I am currently building 15x 5975WX workstations, all identical. Using the Noctua NH-U14S TR4-SP3 with dual fans attached to heatsink and dual exhaust fans directly above it in the case.
Testing with Prime95 on small FFTs all cores to check CPU thermals I am seeing up to 15 degrees difference between CPU’s. I have 2 systems just reaching under 80 degrees Celsius, a lot of the other systems are happily sat around low 70’s and a couple sitting at the mid 60’s. Bearing in mind I can see this temperature difference at the same time, so it can’t be down to ambient temperature. The idle temperatures were all pretty similar as well.
Could this be down to differences between the processors?
I have already tried replacing the thermal paste on one of the high temperature processors and saw little to no change.
For anyone who may be interested, I swapped heatsinks, replaced thermal paste and saw no change and in the end tried our demo 5975WX in one of the systems where the CPU was running hot, it ran significantly cooler around 70 degrees Celcius when under full load. So we have decided to swap these two hotter processors out in this instance.
Yea with that many identical systems you start to see differences.
Certainly could be differences in the IHS shape (slightly convex or concave) or the solder used to attach the IHS to the dies. You could check the vcore under load reported by something like hwinfo, maybe it’s using a slightly higher voltage?
Did you ever consider the placement of the PC in the room? From the outside looking in, maybe the coolest PC is sitting near some vents, etc., and allowing the room’s flow and cooling to affect the PC’s temperatures. But I am unsure if you did all the PC testing in the same location as the room. If that is the case - I digress.
Thank to everyone for their input, much appreciated.
So vcore, I have only done some brief testing but this is an interesting one. On the hotter system it was lower averaging 0.271 where as the cooler system averaged a higher vcore at 0.275. Possibly cooler processor is able to run a higher clock speed resulting in higher vcore, I did notice that in fact on a couple of systems. Maybe 100MHz in it on all cores.
I would love to delid them but I doubt I could persuade my boss, it would be a costly endeavor if it went wrong.
Yeah location wasn’t really a factor, there are no air vents that side of the room, one of the coolest systems was in fact sat directly next to one of the hottest. They were both facing the same way in parallel so one wasn’t exhausting on to the others intakes. I even opened various doors to our workshop to create a draft and they didn’t significantly alter the temperatures between systems, only made them all run marginally cooler.
Looks like good troubleshooting methodology. I’m surprised to see such variance. ~8 C seems like more going on than just the silicon lottery. Not at all surprised that delidding is not an option though. I certainly wouldn’t want to explain how I bricked a $3k+ CPU. What’s your plan with the hotter chips you swapped out?
We have a good relationship with our distributor that we purchased them from, so they have gone back and we have finally received and installed new cooler chips last week. We saw temps on the new chips of around 70 degrees Celsius, which I was very relieved about as I was concerned we would get another hot chip or two.