mettelus
Out of curiosity Jonesey, have you done an SPL check? Noise unto itself is subjective, so is difficult to understand the situation.

No, but I can do
 
Thanks for all the ideas so far chaps, didn't get chance last night to investigate further but intend to tonight.
 
FYI: graphics card is passive - no fan!

www.quietpc.com ftw!  I get my CPU and case fans from those guys.
 
Make sure your PC's internals are being sufficiently cooled by the case fans (i.e. ensure you have sufficient case fans fitted), otherwise the CPU fan will ramp up more than it'd otherwise have to which is possibly why your computer gets louder over time.  Do you have any (temperature) monitor software which you can run to investigate this?
 
Also, what is below the PC in the rack?  Anything that can generate much heat?
 
[edited re: gfx card is fanless]

 
I have also used Noctura cooling and fans. They work very well and are quiet

Note that mounted HDs (even when mounted via anti-vibration mechanisms) can cause a case to resonate.
If you have a case with many 3.5" drive bays (and you notice this type of noise), try moving the HD/s to different tray slots.  Essentially, you're "tuning" the vibration... so that its source is in a less sympathetic position (won't resonate).
 
You'll also notice this type of issue with mounted toms (drums)... albeit in reverse.
Even with higher-end drum kits, where you place the tom on the tom-mounting arm is critical.
At one spot, the tom will sound completely dead... as the vibration/resonance cancels out (similar to a dead-spot on a bass neck).  Move the tom to a different place on that same tom mount... and the tom's sustain is long/clean (unimpeded).
 
Back on topic...   
 
Once you are used to a quiet PC, the genie is out of the bottle.  There's no going back.

I wonder how far away we are from being able to use this kind of tech for cooling:
 
http://smallformfactors.mil-embedded.com/articles/immersion-form-factor-server-class-systems/
 
e.g.
 
https://www.youtube.com/watch?v=CIbnl3Pj15w
https://www.youtube.com/watch?v=Z_X_hgtlJpA
 

Jim Roseberry
Note that mounted HDs (even when mounted via anti-vibration mechanisms) can cause a case to resonate.
If you have a case with many 3.5" drive bays (and you notice this type of noise), try moving the HD/s to different tray slots.  Essentially, you're "tuning" the vibration... so that its source is in a less sympathetic position (won't resonate).
 
You'll also notice this type of issue with mounted toms (drums)... albeit in reverse.
Even with higher-end drum kits, where you place the tom on the tom-mounting arm is critical.
At one spot, the tom will sound completely dead... as the vibration/resonance cancels out (similar to a dead-spot on a bass neck).  Move the tom to a different place on that same tom mount... and the tom's sustain is long/clean (unimpeded).
 
Back on topic...   
 
Once you are used to a quiet PC, the genie is out of the bottle.  There's no going back.

Hi Jim, thanks for chipping in.
 
There are now only 2 HDD's (used to be 3) and yep, my old machine was nigh on silent so I was expecting the same from this box.
 
Home time soon....

A case that's intended to be low-noise is going to have variable-speed fans. If the fans are always running at high speed it could be an indication of inadequate airflow. This would not surprise me in a rack-mounted device, wherein larger, slower-moving fans might not be practical.
 
If possible, I'd situate the computer at the bottom of the rack so there's nothing producing heat below it, and leave a 1U gap above it. You can get rack filler panels that are vented. I suppose you could test this by measuring noise levels immediately after turning the computer on, and comparing that to a couple hours later. This could be accomplished by mounting your SPL meter on a mike stand so it's locked in a fixed position.
 
This is assuming the problem is really fan noise, rather than hard drives. The difference is that fans produce broadband semi-white noise that can be mitigated with acoustical absorption, whereas case rattles, resonances and hard drive spindles have to be treated mechanically rather than acoustically.
 
All of this could be moot, though. If the noise level as seen at your microphone is 30 dB or more below the ambient room noise, it's probably not affecting your recordings anyway. My computer is only 3 feet from my microphone and with 5 fans is by no means a "quiet" computer. But it's on the dead side of the mike and there is acoustical absorption between them. Even with the mike gain cranked, the fan noise picked up is down -60 dB or more, so I don't worry about it.

