APOLLO (2 CPU LGA1366 Server | InWin PP689 | 24 Disks Capacity) - by alpenwasser [COMPL. 2014-MAY-10]

[alpenwasser]

Nice, the mill looks new, is it? i'm on the lookout for one but thinking of building myself a CNC...

Nah, he's had it for a few years now, but he does take very good care of it

and it doesn't get that much use, so indeed it does still look in great shape

(even in person ). It's not a CNC though, still operated by hand. But it is

very well built.

[Vitalius]

 

Tought about it, but no. Couldn't find a solution what would not have cost me

tons of money. Lamptron makes some nice rubber washer mounts, but those cost like

5 USD per HDD, which makes >100 USD for all drives. Nope.

Not even eBay was helpful in this unfortunately.

Let's be honest though: Before a few years ago, none of our drives were ever

mounted with dampening and things seemed to work fine for many decades without

it, so it should be OK. Noise isn't really a big issue in this machine. The only

real concern is the HDDs getting damaged from the vibrations. If I suddenly

start losing disks left and right I'll know why, RMA them and make some modifications.

The important data is backed up anyway, and most of the drives are new-ish, so

that would still be a manageable scenario.

So, for the time being I'm just going to see what happens and keep an eye

on things.

 

Well, she was more of a motivational coach than an actual help.

Just a thought regarding this. Seagate had some thing where their drives were mounted opposite to each other. Like, the first drive is right side up, the next drive is upside down, rinse repeat until out of drives. This way the vibrations cancel each other out since the platters are spinning in opposite directions of each other. I'm not sure how effective it actually is, but it sounds like it could work. And it seems like you could try that without making any modifications to your rack. i.e. just put each 2nd drive in upside down.

The metal part of the HDDs might be much closer, but the PCB side will have more room as a consequence which is the side that gets the hottest from my experience, so more open air may be a good thing here. 

Just something you might try.

 

Nah, he's had it for a few years now, but he does take very good care of it

and it doesn't get that much use, so indeed it does still look in great shape

(even in person ). It's not a CNC though, still operated by hand. But it is

very well built.

Nice.

Not related to the above posts:

Alpen, please put the images in spoilers. When I went to a recent post on the page before this, I would start reading, then all the images started loading in which push the page down and change what I'm looking at. I can't tell you how annoying that is. D:<

[alpenwasser]

such upside down

 

Hm, that actually sounds rather interesting. Airflow wouldn't really be much

of a problem, I have pretty powerful ventilation for the drives, but it could

be that they don't actually fit into the racks like this.

I shall try it out, thanks for the tip!

 

Not related to the above posts:

Alpen, please put the images in spoilers. When I went to a recent post on the page before this, I would start reading, then all the images started loading in which push the page down and change what I'm looking at. I can't tell you how annoying that is. D:<

NONONO! We hates spoilers, we hates it! :angry:

Spoilers are ugly and we are too lazy to click, scrollwheel is less work.

(but yeah, I know the jumping around thing is annoying, I just hate spoilers more

than I hate that, sry)

[MG2R]

 

/thread.

[NerdNinja]

Hey @alpenwasser Maximum PC just shared this build on facebook! Congrats to the publicity man!

[alpenwasser]

Hey @alpenwasser Maximum PC just shared this build on facebook! Congrats to the publicity man!

Haha, awesome! I was wondering why my server was suddenly getting bombarded.

Might need to do some optimization for the dynamic image resizing then, seems

the script uses a bit much CPU and RAM as it is now. Ooopsie.

Thanks for the notification, much appreciated!

EDIT:

I noticed that the dynamic image resizing puts a lot of stress on the server's

CPU and it sends out errors on occasion. I've reduced the source image files'

resolution quite a bit to alleviate the issue, and will be looking for a more

permanent solution in the coming few days. Need to do some script-fu it would

seem.

Sorry for the inconvenience, and thanks for stopping by.

EDIT2:

Oh, they also shared it on Twitter! https://twitter.com/maximumpc/status/451821285924827136/photo/1

[jscho]

 Hm, that actually sounds rather interesting. Airflow wouldn't really be much

of a problem, I have pretty powerful ventilation for the drives, but it could

be that they don't actually fit into the racks like this.

I shall try it out, thanks for the tip!

