streamlined server project .. part 2

Status
Not open for further replies.
Wondered if I could ask for a few opinions on the second phase of my Digital Music server project. In brief, the project is defined as "I already have lots of cds and lps but wanted to take the opportunity of finding a free desktop and building a lean, simple audiophile server".

Basically so that the most current-future topology in music collecting technology -- soundfiles -- doesn't leave me behind, with my miles of shelving for the physical collection, my phonostages and other oldschool gear.

Oh, and it should be nearly free most of the way. It has been so far. It should cost money only when absolutely necessary, and/or when some option or function is just too tempting, too worth the expenditure. Here's the earlier stages : http://www.diyaudio.com/forums/pc-based/263045-streamlined-server-project.html

At present, it's in an intermediate stage, where I have functional playback off the harddrive via foobar, a minimal set of drivers for vid/audio, and nothing else on the machine at all. Forget exact numbers but well below a gig in physical memory use during playback. Machine runs an Intel pent-4_2.2ghz board.

My question at this moment is that I ran across these clever-looking enclosures online Streacom | Computer Hardware & Accessories and just wondered--- don't flame me if it's the stupidest question ever asked here-- could the hardware in my tower move into a *fanless* enclosure ?

There are two fans at present in the cpu tower, and they're noticeably present during quiet playback. But things are running so minimally -- I'd expect well below the thermal averages of a machine that is doing the multi-tasking norm: busy hard drive and wlan, netbrowsing while doing email checks, an open word or excel program maybe, all at once ...

Contrary to all that, my server is offline, has no programs running other than system and foobar, and is extremely cool doing the pretty minimal task it's doing. The psu I think is a lam-stack linear type (I'll check on that). But almost certainly way over-capable of supporting a processor doing 16_44 wav-file conversion, so must be quite some distance away from capacity.

So I wonder what the caveats and risk might be of perhaps installing in an appropriate fanless enclosure... If necessary, it wouldn't be hard to 'outboard' the psu to another chassis, if that would help a lot. The hard drive here is expendable, too, so maybe that goes to a solidstate drive in the next enclosure ... if that would help. At present it's a plain vanilla ide western device drive. (Are there cooler-temp kinds of drives?) Or maybe the hard drive goes outboard ? I just don't know.

I presume the big heat-monsters are a) processor b) powersupply and less so c) primary drive.

Question then-- is any fanless enclosure going to be 'appropriate' here, or is the whole idea wrong..? Does 'outboarding' any one element, or all, start to be a game changer ?

(oh and fwiw, yes of course it's simpler to buy a new fanless chromebox, macbook, intel NUC, or something rather than doing this; but that's not the project. The project is already in motion.)

Thanks for any and all comments, (and feel free to say it may be delusional to take components that are fan-cooled and expect anything other than rapid overheat and fail... )
 
Last edited:
Why not consider some of the 'thin client' computers.. all fanless and have enough CPU power and RAM for stable work..
Just today I try Daphile on Fujitsu Siemens Futro S500, plays without problems hires and DSD files..
 

Attachments

  • Futro S500.jpg
    Futro S500.jpg
    7.1 KB · Views: 313
Doing some testing with the thermal-check program called Cpuid HWmonitor.

If I can compare normal temps at motherboard, psu and hd to temps while running foobar playback, I can make some decisions about 'fanless'.

Suppose if it looks encouraging, I might try unplugging the fan while watching the numbers change, too ... if only to see if the change is enormous or not.

Been reading up online too, and there are a few tips out there about possibly keeping the fan(s) but "five-volting" them, thus making twelve volt fans super quiet, if half as fast, but at least maintaining the thermal flow thru the machine.
 
A couple thoughts:

1. Primary drive. Doesn't really run very hot, but it is indeed spinning. Easy fix: solid state drive.

2. CPU fan. Desktop motherboards typically have desktop processors, and these can run hot. Your motherboard may be "underclock" able, which will reduce heat generated. Also, there are a variety of such fans available, some very much quieter, and very inexpensive. Your motherboard may also be able to adjust the speed of a cpu fan based on temperature of the cpu. Modern motherboards can do that.

