SuperChassis 847E16: MB/CPU/HDs recommendations

[Ericloewe]

My bet is on a defective mounting solution. Might be the motherboard that has a socket that's out of tolerance (sounds unlikely, to the extent required to explain this), the heatsink that is deformed or otherwise damaged, ...

 

[Bidule0hm]

Yep, try to swap the heatsink without changing anything else to see if the problem stay on the same socket or if it follows the heatsink.

 

[Tenek]

I swapped heat-sinks. The temperature didn't follow it.
I tested with smaller amount of thermal paste. That is how it looked after the last time:

After this, I tried to apply pressure evenly during installing heat-sink to avoid missed corner.
Temperature went down to 78C on hottest core. Still too high. BTW I think I was running Prime95 for too long and so the system halted.

Like I mentioned earlier, setting fans to the full speed (I have AN-0126L4 with 7000 rpm) bring temperature down to another 4C.
I'm thinking to get a more powerful fans (noise is not a concern anymore), for example this 11000 RPM fan that is compatible with the case.
I think replacing 4 fans that sit across CPUs should do the trick and bring the temperature down to about 72C. Which should be acceptable.
Makes sense?

PS: I also could get a new mother board for testing. But only if it is really necessary....

 

[jgreco]

I think we have a winner here:

My bet is on a defective mounting solution.

I think that looks like there's some really uneven pressure. Try cleaning it all up and then reapplying with the heat sink 180' in the other direction. See if the spot moves to the other corner. If it does, get a new heatsink. If it doesn't, I'd say the socket is "bad".

 

[Bidule0hm]

I think we have a winner here:

"My bet is on a defective mounting solution."

I think that looks like there's some really uneven pressure. Try cleaning it all up and then reapplying with the heat sink 180' in the other direction. See if the spot moves to the other corner. If it does, get a new heatsink. If it doesn't, I'd say the socket is "bad".

I agree 100 % ;)

 

[Tenek]

rotating heat-sink didn't show missed corners. I'm talking to seller to see if they have another mother board they could send me for testing.
Meanwhile, I would like to share the best thermal paste application I ever seen:

 

[Robert Trevellyan]

 

[diedrichg]

ROTFF! LMFAO!

 

[jgreco]

Meanwhile, I would like to share the best thermal paste application I ever seen:

What... the... frak.

Egads. To the imbecile who did that, I'd like to point out that maybe this could have been excusable in the 1990's, but this is frickin' 2016 and if you're not certain what you're doing, there's probably about a thousand YouTube videos that will show you varying levels of competence but all of which would stop you from such an ... idiotic ... mistake.

 

[Bidule0hm]

Looks like a 90's socket however...

 

[Mirfster]

Would be cool if the system still booted...

 

[DJ9]

Looks like they used the entire tube.

Maybe they have really large peas where they are from. ;)

 

[Ericloewe]

Looks like a 90's socket however...

It says Socket 467, which is a Pentium 4 socket. Pretty much the only thing that'll fit in there, I believe, since later Pentium 4s moved to LGA775 and Pentium 3s were mounted on daughterboards.

 

[Tenek]

I bought server from these guys.
They said that they get 80 Degrees as well. Not sure, either all their servers are too hot or it is normal for this particular configuration. TCase for my CPU is 80C. Also I read somewhere that TCase temp and core temp are the different things. Dunno.

 

[Tenek]

I ran Intel® Processor Diagnostic Tool and both CPUs passes validation.
I'm not sure if running this tool made a difference, but I tried to run Prime95 again and temperature averaged at 72C (ran it over 30 minutes. It spiked to 79C once tho).
Although before I noticed that CPU were pushed to 114W and this time it was about 100W only. I ran the same tests as before.
Anyway I feel much better now about my case. Sounds like I should be fine.

 

[diedrichg]

Although before I noticed that CPU were pushed to 114W and this time it was about 100W only. I ran the same tests as before.

So you found out that it was overclocked?

 

[Ericloewe]

So you found out that it was overclocked?

You can't overclock Xeons, so that seems unlikely.

It does bring up the very important matter of how much power these cooling solutions can handle. Some cooling solutions might only be rated for the lower-power Xeon E5s, whereas the huge god-knows-how-many core beasts go up to 165W, IIRC,

 

[jgreco]

Looks like they used the entire tube.

Maybe they have really large peas where they are from. ;)

That's not a pea, that's a green apple.

 

[Tenek]

I would like to thank everyone for the support. Thanks a lot guys!

 

[Tenek]

I'm building a server closet at the moment. It will get portable AC inside. Also I was going to have a fans connected to thermostat as redundant cooling solution.
But, I recently came over that extension unit for my UPS: APC AP9631 UPS Network Management Card 2 with Environmental Monitoring
Per this thread it sounds possible to automatically power down server after temperature exceeds maximum. I'm thinking to get this unit instead installing fans and thermostat.
Sounds like a workable solution to keep server safe?

UPS sits at the very bottom in rack right under the main server. So temperature readings should be pretty close to the server environment.