 

I was actually thinking about this the other day, because someone in a YouTube video's comments section was complaining that the uploader was using WD Greens in a NAS and that the vibrations would kill them as they aren't NAS grade drives (he had 6 WD Greens in a Fractal Node 304) but the way the HDDs mount in the Node 304's cages is that they are mounted opposite from each other, so vibrations would be 'going the other way' (for lack of a more sophisticated term, centrifugal vibration I believe (?), can't remember)

Mounted like so:

 

As @Vitalius pointed out, mounting the HDDs like that could help with the vibration!

[MG2R]

upside-down-ness

I'm having trouble believing that this would alleviate the vibration problem. If both drives would have the same vibrational pattern and if they would somehow be synced up, then yes, it would cancel out the vibrations. But, that's two if statements that aren't easily satisfied. The vibrational pattern on both drives will be different, as the inaccuracies in both drives are different (statistically the same, though). The drives don't run in sync, so it would be absolute luck if the patterns (even when they're the same) match up and have a 180 degrees phase shift. Theoretically, you would have just as much chance of cancelling the vibrations by mounting them all in the right way.

 

At least, that would be my insight on the matter. Please do share any other views.

[alpenwasser]

I was actually thinking about this the other day, because someone in a YouTube video's comments section was complaining that the uploader was using WD Greens in a NAS and that the vibrations would kill them as they aren't NAS grade drives (he had 6 WD Greens in a Fractal Node 304) but the way the HDDs mount in the Node 304's cages is that they are mounted opposite from each other, so vibrations would be 'going the other way' (for lack of a more sophisticated term, centrifugal vibration I believe (?), can't remember)

Mounted like so:

 

As @Vitalius pointed out, mounting the HDDs like that could help with the vibration!

 

 

I'm having trouble believing that this would alleviate the vibration problem. If both drives would have the same vibrational pattern and if they would somehow be synced up, then yes, it would cancel out the vibrations. But, that's two if statements that aren't easily satisfied. The vibrational pattern on both drives will be different, as the inaccuracies in both drives are different (statistically the same, though). The drives don't run in sync, so it would be absolute luck if the patterns (even when they're the same) match up and have a 180 degrees phase shift. Theoretically, you would have just as much chance of cancelling the vibrations by mounting them all in the right way.

 

At least, that would be my insight on the matter. Please do share any other views.

Yup, having thought about it a bit, I'm inclined to agree with ^this. Basically

what you get, no matter how the drives are mounted, is the superposition of several

vibrational patterns. It would be an insane amount of luck if they cancelled each

other out. Just about as likely as them being in resonance and reinforcing each

other, which would be very bad and which I haven't heard happenening in practice

(at least not yet). Seeing as the bad case doesn't really seem to be a problem, I

don't think the optimal case is very likely either. At least if I've understood

my wave function theory correctly.

As for the Node 304: The reason I'd say vibration is a non-issue is because the

drives are mounted on nice fat rubber mounts anyway.

But thanks for the help guys, much appreciated.

[jscho]

-snip-

 

Ah touche...   Would it be possible to make some DIY rubber mounts? Would that be too far of a stretch? Would be incredibly cool though!  :lol:

Knowing you, the almighty Alpenwasser, you'll find an excellent solution. We have faith in you! 

[alpenwasser]

Ah touche...   Would it be possible to make some DIY rubber mounts? Would that be too far of a stretch? Would be incredibly cool though!  :lol:

Knowing you, the almighty Alpenwasser, you'll find an excellent solution. We have faith in you! 

Thought about that too, but then you can't really slide the disks in and out anymore,

since rubber is sticky. Plus the rails and recesses are all designed to fit the screws

now, it'll be a lot of work to change it. TBH I'll just run them on the screws for now,

see what happens, hoping I'm not tempting fate too much.

If you suddenly hear me screaming about all my disks dieing you'll know what happened.

But thanks for the vote of confidence.

[Vitalius]

Thought about that too, but then you can't really slide the disks in and out anymore,

since rubber is sticky. Plus the rails and recesses are all designed to fit the screws

now, it'll be a lot of work to change it. TBH I'll just run them on the screws for now,

see what happens, hoping I'm not tempting fate too much.

If you suddenly hear me screaming about all my disks dieing you'll know what happened.

But thanks for the vote of confidence.

Idea: Use U Channel on the recesses. It's rubber isn't it? Then it doesn't have to come out. 

i.e. cut a really short piece of U Channel, sit a single bead of glue on the side wrapping around the metal (not too much), then put it in the recess and set an HDD on top of it to hold it down until the glue dries (fans being on should help this).

Ghetto, but it would server it's purpose I believe.

