I got a pc that suffers from severe heating problems because of its poorly designed passive cooling case. I tried several strategies to put temperature under control, did a lot of tests with thermald, but ended up hacking a fan into the case. However, I still have problems with ssd temperatures.

Right now, as I write this, the nvme ssd is at 67ºC and easily goes beyond 80 or 90 during data transfers. I already lost 2 ssds before realizing this was the problem, because I was controlling just the cpu temperature.

There's no physical way to force airflow where the ssd is, so I was wondering if there's some way to throttle data transfer of something to prevent the ssd temperatures from reaching a threshold. All I could find was a vibecoded github repository. I don't care if data transfer gets ridiculously slow, as long as the ssd remain safe, and I can't consider getting any new hardware right now.

I'm currently running fedora 44 on that machine

Edit: I thought there was an easier software mitigation to that, but apparently, solutions will tend to the hardware side, so here are the hardware details:

  • It's an unbranded chinese cheap mini pc with a heat dissipation case that doesn't dissipate well
  • The ssd is mounted on the back of the motherboard. The model is a sandisk M.2 1TB 2280
    Here's a photo of the case, for the reference (don't judge me lol):

image

you are viewing a single comment's thread
view the rest of the comments
[–] 1 point 4 hours ago* (last edited 2 hours ago)

I’m gonna start off with an apology because my flippant tone may have made it seem like I was targeting you when really I’m running on all eight cylinders trying to blast out accurate, useful posts that are fun to read. I’m sorry I hit you with a little friendly fire, that wasn’t my intent.

Working backwards: ca glue doesn’t make a chemical bond with the plastic of a semiconductor package, it creates a solid material out of itself using water that relies on the surface dressing of the stuff it’s being used on for mechanical purchase. That’s why you gotta use sandpaper to rough up stuff you’re trying to glue together first. That also explains why supergluing a nut to an ic doesn’t hold like you’d think it would and in the case of for example the big boy fets and voltage regulators of a video card creates a weaker bond than the solder to the pcb (I was carrying a mining gpu by the bolt glued to the vrm and the glue failed and the card dropped three feet onto concrete).

Removal of ca glue can be easily accomplished with acetone. Be careful using acetone on or around electronics because it will eat certain plastics (the “a” in ca glue stands for acrylic!).

What you do need to be careful with when using ca glue is natural fibers. For reasons I have never felt (see what I did there?) the need to investigate, it reacts with them and can get really hot.

I’m gonna combine the other two sections of what you wrote into one because there’s a misconception there.

Based on the ops own post, the ssd is idling at 67 or so c. We wanna keep it below 90 for [explanation redacted].

That doesn’t mean there is 67c air inside the pc case. That would be crazy. It means the ssd can’t effectively dissipate its heat into the surrounding air at idle until it reaches 67c.

The equation for that is (kATd)/l.

K is the ability for the material to conduct heat, A is the area over which it’s happening, Td is the difference in temperature and it’s all over l, the distance it’s happening across.

Basically every part of that equation is bad numbers when we’re using a bare die into the ambient air.

But why?

Well straight off the bat, air sucks. Heating it up is way harder than heating up for example plastic, water, oil, metal or human skin. That’s why when you run out of heating oil for the furnace and use the coil stove burner to heat the house you can keep your hands cozy by rubbing them over it but get a crazy burn when you put them on the coil. Air is such a bad conductor of heat that the difference between “ahh” and “AUGH!” Is a couple of inches.

But plastics suck too! Remember that the part of the ssd that gets hot is the big ol chip and the part of the chip that gets hot is the silicon die inside it. Plastics are in general so bad at conducting heat that in many cases we use them explicitly for that purpose in industrial engineering. If you’re worried that some parts variable temperature will have an effect on some other part then you can plop a chunk of nylon or something down in between em and suddenly the temperature shifts are good enough for government work!

So the ssd is getting too hot before it can transfer its heat to something else (the air). It’s being hampered by how shitty the two materials plastic and air involved are at doing this job but that isnt the whole story! We havent even talked about surface area or thermal mass!

The chip could effectively conduct away more of its heat if it were bigger. If you have a one inch square chip it can only radiate (I’m playing fast and loose with the words here but all throughout this post I’m actually only talking about conduction, the amount of joules lost to terahertz radiation in a ssd chip is negligible) into the air close to the surface of the 1 square inch chip. If it were three square inches it would be able to heat up three times more air! A heatsink will usually have a knurled surface with fins and other hard to dust features that raise its surface area super high. It’s not uncommon to see a little bitty heatsink with an effective surface area of a foot square!

We can also solve for bigger in a different direction, and imagine a chip with a one inch cubed package size as opposed to one inch square. Now not only do you get to take advantage of the 5x increase in surface area, but you have much more mass to mess with. If it takes one joule of energy generated in the ssd chip die as it recalls futa SpongeBob memes to heat up the normal 1 inch square package surface one degree, the one inch cubed package can take much more bikini bottom dick pics before it heats up one degree. Because it’s bigger. See what I did there? Futa, bigger? Never mind.

That’s called thermal mass.

Oh yeah and I guess in a heatsink there’s usually some kind of conductive material to make a good connection between the package and the heatsink surface. I didn’t talk about that but I’m running out of energy so just imagine a good explanation of why even something not specifically engineered to do the job like ptm is still better than shitty ol microscopic air pockets.

So to recap: the heat is coming from inside the ssd. We want to get it away from the ssd. We can use a big thermal mass to the a huge surface area made out of a material that conducts heat really well coupled to the die with some solid or liquid to both reduce the effect a joule of heat generated has on the die temperature and increase the rate at which that heat is conducted into the air.

Adding a heatsink improves literally every part of the math for getting parts cool and if that doesn’t convince you I encourage you to experiment on your own with your hair dryer and see what happens.

E: I didn’t even talk about the duty cycle of the chip getting hot. Unlike your computer cpu which even in a real low pee state still does something, an idle ssd controller is pretty much off (they still fiddle fuck around with blocks and do other dumb shit but they’re consuming less energy than an idling cpu and therefore need to dissipate less heat. The point stands.). It’s only under some load that heat gets generated and needs to be dissipated.

So if you’re just looking through your futa SpongeBob memes folder and having a chuckle the spikes of load are probably not enough to even heat up the heatsink appreciably while transferring your friends usbc futa SpongeBob memes folder on to your ssd would result in the need to rely on the thermal mass of the heatsink and its big surface area and its much better ability to transfer heat into itself from the chip and from itself into the air.

  • source
  • parent