Cooling Do I need special thermal paste for an Intel Atom SOC chip?

zovc

King of Cable Management
Original poster
Jan 5, 2017
852
602
Hey!

So, I assembled a system the other day with my Supermicro A1SRi-2758f, before installing the board I removed the CPU cooler to check its application of thermal paste... it wasn't great. I tried taking some rubbing alcohol to the thermal compound and it didn't have much effect at all. I concluded this paste must be different from what I'm used to and reapplied the cooler.

I'm assuming that's where my problems are coming from, but HWiNFO64 is saying that my CPU is NOT thermal throttling, and the temperatures are staying below 55C (max, average ~51C). In spite of the CPU allegedly not thermal throttling, it made it all the way down to 1.18Ghz after leaving the computer on overnight.

I restarted the board and booted into Windows 10 (long story, using the computer as a temporary desktop) and the System information recognizes that I'm using a 2.4Ghz chip, but Resource Manager and HWiNFO64 both reported that I was clocked at ~2.20Ghz.

I emailed Supermicro and the tech suggested that the CPU is slowing down because I'm not doing anything. So, I installed Geekbench 4 and got this result, (1145sc, 5203mc) my CPU staying at ~51C throughout the entire benchmark and my clock speed staying at ~2.20Ghz the entire benchmark. The technician I spoke to said that's fine, the CPU should go back up to 2.4Ghz if I run something taxing. I said okay, but was left kind of miffed that a CPU benchmark didn't tax my chip enough to get it up to its rated clock speed...

So after work today, I came back to my system being at 1.18Ghz again... I ran Geekbench 4 and got this result (729sc, 3334mc). My clock speed stayed at "x12 ratio" the entire time according to HWiNFO64. The load didn't change my clock speed at all. Maybe there is something wrong with how Geekbench works?

So I tried CPUmark, and I got a score of 1762. My clock speed didn't change throughout the entire benchmark...

The only thing I can think is that this chip throttles itself around 50C and that I disturbed the thermal paste enough to make the CPU need to throttle itself. But HWiNFO64 says it's NOT thermal throttling...

Anyways, I emailed Supermicro again and hopefully I hear back from them, but I'm also looking to replace this thermal paste if I can in hopes it's my own error here. Do I need a special kind of thermal paste for this application?
 
  • Like
Reactions: Ceros_X

confusis

John Morrison. Founder and Head writer SFF.N
SFF Workshop
Editorial Staff
Moderator
Jun 19, 2015
3,541
6,094
sff.network
I wouldn't imagine so - I've replaced the paste on many an Atom unit (netbooks, laptops,. itx boards, POS units) with regular stuff and had no issues.
 
  • Like
Reactions: zovc

zovc

King of Cable Management
Original poster
Jan 5, 2017
852
602
Did rubbing alcohol take the old paste off for you? Or did you use something else?
 

zovc

King of Cable Management
Original poster
Jan 5, 2017
852
602
Maybe I just didn't give it enough time to work its magic.
 

zovc

King of Cable Management
Original poster
Jan 5, 2017
852
602
Supermicro offered to RMA the board even if its outside of its warranty, that's really nice of them. I'm gonna try to reapply thermal paste to the CPU at lunch tomorrow and if it doesn't boot up at 2.4Ghz, I'm gonna ship it off.
 

zovc

King of Cable Management
Original poster
Jan 5, 2017
852
602
Reapplied thermal paste. System booted into Windows 10 running at x24 clock multiplier. That's a good sign!

Main indicator thermal paste was the issue, to me, is that instead of constantly running at ~51C, the system booted at ~26C and only went up to 40C during benchmarking. It wasn't connected to the internet so it didn't upload its geekbench results, but I'm pretty sure that because it stayed at x24 the entire time means it would be my highest score yet.

I'm sure thermals would be better if I had a fan in there but at the moment the system is fanless.
 
  • Like
Reactions: Phuncz and confusis