3. Power supply fan. Modern low-noise power supplies also have variable speed fans, and you can get low-noise power supplies.

I built a modern, 8-core system for my wife to use a couple months ago. Solid state drive, low noise fans, etc. You can barely hear it in a dead-silent room with your ear on the case.
 
Some fans don't have enough torque to start up at 5 V (they might run just fine once they get up to speed).
Yes, good point. That's more or less true of the fans in my laptop riser platform-- they'll go really low on the rheostat dial, but must be kicked at the start with full voltage.

I didn't consider that, but now know why this switching scheme makes some sense :
Get 12V, 7V or 5V for your Fans | silentpcreview.com
Boot up at 12v and once in playback, and at-speed, switch down to the 7v or 5v.
 
A couple thoughts:
All good points, there.

This kind of reminds me of a project I was near enough to witness, but thankfully far enough away to avoid the mind-melt. In short, the idea was to soundproof (or 'blimp') a portable generator so that it could be rigged to the back of a bus to facilitate lighting in a motion picture context. It would need to be quieted so as to record sync sound. The ongoing fiasco for the project mgr was that every attempt to enclose the machine introduced aspects that made the heat rise. Every degree of heat rise impelled expanded intake and exhaust methods. Which increased overall sound levels.

Efforts to 'open up' the machine to improve thermal currents succeeded in laying bare the noise more effectively. Closing up, of course, quieted but increased heat.

Anyway, I can see some physical buffering might be applied to the HD without too much committing to keeping the change. Even perhaps just a vented double-enclosure built up from the existing single-wall enclosure might help there. Layer of dynamat in-between, maybe.

The PSU, yes, might just have to eventually be changed out to a fanless, and conceivably outboard model. That all makes too much sense.

The CPU I'll just have to do the therm testing to see what I'm actually up against. There are also massive cooler-heatsink things that gamers use that might just relieve the heat to usable levels. I'll report what I find with the testing.

One thing that I keep picturing is leaving the (side, typical) cover off the tower, situating it on it's side so that the opening is upward-facing, and laying on some kind of steel mesh in place of the cover. Therms would rise directly up from the processor, and perhaps with undervolted fan ....

Hmm, starting to sound a bit like my generator story above.....
 
Have run into a snag with Temperature Monitoring.

After failing to be able to monitor processor temp and power-supply temp with my usual Cpuid-HWMonitor, I tried another software and came to the same result. After some research, it turns out that Intel's Pentium 4 doesn't gather, distribute, or use temp-monitoring info-- in the same way as other processors.
This is a setback for the route to silent operation.

Can't think of how else to proceed with calculating whether sinks and external configuration for components would be feasible; even if I already knew, there's still no way to test it after that configuring is done...

Any thoughts about a workaround welcome...
 
My question at this moment is that I ran across these clever-looking enclosures online Streacom | Computer Hardware & Accessories and just wondered--- don't flame me if it's the stupidest question ever asked here-- could the hardware in my tower move into a *fanless* enclosure ?
Maybe, but not one of those, I don't think. There have been reviews on them, and they seem to rely on the CPUs throttling to handle the models they claim.

There are two fans at present in the cpu tower, and they're noticeably present during quiet playback. But things are running so minimally -- I'd expect well below the thermal averages of a machine that is doing the multi-tasking norm: busy hard drive and wlan, netbrowsing while doing email checks, an open word or excel program maybe, all at once ...
Today, if you don't add a video card, you can use a bit less power actually using the PC. Pentium 4 CPUs were hot back then (though a 2.2GHz isn't too bad), and new CPUs will idle under 10 Watts, making for a typical 20-50W load, if not stressed heavily (very dependent on parts other than CPU, for how low it can go when idle).

I presume the big heat-monsters are a) processor b) powersupply and less so c) primary drive.
CPU, CPU voltage regulation, and that's it, if you buy relatively new. Drives get hot in terms of temperature, but use little power. Power supply efficiency varies.