If I am reading you correctly, the noise diminished or disappeared when the case was removed. If so, then it is not so much an issue with the noise source as with the enclosure. But if removing the case also removes the fans, then the issue is less clear, and if it also involves removing the unit from the rack, then rack vibration is still in the differential. The full test is to assemble/disassemble the unit on a noise absorbing pad, and unplug the case fans if they are attached to the removable section. You can also examine the questionable components using a poor man's stethoscope, your ear at one end of a piece of pvc or other handy pipe and the other end close to the component in question.
 
Assuming that the problem is isolated to the case environment there are a couple of possible issues. Vibration is one possible mechanism, something in the case is picking up a frequency from the root problem component, and that is setting up a sympathetic vibration that either magnifies the sound or re-radiates it at a more annoying frequency. That problem can be less with a heavy case (imagine replacing the paper of a speaker cone with plywood) but there are still possibilities that have to do with connections of the case components like a loose screw. The other issue is acoustic reinforcement, sounds inside the case can be bounced around in such a way that there is positive reinforcement of the waves, like the chamber of a violin. A 2mm sheet is about 13 or 14 gauge, and may not by itself be enough to completely kill vibration, especially if the natural resonance of the sheet reinforces the noise. Attaching a mass to an off center location on the (usually largest) sheet could alter or eliminate that resonance. Rubber grommets or fiber or plastic washers or tape along the joins and screws may help. In a rack mount there is very little room inside the case, but a noise absorbing material there would be most effective at killing echo/reinforcement (you put the bass traps inside the room). Wrapping a noise reducer around the outside of the case is not going to prevent sound from getting out via the vents. The traditional way of handling that kind of escape is with baffles in a duct connecting to the vents, but that takes a lot of space, albeit external to the unit, and the fans that are designed to move air a short distance may not be up to the task. Adding a large fan inside the duct creates another potential noise source. I do very little recording, and I grew up on vinyl and AM radio, so my brain easily filters out the low level crap when I am listening. If I have to record, I do it in a closet packed with clothes and bedding (not created specifically for this purpose) and run a mic cord to the room downstairs where my computer lives. 
 
 

bitflipper
All of this could be moot, though. If the noise level as seen at your microphone is 30 dB or more below the ambient room noise, it's probably not affecting your recordings anyway. My computer is only 3 feet from my microphone and with 5 fans is by no means a "quiet" computer. But it's on the dead side of the mike and there is acoustical absorption between them. Even with the mike gain cranked, the fan noise picked up is down -60 dB or more, so I don't worry about it.

I agree with this, and is a good "practical" point. Because of this thread I checked my machine's SPL (simple phone app) and was 21dB with the machine off and 26dB with the machine running with the meter at "normal mic location." Simply rubbing my fingers together spiked it into the 35dB range (or even thing like shifting in chair (40dB), picking up/rubbing a piece of paper, etc.). It is easy to get fixated on something that we notice. With the same app my vocals exceed 70dB consistently, so the S/N ratio is pretty big and easily managed.

mettelus

bitflipper
All of this could be moot, though. If the noise level as seen at your microphone is 30 dB or more below the ambient room noise, it's probably not affecting your recordings anyway. My computer is only 3 feet from my microphone and with 5 fans is by no means a "quiet" computer. But it's on the dead side of the mike and there is acoustical absorption between them. Even with the mike gain cranked, the fan noise picked up is down -60 dB or more, so I don't worry about it.

I agree with this, and is a good "practical" point. Because of this thread I checked my machine's SPL (simple phone app) and was 21dB with the machine off and 26dB with the machine running with the meter at "normal mic location." Simply rubbing my fingers together spiked it into the 35dB range (or even thing like shifting in chair (40dB), picking up/rubbing a piece of paper, etc.). It is easy to get fixated on something that we notice. With the same app my vocals exceed 70dB consistently, so the S/N ratio is pretty big and easily managed.

An excellent point, and reason enough I think to come out of the closet, except for the fact that my computer lives in a room full of file cabinets and the wife's trousseau makes for a cheap anechoic chamber.