Not-so-SFF 34.7 x 20 x 37.2 cm (26 L) case project :
I'm designing it with the following constraints/objectives :
- "Classic ATX tower" level of component compatibility
- Minimal accessories cost (no PCIE riser, PSU internal cable...)
- Accessible DIY construction with minimal tooling
- Efficient dust filtered air cooling
- Maximum accessibility of the dust filters
- Low noise level
- 4x 3.5" bays
- Laser cut 2 mm aluminium (side, front and top panels)
- Laser cut 1.5 or 1 mm steel (dust filter frames, internal pieces)
- Screwed assembly (10x10x10 6xM3 cubes + corner brackets)
- Mini-ATX board
- 160 mm ATX PSU (with additional 50mm "rats nest" for cable non-management)
- 3-slot GPU : max length 330 mm - max height 170 mm
- CPU cooler : max 165 mm height - max 12 mm overhang over the ITX board side
- 4x 3.5" HDD bays
- 2x 2.5" HDD bays (don't need them, but the option is there)
- 3 (almost) isolated chambers for PSU, GPU and CPU
- All intakes with large 15mm wide cutouts covered with 100-200 µm (TBD) stainless steel mesh glued on easily accessible magnetically attached steel frames
- All intakes feature a "pressure buildup chamber" to prevent fan performance hindering and for noise reduction.
- PSU chamber : bottom intake (10 mm pressure chamber), rear exhaust.
- CPU chamber : 140 mm fan front intake with a 90 mm pressure chamber, unobstructed exhaust directly in front of the CPU cooler. Rear exhaust with a 120 mm fan + cutouts in the rear panel
- GPU chamber : designed for a deshrouded+defaned GPU. Top intake with a 15 to 30 mm pressure chamber (depending on GPU thickness) for 2 x 120mm fans sitting directly on the GPU radiator, using a height adjustable fan support. The fan support entirely seal the intake to prevent any hot air recirculation. The GPU / CPU chamber separator is designed to channel the air flow toward the sides and to prevent it to leak into the CPU chamber. Large cutouts on each side panel allow passive exhaust through positive pressure.
This is something I was thinking about for a while. My last and only ITX build was in a Sugo SG13 nearly 10 years ago, which is now taking dust in a garage. My current setup is a micro-ATX build in a Define Mini, and it's end of life is getting near so I'm thinking about rebuilding a new ITX setup.
I've been closely watching the ITX developments for a couple of years and to my despair, the market seems overly focused on watercooling and tiny case volume. I get the excitement around that, but from a practical view, I see little point in shrinking case volume once you get past the 10-15L mark.
Unless you switch to a vertical layout, desk footprint of a 15L case or a 25L case will only vary by a few centimeters, and anything larger than 15L is definitely not something you can easily carry around. I also think there is an inflection point at around 25L where volume shrinking start to affect all metrics : cooling performance, noise level, cost. For most components, there is still a significant cost and/or performance penalty to size shrinking : SFX PSU vs ATX PSU, CPU air coolers, 2.5" vs 3.5" drives...
There aren't that many cases designed for mini-ITX in the 25-30L range. At that volume range, many cases switch to micro-ATX compatibility, becoming quite volume inefficient as a side-effect. For my new build, I tried to find the following unicorn : a mini-ITX case for ATX PSU, full length (minimum 300 mm) 3 slot GPU, 160 mm CPU cooler and at least 3 3.5" bays. I would expect those specs to fit easily in a sub-30L case. Granted, the need for 3.5" bays is a major issue.
I didn't find any horizontal mainboard layout case that fit the bill. I don't really like that layout anyway, it imply a much wider desk footprint and air flow is usually pretty bad in that setup. Tower cases like the NZXT H210 and Fractal Design Define Nano S (both ~27L) or Phanteks Evolv ITX (34L) have the same shitty layout with the PSU and HDD cage at the bottom blocking the (max 2 slots) GPU intake. There are only 2 3.5" bays in the Evolv ITX and Nano S, only 1 in the H210.