Question then-- is any fanless enclosure going to be 'appropriate' here, or is the whole idea wrong..? Does 'outboarding' any one element, or all, start to be a game changer ?
The whole idea is wrong, just because it's difficult to do well. Most motherboards expect more than just passive air flow to cool the parts on it, and the space taken up by heatpipe coolers that can run passively is enormous. Big very slow fans, behind a cover, so you do not have direct line of sound to them, will do the job very well, to the point of inaudibility. But, 120mm and 140mm fans are not found on most prebuilts, and are relatively new, in terms of being common, with DIY.

(oh and fwiw, yes of course it's simpler to buy a new fanless chromebox, macbook, intel NUC, or something rather than doing this; but that's not the project. The project is already in motion.)
Buying new, but not a NUC, would be best, and easiest, or at least not too old. Since NUCs tend to have fans, which are tiny, and so become quite audible, I don't see them as appropriate.

What you could do is this: a new system, with a relatively big quiet-oriented heatsink (given the low power used, there are many to choose from: basically any without very tight fin spacing), and low power processor, then put it in a MicroATX that takes 120mm or 140mm fans, and isn't too flimsy. Get nice quiet fans (Noctua NF-12P-PWM are a 'safe' option), and low vibration HDDs (soft-mount them nonetheless).

Or, if you have a NAS somewhere else, get a tiny book PC, that only handles the playback, and has no space for the HDDs. New ones can be had for $200-300, that are better in every way but small size to a NUC. Make sure to check if it has a fan or not, or build one so you know it won't.

Why?
Power supplies: we now have affordable semi-passive power supplies. Buy one that's rated for 550W for a 50W computer, and it's fanless. Or, you can get a low-power setup and use a PicoPSU, and be fanless, without paying a mint for a fanless high-power PSU.
CPUs: AMD and Intel both offer CPUs with maximum draws under 40W, and the Atom and Jaguar based models even less (many under 20W).
Video: new integrated video uses low power, offers good analog signals, digital display options, and doesn't eat into CPU performance.
HDDs: you pay a bit of a premium, but WD and Seagate both make 5400 RPM "NAS Drives" that are low-vibration. They still need soft mounting, but are very quiet. That said, a NAS is another room totally takes care of that issue (keep in mind that an always-on desktop PC with a shared directory is as much as a NAS as some Synology box).
Cooling: using new sockets and CPUs means being able to use new heatsinks, many of which can run lower-power CPUs without their own fan being added, and without needing ducting to make that happen. With a low-power CPU and integrated video, there's no reason that a small case cannot be cooled well by a sub-500 RPM 120mm fan (actually, at idle, my gaming rig, with a Xeon E3 and GTX 970, can be cooled by a single 120mm at 180 RPM, so I'm being very conservative, with the 500 RPM figure!). With a low power CPU to start with, a quiet fan running at the lowest RPM it will stay on should be plenty (one of several reasons to use PWM fans only). New integrated motherboard/CPU combos without heatsinks, using low power CPUs, can be used passively, if no other parts are problematic, too, making it even easier.

Finally, parts that old increase the chances of failures from age, some of which can take other components with them.

Retrofitting old parts is going to be harder than doing it new. I mean, fine for getting the hang of all this software based audio management, but for a final product, where you might risk hearing it, no. The last decade has seen a huge move towards greater efficiency, lower power, and quieter cooling.

With careful parts selection, you can make inaudibly quiet powerful new PCs. With performance and local storage requirements reduced, like for this sort of appliance, you can do it with less and less care in parts selection, and often at a lower and lower budget. With the right parts, you won't even need a cooling fan for the case.

That ASRock J2900, for example, if it's coil whine free (Jetway's aren't, IME), in a nice MiniITX case, with a PicoPSU, and small SSD, should work great for a totally passive system, with higher performance than your current one (maybe not by more than 2x, but still faster). Even buying a Windows license for it, you will probably not spend more than $450, and might be able to squeeze it in $350, depending on what is available.
 
