability. This results in "near-passive" operation of the orbiter while minimizing vibration transfer to the server chassis as much as possible. 

 

In other words, I specifically chose to use a Porsche as a pace car traveling at 30kph and pray it doesn't blow up from underutilization - lol.  

 

For reference, I can't hear the orbiter over my air-cooled work PC during normal use. But if you are highly sensitive to noise, one of those freaks that listens to music in an anechoic chamber, or just don't want to risk hearing a single decibel...you could try replacing the stock fans with Noctuas (I think the orbiter uses "be quiet!" fans by default) or try disabling the fans all together (gven the high cooling efficiency of this loop design, very low pump utilization, and the thermal mass of the internal radiators, I suspect you could get away with disabling the fans...though I haven't tested it.). 

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...