The Phanteks P200A (30L) is as close at it gets to what I need. It is a classic tower layout with the PSU at the top, 3 slot GPU support with bottom intakes and 2 x 2 (optional filmsy mounted) 3.5" bays. It is a decent case, very reasonably priced, but it has a some flaws : the 3.5" bays are obstructing the front fans and the PSU intake is internal (in front of the CPU cooler).
I could have gone for that case (and in the end, maybe I will), but I want to see if I can design a better layout.
GPU positioning
I really like the concept of a direct intake deshrouded GPU, like what is commonly done in the NR200, NCase M1 and what could potentially be done in the P200A. I've read and watched as much as I could find on the topic to compile feedback on that setup.
The idea only works well if (ideally both) side panels are open so the GPU radiator exhaust can escape the case immediately. Otherwise, that air either get sucked down by the GPU fans if the bottom of the case has openings, causing recirculation of the hot air in the GPU, or gets pushed up toward the CPU area, causing a severe degradation of CPU temps.
First conclusions :
- The bottom of the case must be sealed to prevent direct (inside the case) recirculation of the GPU exhaust down toward the GPU fans intake.
- Large openings must be made in the side panels, directly in front of the GPU radiator fins.
- The air flow must be prevented from moving past the GPU backplate. Ideally, the backplate must be "extended" as to seal completely the GPU from the rest of the case.
With the "GPU at the bottom" case configuration, the fans performance is massively hindered by the lack of breathing room clearance under the case. Thicker feets quickly increase the case occupied volume, so not a good solution (plus it looks weird).
Moreover, in that configuration, the intake air flow is parallel to the desk surface, in the same plane but in an opposite direction as the GPU exhaust from the side panel. This is a perfect "flow loop" geometry, massively increasing the likeliness of the side panel hot exhaust being recirculated by the GPU fans.
The obvious solution is to turn the case upside down, GPU at the top. No more "breathing clearance" issue for the fans, and intake flow from the top is now perpendicular to the side exhaust flow, reducing the likeliness of recirculation. Ideally, I would prefer a case with a solid top panel, but you can't have it all. And unless you find acceptable for the PSU to intake from inside the case (severely hindering the usefulness of a semi-passive PSU), the PSU intake would have been at the top anyway.
Fan front pressure chambers
A fan air flow and noise performance decrease exponentially when the air flow in front of it obstructed. I didn't do any testing, but what information I could find online finished to convince me but that a single unobstructed fan is worth 3 fans with dust filters strapped right in front of them and a mesh panel 15 mm away.
That's the reasoning behind that massive chamber in front of the 140mm fan, as well as putting as much space as possible in front of the GPU and PSU fans. This allow more filter surface area, less air resistance, a slower and less turbulent air flow, resulting in better fan performance and less noise.
The front chamber setup allow a straight unobstructed flow path through the length of the case for the main CPU chamber.
Possible caveats
- I'm betting that by optimizing its efficiency, the single 140mm fan will provide enough fresh air in the CPU chamber. I initially started with a 180mm fan there. However, this only worked with only 3 3.5" bays instead of 4, required to raise the case height slightly, and the fan would have been slightly in front of the PSU backside, likely causing a lot of turbulence. I could technically fit a fourth 3.5" bay on the side of the PSU, in place of the 2 2.5" bays, but this is a tight fit and it would make cable management a lot messier.
- The GPU/CPU chamber separator isn't a perfect seal. To account for varying GPU backplate thickness, I have to let a few milimeters of leeway there. The separator plate overhangs quite a bit over the backplate, but still, there is a possibility that a significant hot air flow manage to squeeze through. Maybe I can achieve better sealing using some kind of compressible foam joint.
- The huge GPU exhaust openings in the side panels would likely e quite effective at letting GPU coil whine noise out. Not sure what the solution could be. I have vague plans to cover the side panels with 4mm sound dampening material, maybe that could help a little.