[alpenwasser]

Idea: Use U Channel on the recesses. It's rubber isn't it? Then it doesn't have to come out. 

i.e. cut a really short piece of U Channel, sit a single bead of glue on the side wrapping around the metal (not too much), then put it in the recess and set an HDD on top of it to hold it down until the glue dries (fans being on should help this).

Ghetto, but it would server it's purpose I believe.

Thought of that too at some point, and that's probably the solution I'll be

going with if I actually start having problems. Trouble is, I'd need to disassemble

everything and drill out the recesses. They are precisely machined to the screw

diameter so that the drives lock into place properly (my neighbour actually did

an awesome job on those, they all fit pretty damn near perfectly, not an easy task

to do I tell you).

[flibberdipper]

server it's purpose

I see what you did there.

[Vitalius]

I see what you did there.

I... wha...

I typoed that. Being completely serious. That wasn't intentional at all and I didn't notice it until you said something. Holy carp that is awesome.

[alpenwasser]

I see what you did there.

 

Bloody hell, good catch!

 

I... wha...

I typoed that. Being completely serious. That wasn't intentional at all and I didn't notice it until you said something. Holy carp that is awesome.

Now I'm not sure if you're doing it on purpose or just keep mistyping.

Either way, I am entertained.

Since I'm here anyway: I've done some script-doctoring and the server load

issues might be solved, or at least alleviated. I can't really test it properly

unfortunately. But we shall see.

[alpenwasser]

Chipset Cooling, Adventures in Instability
 

 

I'm trying  something new  with the  formatting  which  requires a

monospace font, let me know if you have issues with it please.

As some may be aware, I  originally had some issues when trying to
get this machine to run  stable. While stress testing with mprime,
it repeatedly and  reproduceably crashed after less  than an hour,
sometimes  even  already  after  a few  minutes. Each  time  after
crashing, it took  me several tries and about 10  to 20 minutes to
get the board to POST again.

After  some  troubleshooting and  running  a  few diagnostics,  it
turned  out that  the 5520  chipset was  running really  hot. It's
temperature threshold  as indicated  by the  system is  95 degrees
Celsius, and when I  was last able to check on  it before a crash,
it had already passed 85 deg C, so I suspected that it was bumping
up against  the threshold, upon  which the board did  an emergency
shutoff and mandated a cooldown period until it would run again.

As an emergency fix,  I took the 80 mm San Ace  fan that came with
the case  and mounted it to  the chipset heatsink with  some waxed
cotton cord, and  voilà somewhere slightly above 70  deg C maximum.


Unfortunately  I forgot  to take  pictures of  that rather  ghetto
setup before dismantling it again  and replacing it with something
more  solid, but  I have  managed to  blow up  some sections  from
another picture  that should at least  give you an idea  of how it
looked.


Some Improvisation

Apologies for the horrid picture quality, as said this is a blowup
from a picture of which this section is only a small part.
(click image for full res)



A More Permanent Solution

The chipset heatsink is just your run of the mill alu heatsink held
on by a spring clamp with some hooks.

(click image for full res)



And the naked chipset after cleaning off the TIM. That stuff was a
bitch to get off, it had dried up rather significantly.

(click image for full res)



Since  the 80  mm  fan is  quite  a bit  larger  than the  chipset
heatsink itself, I needed to either replace the heatsink or modify
it  in  order  to be  able  to  mount  the  fan to  it. I  took  a
rather crude,  but very effective  approach: I took an L  piece of
aluminium,  drilled two  holes across  the heatsink,  cut some  M4
threads on  those two holes  (which worked despite the  holes only
going through the fins and  not being continuous), then bolted the
L piece to the heatsink with two M4 screws. Works like a charm.

Don't mind the  unclean alu bits from the drilling  and cutting on
the  heatsink  between the  fins,  it  wasn't really  possible  to
properly clean that off and make the holes as clean as one usually
does.

(click image for full res)



And from the other side...

(click image for full res)



The fan itself is held down by three screws, two in the L piece...

(click image for full res)



... and  one in the corner  of the heatsink itself. The  bent fins
are  from  drilling  and  cutting  the  thread,  they  got  a  bit
structurally  weak  at their  edges  due  to that. Doesn't  impair
functionality, so  not such a big  deal since it won't  be visible
anyway.

(click image for full res)



And the whole package:

(click image for full res)



The heatsink unit mounted on the M/B. You need to unmount the fan
to do that. You can again see the bent fins here.