Last edited:
ASrock Q2900 $111
4GB DDR3L: $35
HTPC case, Silverstone ML06B: $55
OS drive: 64GB U110 (about the minimum for a comfortable Windows installation, and it's a good drive, if not the fastest): $55
PicoPSU 80W + 60W brick + bracket: $36
Random metal plate to cover ATX PSU hole: a little scrounging, cutting, and drill work 🙂.
Windows 8.1 full: $122 ($102 for OEM 64-bit, $0 if you go Linux)
It comes to around $420. Zero fans. Other cases might be better for using internal HDDs, rather than streaming over a network, and could add $30-50 to the cost, depending on how much aesthetics matter (choosing that $55 case was due to a nice looking bezel, not technical needs).

Sure, that's a bit of money, but it's hardly outrageous, and depending on case, can offer a lot of flexibility (larger cases waste space, but may allow full-height cards, and multiple soft-mounted 3.5" HDDs). Once you are comfortable with your software and storage setup, just go new, and not prebuilt.

Those Atoms should be 1.5-3x as fast as your P4, per core, depending on what you might do that brings the P4 to 100% CPU, along with giving you 3 more CPU cores. The fanless small boards like the one selected also tend to be made to be able to run truly fanless. That is, all the power hogs are under the heatsink, and use little enough power that the heatsink alone will do, as long as the place it is installed in is well ventilated.
 
Last edited:
@bloodfromastone Thanks. Lots of great ideas there. You may have missed the first part of this project, though. The frame around this is "got a free, mint, 2003 winxp machine, have a hundred wav files, might be fun to make a digital server-- a minimal one, streamlined and barebones, that costs a nice, minimal, next-to-nothing price." For the experience of diying it, hacking it, basically, out of what's on hand.
Retrofitting old parts is going to be harder than doing it new. I mean, fine for getting the hang of all this software based audio management, but for a final product, where you might risk hearing it, no. The last decade has seen a huge move towards greater efficiency, lower power, and quieter cooling..
As I've mentioned in the beginning of this, I'm at no loss for yet another source in my system. I've got a range of competent music sources around here; this is, indeed, an occasion of "getting the hang of all this software based audio management". I'm not after giving Sooloos or anyone else a run for the trophy. (And not at any risk of same, either!) A "final product" this most certainly isn't, even if it turns out to be very functional in the end.

Your post could well be a good start on the next chapter, or more like it, the next volume --- having done it once to see how things fit together. For now, it's just a matter of patiently adapting a full-service desktop into a minimal Server, but using the absolute minimum to do it.
 
Separate temperature sensor (i2c or 1-wire)?
Yes, it's looking like this is the right track. The sense I get from reading about what Intel did with their Pent4 is that rather than sending data with numerical value to a central collection point --- (bear with me here, & do recall I have no expertise here at all--) but that they made certain conditions swing the speed of the processor downwards if too hot. Basically without any intervening monitoring station, or at least without one that is very accessible.
" . . . Intel implemented a new technology, Thermal Control Circuit, in its Pentium 4 CPUs. This technology is intended to ensure stable work and protect the CPU against overheating. Every Pentium 4 has two built-in thermal diodes. One thermal diode reports the CPU temperature to the hardware monitoring system of the mainboard. The other is placed in the warmest spot of the die, next to ALU units and is a part of Thermal Monitor circuit.

AMD Athlon XP cpus also feature a similar thermal diode, but the two processors do differ a lot here. The thermal diode of Athlon XP reports the CPU temperature to the mainboard. A special logic unit of the mainboard processes the received data and shuts down the PC when the temperature exceeds a certain critical value. Of course, all the unsaved data will be lost in this case. Thermal monitoring in Pentium 4 is based on a different principle: the system should work stable even when the CPU temperature notched the critical value. The system should only shut down in case of emergency. This means that the CPU must prevent itself from heating up further and continue providing stable work of all applications." ..from: X-bit labs Investigation: Influence of Intel Pentium 4 Core Temperature on CPU Performance - X-bit labs

So if I get that right, an internal failsafe, and not a set of data that is accessible. Doesn't matter really, if the standard Temp Monitor softwares cannot--- and if you look, they do add a pent4 disclaimer--- cannot pull up temperatures to show the user.

Thus yeah, think you're right-- it looks not like software to tap any internal data, but an add-on piece of *hardware* that can simply transfer data about processor, (or more likely processor heatsink, since that would have to remain in place for initial testing)-- at fan design-speed, at lower fan speed(s), and perhaps at no-fan. And something to monitor alternate heatsinking and cooling strategies, once decided upon.

