News SilverStone Reveal Case For Raspberry Pi 4

Jun 17, 2020
1
0
10
If the heatsinks don't touch the top part of the case, it would be sensless. Anyways, in the picture, the second heatsink is located on the usb controller, not the ram. The ram is the bigger Chip next to the cpu, or not?
 

deesider

Honorable
Jun 15, 2017
298
135
10,890
Had it been available at back in December, I'd've definitely considered this for my Pi4 in lieu of the Flirc case, though thus far, I have ZERO complaints about the Flirc.
The Flirc case appears to be an all encompassing aluminium shell with soft touch panels on top.

How much is the wifi functionality affected?
 

CooliPi

Reputable
Oct 4, 2019
72
22
4,535
The picture shows small heatsinks on the top of a BCM chipset and USB bridge, NOT ram. RAM is not power hungry. USB bridge can thermally throttle and stall data transfers if it's too hot.

If the heatsinks aren't connected to the case, they're barely helpful. The FLIRC case has it connected (it's just a protruding column of the case material - however no direct cooling for the USB bridge), so I assume it'd be better than this.

Last but not least, we've started with a "no compromise" approach to RPI4 cooling and ended with CooliPi, about 9 degrees better than FLIRC under the same conditions. The problem with WiFi signal haunts all closed, metallic cases. So CooliPi has 3D printed plastic base and the heatsink directly touches all the hungry chips.

Remembers me we should continue our effort to offer it in the U.S. through Crowdsupply....
 
  • Like
Reactions: King_V and bit_user

bit_user

Polypheme
Ambassador
I'd rather see SBCs adopt a standard PC form factor, like mini-STX, than have PC case manufacturers make custom cases for particular boards.

Also, people would do well to use copper heatsinks with more surface area + a decent thermal compound!
 

deesider

Honorable
Jun 15, 2017
298
135
10,890
Also, people would do well to use copper heatsinks with more surface area + a decent thermal compound!
The great thing about thermal pads is that you can just stick the heat sinks on. When using proper thermal paste, a mechanical means for attaching the heat sinks is also needed...
 

bit_user

Polypheme
Ambassador
The great thing about thermal pads is that you can just stick the heat sinks on. When using proper thermal paste, a mechanical means for attaching the heat sinks is also needed...
Theoretically, yes.

I have installed heatsinks in a DVD player, a game console, and a first-gen Pi. In each case, I just used thermal compound, since the heatsink was sitting level, and it's been fine.

When the stakes are comparatively low and everything is sitting flat, you can get away without a mechanical means of securing the heatsink.

However, I do hope that Pi v5 has a pair of screw holes on opposite corners of the CPU, to support a mechanical restraint. I would prefer to do it "the right way", even if I think I could probably get away with something less.
 

CooliPi

Reputable
Oct 4, 2019
72
22
4,535
.....
When the stakes are comparatively low and everything is sitting flat, you can get away without a mechanical means of securing the heatsink.

However, I do hope that Pi v5 has a pair of screw holes on opposite corners of the CPU, to support a mechanical restraint. I would prefer to do it "the right way", even if I think I could probably get away with something less.

I don't expect them to do it - it takes valuable, precious space. They're battling for every mm^2 of the board area. I've read that a prototype of Raspberry Pi 4 was 5mm longer (in the longer dimension) and the've been optimising itt for half a year to shorten it. The movement of an ethernet jack to the other side was necessary.

The RPI4 has a quadrature of holes, which can be used for heatsink attachment - exactly what we did with CooliPi 4B. But - given they're far away from the SoC - you need some mechanism to create force from the underside of the board against a heatsink. We've bet on the PCB rigidity and flexibility and it paid off. The heatsink is designed in a way in which it's fastened by four screws (hexagonal nuts) BUT the Aluminium island above the BCM chipset (and the two other chips) is 0.3mm closer to the PCB than it should nominally be. This distance is enough to exert a small force on all the chips. It makes a triangle of contacts (every one of the 3 chips), defining a plane - the PCB is bent a little bit between each corner and this middle triangle.

This strategy has paid off well. It's a good balance between the exerted force, PCB bending (not too much) without the necessity to create any support mechanism under the board to force it against the heatsink. Using thermal paste is highly recommended.

I've added a page to our support site here
https://www.coolipi.com/Direct_contact.html
Some users were keen to see it, so there it is.

Your idea would help only if it were useful to install a heatsink smaller than the board. That's not the case - any decent coolercase needs to cool by its chassis to gain surface area. Any small heatsink inside a case is worthless. And with the 3-8W power budget, you need something between this size (SilverStone case) and CooliPi on the upper side of the size spectrum.

Any future SoC will be either underclocked to fit into some rectricted power envelope, or running at full tilt but with higher power draw. And power users will demand more performance for desktop anyway. So, with any future process technology RPF could pay for, we power users will demand to get the most of it. And that means (again) the biggest heatsink money can buy.

And any big heatsink can be mounted as CooliPi 4B - using existing holes.
 

bit_user

Polypheme
Ambassador
Your idea would help only if it were useful to install a heatsink smaller than the board. That's not the case - any decent coolercase needs to cool by its chassis to gain surface area. Any small heatsink inside a case is worthless. And with the 3-8W power budget, you need something between this size (SilverStone case) and CooliPi on the upper side of the size spectrum.
I was just saying what would work better than Silverstone's solution - not that it would be perfect.

What I described is adequate for my 2-3 W, first-gen Pi.

with any future process technology RPF could pay for, we power users will demand to get the most of it.
Eben Upton said in a Q&A on this site, just over a year ago, that they expect to stay on 28 nm for a couple more generations, at least. He said it currently offers the lowest cost per transistor, which is what they're optimizing for.

And that means (again) the biggest heatsink money can buy.
Yes, cooling will likely continue to be a limiting factor.