(click image for full res)



And mounted, with the fan:

(click image for full res)




That's it for today, thanks for stopping by.


-aw
 

[Raboo]

Cool idea for chipset cooling.

[flibberdipper]

You would think they would put better cooling on that if that was going to happen... Just like my ASRock board...

[alpenwasser]

Cool idea for chipset cooling.

 

I see what you did there!

 

And yeah, I'm rather happy with the solution, it's very solid and gets the job done.

 

You would think they would put better cooling on that if that was going to happen... Just like my ASRock board...

 

 

My line of thinking is that this is a server board, made to be placed in a chassis with massive

airflow being pushed through the case front to back not something like what I have here, hence

the lack of chipset cooling. But yeah, fun times... :rolleyes:

[CJPowell27]

2 puns in 2 pages good idea for cooling the chipset Alpen

[alpenwasser]

Disk Ventilation


Although disks  have become  quite frugal  when it  comes to
power consumption these days (at least some of them) and HDD
cooling is not really a  huge issue for most people, packing
24 disks  as closely  together as in  this build  will cause
heat issues without ventilation. There is no need for 3k rpm
Delta fans  though, a  whiff of cool  air breezing  over the
disks should do the job nicely.

For  this purpose,  as you  may have  seen in  some previous
pics, I  have chosen 6  120 mm Papst fans,  specifically the
4412 GLL  model, and am running  them at 7 V. The  fans draw
air in through a vent area,  and it then gets passed through
the M/B compartment and out the back.

Each fan is fixed to a rail  riveted to one of the disk rack
panels with two screws.


You've  seen this  before,  but for  completeness' sake  I'm
adding  the pics  of the  bushings used  to prevent  the fan
frames from being crushed to this update as well:

(click image for full res)



I exchanged the  copper screws for some silver  ones, and in
the process  added some  dampening foam between  the mouning
rails and the fan frame.

(click image for full res)



The whole fan panel assembly:

(click image for full res)



While doing  some test runs,  I noticed that a  rather large
amount of  air was being  expelled through the front  of the
case instead of  going into the M/B compartment  and out the
back (I wasn't  really surprised by this seeing  as how open
the front was). Obviously, this was not optimal. So I took a
1.5 mm panel of alu and bolted it to the front.

Because the existing  front has a few folds in  it, I needed
to do some cutting on the case first.


(click image for full res)


(click image for full res)


(click image for full res)


(click image for full res)



After having  done that, I  turned my attention to  the side
panel, making  an opening for the  ventilation. I thought of
several ways of doing this, but  all of them were a bit more
complicated than I'd  have liked them to  be. Cutting such a
big  hole  with  a  dremel  isn't  really  practical,  so  I
considered doing it  with our jigsaw, but after  doing a few
test cuts I didn't really like  the result as I couldn't get
a  straight enough  cut. And  the cut  needed  to be  clean,
because there's no  space to fit a U channel  over the edge,
and I  don't really like the  idea of covering it  up on the
outside.

Anyway, the guy  just used a nice big angle  grinder for the
cut, and since  he's a metal worker by trade,  it turned out
almost perfectly straight  (not 100%, but it's  still cut by
hand, after  all ).  After that, I  painted the  bare edge
with some model paint to not have the blank metal staring at
me.

I thought about painting the mesh, but at the moment I don't
really have the  time, plus I kind of like  the look of this
bare piece of alu, so I've left it as-is.

(click image for full res)



The mesh  doesn't cover the whole  fan area (nor is  it very
open with those  rather narrow slots), but there  is no need
for high-power ventilation here, so this is not a big deal.

(click image for full res)


It's fixed to the inside of the panel with some double-sided
adhesive tape.

(click image for full res)



And in its final config:

(click image for full res)



Drive temperatures  hover between 28 deg  C and 35 deg  C at
the moment, ambient is about 23 deg C.


Until next time,
-aw
 

[LukaP]

Really nice build, been following it, have finally subbed

Good luck

[Vitalius]

If you want straight lines, what you do is you first draw where you want the lines at (chalk or erase-able marker). 

Then you use a jigsaw to cut as close to that line as you can without going outside the line (be careful, no take-backs).

Then you use a fairly tough file (as in, for grinding stuff) to grind the edges until they are as close as possible to being straight using your drawn lines as a guide. 

It's more work, but if you are good, you can easily get straight lines (99.999% straight anyway) that way.

[wpirobotbuilder]

It somewhat resembles a PowerEdge VRTX. Except much cooler because it's got custom hardware inside.