Same sensor could conceivably monitor Psu temps, just by relocating the hotspot, or wand, or whatever it is. Well that all sounds like something. Off to the research now.
 
@bloodfromastone Thanks. Lots of great ideas there. You may have missed the first part of this project, though. The frame around this is "got a free, mint, 2003 winxp machine, have a hundred wav files, might be fun to make a digital server-- a minimal one, streamlined and barebones, that costs a nice, minimal, next-to-nothing price." For the experience of diying it, hacking it, basically, out of what's on hand.
...
Your post could well be a good start on the next chapter, or more like it, the next volume --- having done it once to see how things fit together. For now, it's just a matter of patiently adapting a full-service desktop into a minimal Server, but using the absolute minimum to do it.

I don't think it's worthwhile: you have an old desktop, and are trying to make a silent audio server out of it. Can be done, but not cheap... Passive cooling costs a lot.

If you're onto cheap solutions, just go for a second-hand (cheap) thin client: a 600 MHz machine with 256MB of RAM and possibly a 128MB CF card will do wonders: a completely silent (no moveable parts) server, that can spit out 24/192 hi-res audio (depending on the USB DAC used). I've been building such servers for the past two years, and can attest that such hardware is absolutely the most trouble-free solution you can get (not to mention the cheapest - bought second-hand, it can be had for less than a Raspberry Pi)... I'm using various thin client HW, like Igel, HP, Neoware, etc.. - mostly with VIA motherboards. For more info on such hardware, this is a good site:
Repurposing Thin Clients

The only thing is, you can't put a standard distro (Windows/Linux) on it: you'll need something very small: I'm using mpdPup, a Puppy Linux distro, which I tweaked for my purposes - it runs a MPD-based audio server (up to 24/192) from a 128MB CF card and 256MB RAM. For hi-rez (24-96 to 24/192), 512MB RAM is recommended, though.

If you're onto something with more "horsepower", look for a fanless Atom board... A fanless solution can also be had relatively cheap second-hand...

Just as an illustration, here's some stuff I've built in the past 2 years (that's my commercial site, sorry...):
HANA LAB - Product Gallery
 
@den-hr
Fact is, I'm using 1oo% free hardware to learn about things. Nothing wrong with that. And as stated, I'm not setting my hopes that my current project here will be anything but intermediate. It's a chance to learn by doing, learn by mistakes, learn without endangering or losing any expensive gear in the process.

I came into this not really knowing too much; about various OS, about mounting and configuring a new hard-drive, about the general physical architecture of the standard Windows Computer. I still have things to learn about cpu usage and temperatures, controlling heat, and also about the merits of various server programs and the meta-tag capabilities of same.

With everything you learn, it seems, you get a double lesson-- first, the 'unavoidable reality' kinds of lessons, things you must do, and second, having had the overview, an understanding of the things that are available but optional. Things that others consider worthwhile that you may wish to customize into your project (or out).

If I didn't care about expense, I'd probably just go to a Meridian or similar hardware, and maybe the JRiver software, and call it finished.

But I do care about expense, especially since I knew I was starting fairly uninformed; I've learned with tube-amps, analog-Lp front ends, speaker-building-- that you will learn an unbelievable amount on your first Diy project, and continue to pick up knowledge as you proceed. No reason I can think of that this would be different.

Thanks for the Thin Clients link. I'll read up every word there. It does seem a natural solution for this kind of thing. For some reason I think it will become obvious when the current {desktop} project has run its course, and can teach me no more. For now, though, the price of the 'schooling' is exactly right.

(*one reason I went this route too is that I didn't want to learn three involved new tasks--- HD configuration/OS-installation, new Server program, and new OS (like Linux)-- all at once. Being already comfortable with Foobar, and windows OS, kept two of those things from being an all-new challenge. All I had to do was install the HD and already-familiar XP. Next time out I'd be more brave about that, having gotten a little bit ahead in baby steps.)

Thanks for the computer audio site of your products, too. Looks like you've hit on a good concept there.
 
Status
Not open for further replies.