An Addlink D60 SSD: 960GB, PCI-e v4, 1DWPD for 5y, big caps. Not terrible... except for power: 4W idle, 12W max. 3.6A, or .4A/contact (M.2 M-key has
9 power contacts), rather close to the .5A/contact spec. Ouch. I didn't even bother to try the factory thermal solution: a sticker. I tried one of
these out with a Thermalright HR-09, and the result was disappointing. So I figured I'd try my old method of epoxying a heat sink to the components.
To prepare, I decided to verify the component heights, and I'm glad I did: the flash and DRAM chips are reasonably uniform, but the Phison controller is around .010"/.25mm lower. That could explain the poor performance of the HR-09: I had used a soft thermal pad (.75mm TG-A1250), but contact with the controller (effectively the single hot component) has to be poor as even good compression would not compensate for the component height disparity. And the HR-09's clamping method does not lend itself to high mounting pressures in any case. Something to think about...
I decided on a shim, to avoid having to modify a heat sink by hand. The additional thermal interface is not ideal, but doesn't seem to be significant. I settled on a 60x20mm shim, to cover the more uniform components. I suppose I could have extended it to ~70mm and stepped it for the caps, but I didn't consider it at the time. The only convenient material I had was 1/8"/3mm 6061 - a bit heavier than I'd like. Shown is the shim in rough form, with steps to clear the flash and RAM chips. Yes, I filed the steps into it - you work with the tools and methods you have handy. I made them a couple thousandths over depth, as close contact with the controller is my only concern.
Underside of the D60.
The D60, finished shim, and heat sink. The notch in the shim clears the tall power components - even with an insulating epoxy, I want to avoid
contact between the aluminum shim and any solder joints. The power controller chip will just have to suffer, as it would in most any other
installation.
Shim applied...
Not my best epoxy job, but far from my worst. Application note: I used Artic Alumina epoxy, as I have it and I have experience with it. The
instructions always say to clamp the parts, but the 2-part epoxy adhesives I've used all behave like the slickest grease you've ever encountered
when heated or compressed, while being rather sticky when not. A clamp that is not itself secured will tend to drag the parts around rather than
fix them in place. If I was assembling more than one or two of these I'd build a fixture to clamp them; these I just squeezed together and left sit.
On a level surface: sticky or not, they will slide, so I keep an eye on them until the epoxy sets.
The assembled D60, with a couple additional heat sink choices if I care to stick another one on (for 25-40% additional surface area).
Another view. The weight doesn't seem too bad (~30g over the bare D60)... particularly when compared to the HR-09 (80g). The heat sink has
~690mm2 of contact area to the shim, compared to ~144mm2 for the Phison controller.
Its performance is a bit less than stellar. Installed in the H5v2, lid on (essentially no airflow), idle, stable:
PM983: 47C D60, HR-09, old: 52C D60, 1 Alpha: 54CI'm not surprised - the thing is a weenie burner. I did not test activity; I may do that eventually. The only relevant data point there is throttling behavior, as the controller is guaranteed to head straight to its spec maximum temperature and nail itself there. For a bit of perspective, the X550 is also (approximately) a 12W device, as are the 2.5" SSDs in the H3v3, H5v3, and IW301, which run quite cool.
Needless to say, I went with the 40% improvement. Here are both D60s now.
For the epoxy job I added the 19x19 heat sink, as it was the largest that would fit conveniently and had the same fin pattern as the 40x20. Added weight is now ~35g.
For the HR-09, this photo is of the old one, but it's unlikely the differences would be readily visible in photographs:
Old: Two 60x21mm pieces of .8mm T-Global TG-A1250;
New: Two pieces of .8mm TG-A1250, 28x21 and 14x21mm on the top flash and DRAM; one 15x15mm piece of 1mm TG-A1660 on the controller; one 14x3mm piece of 1mm TG-A6200 on the back underneath the controller (to help prevent flex); and one 60x21mm piece of 1mm TG-A6200 over the back side.
I had the parts all laid out for a photo, then forgot to take it. I assembled the parts, set the thing upside down, set a chunk of aluminum plate on it to keep it flat, and stacked a couple boxes of bullets on it and left it for a couple hours to squeeze down. The bullets weighed ~12lbs (200x405gr, plus plastic boxes). I then tightened the screws with the stack still in place.
Incidentally, getting the old setup apart was not easy. Those T-Global pads get sticky.
Both SSDs about to get cooking in the H5v2, lid on...
D60, HR-09, new: 54C D60, 2 Alphas: 54C
Well, that was rather anticlimactic. Apparently the ambient environment was not as stable as I had thought. The two SSDs have much different ramp rates (up and down), likely due to thermal mass (up) and thermal resistance (down); both numbers are lower for the epoxied one (thus it reacts faster). Stable temperatures (with lid on/off and various levels of somewhat indirect air flow) ended up basically identical. The original thermal pad installation on the HR-09 seems to have been much more forgiving than I expected.
I'll probably pair the "2 Alphas" D60 with the PM983 in my TV machine (a moderate air flow environment).
Most 2.5" or EDSFF short devices should be superior to M.2. (I specify short as the long devices are kinda ridiculously long... not that I'd turn down a deal on one.) Hopefully these will be the last M.2 devices for me.