UPDATE 2022/06/04: A few details kept nagging at me, mostly that if I push the HSF onto the CPU heat spreader very little heatsink compound would transfer from the HSF to the heat spreader. This made me think there was still something physical preventing ideal contact. To ensure the brackets were not hitting anything, I moved them to the top of the heatsink. This change necessitated getting longer screws, and longer screws required different springs.
An aside regarding springs: the original springs (both from the stock HSF and the extra sets I bought from Alpha Novatech) were too long with the 19 mm screws (SS-19-4.2-M3-4.2, they were pretty much fully contracted and at their solid length, which could cause them to take a set, ie they would likely warp and continually reduce their spring coefficient over time), which means the distance from the top of the shoulder screw to the bracket was 3.5 mm. This short of a distance meant it was very unlikely to find a spring that would work because most have solid lengths of at least 3 mm, so I needed to use longer shoulder screws. Changing over to the 22 mm screws (SS-22, both use smaller diameter e rings than the stock screws, but you can also order them from AN) the working distance would increase to 6.5 mm. Looking on the AN site, I took all of the spring data and dumped it into a spreadsheet to figure out which springs could be compressed to 6.5 mm and be at about 70 % of their working range (this is the limit AN gives before springs take a set). From there, I chose the strongest spring that met these criteria (if you would like a copy of the spreadsheet, DM me) and chose S001YJ08. The end result was idle temps from 28 C to 30 C and temps under full load are 40 C to 43 C.
Now on to the X11SDV fun times. Trying to swap the cooler from an X11SDV to my test X10SDV was a no go. Turns out the X11SDV HSF holes (left to right) are 3 mm closer than those on the X10SDV and the front-to-back hole distances are far longer. Womp womp... One revelation that came right away is the stock HSF springs from the X11SDV are _stiff_, which might explain the lower temperatures. Additionally, the shoulder screws are fatter and have coarser thread so using the same screws for an X10SDV and an X11SDV will not work. I also noticed the HSF base is not long enough to make contact with all of the heat spreader, so I should have been using a copper shim (self -= 1). Going back into OnShape, I took the X10SDV design and modified it (the first one did not work, which is how I found out the 3 mm problem). Once the second set of brackets arrives next week I will give them a try and post another update.
UPDATE 2022/04/24: The temps were not what I was expecting, because this same modification on my 3 x11sdv (1x 8 core and 2x 12 core) resulted in idle temps around 20 C and load temps averaging about 40 C. The 1541 first converted below turned out pretty well, but the other 2x 1540 and 1x 1541 nodes I converted below had idle temps of 33 C and load temps of 55 C. This makes no sense to me with these older models having TDPs of 45 W vs the 90ish W for the X11SDVs. After taking these apart a couple times I finally figured out what was going on. The VRMs between the CPU and the ethernet ports are tall enough to keep the HSF from sitting flush on the CPU because the ends of the heatpipes extend over the top of these VRMs.
The two solutions I came up with are:
1) Re-design the aluminum brackets to shift the HSF back the other direction, in which case the ends of the heatpipes would not touch the VRMS (and the HSF would no longer be centered on the CPU).
2) Use a pair of pliers to bend up the ends of the heatpipes.
As you likely guessed, I went with option (2). This lowered the load temps another 5 C and, so far as I can see with the limited viewing angles, the heatpipe ends are no longer touching the VRMS.
Some day I plan to swap a setup from one of the 1540 nodes with one of the X11SDV nodes to see what difference it makes.
@TXAG26 I am very sorry for the long delay, my 6 (now 8) month old and a new job opportunity kept me from my computers.
Here is the bare board with the brackets in front, now to drill and tap the brackets. I did not have a 3.3 mm diameter drill, but Ace Hardware had a bunch of #30 drill bits which worked just fine, and then an M4 tapered tap with a couple drops of Tap Magic. I kept the HSF in its cardboard packaging to protect it during the process and keep as many metal shavings out, as possible. I also used a chamfering bit to put a bit of a taper on the hole to make starting the tap easier. I used a tapping stand, as opposed to free tapping the holes, because starting them was difficult when attempting to free hand them. Third thumbnail depicts using a hand tap to cut the threads a second time and get more of the chips out.
On to assembling the brackets, the parts all of the shoulder screws, springs, washers, and C rings came from the OEM heatsink.
I tried to find replacements, but have yet to successfully do so. AlphaNovatec has a good assortment of shoulder screws, but none with the proper C ring-to-tip distance. Be careful of any washers you get, I picked up some PTFE ones from McMaster, but they were tall, but the ID and OD were similar and somewhat too large so the spring didn't seat very well on top of them. Lastly, C rings are ideal because they easily stay clear of the M4 screw holes, but the circular kind with an internal star often have too large of an OD.
Here is the completed assembly using low-profile stainless steel M4 x 10 mm screws from McMaster. Were I to do this again, I would order ultra low-profile screws to ensure there is more distance between the screw head and the mobo (though not right now b/c they cost $4 per screw). The current low-profile screws leave 0.5 to 1.0 mm of distance, and I am loathe to dremel or file them which would render them noticeably more magnetic.
After lots of isopropyl alcohol and compressed air to clean up the HSF, next came the installation. Ideally I would have figured out a method for getting at the shoulder screws without bending or cutting the fins, but using a low-profile screwdriver would have taken an eternity. Once again, Ace Hardware came through with a narrow-shaft flat blade screwdriver, which only required bending the fins a bit (previous incarnations used a 1/4 in drill bit to make holes through the fins, which was ugly and made me feel dirty).
Now on to the installation, which took a bit because I had to add AS5/install the HSF/uninstall the HSF several times to completely coat the CPU die.
Finally, testing the end result! With the stock Noctua 12 mm fan the load temps (using stress --cpu 8) for a couple days and the temps varied from 36 C to 41 C, which is over 10 C lower than the custom HSF I previously created. Another big benefit is the noise is incredibly quiet relative to the 2x Sunon 60 mm fans.
If you have any questions, just let me know.