X10SDV Thermal Paste

[crayonshin]

Hello all! I've recently put together a SuperMicro X10SDV-6C+-TLN4F inside a CSE-721TQ-250B Super-chassis. It's a great system but a bit loud for my taste. I just saw Patrick's article on quieting it down using a Noctua CPU fan so I am going to give that a try. As long as I'm taking the motherboard out, I thought I may as well see if I can take off the heatsink and replace the factory thermal paste in the hopes of lowering the temps a bit.

Has anybody tried this on their X10SDV board? I just wanted to make sure the heatsink is actually removable and uses standard paste and is not using a thermal pad or is permanently attached somehow. Or should I just leave it alone?

Thanks!

 

[pricklypunter]

I would leave well alone unless it's obviously not making good thermal contact. The factory fit should already be optimal

 

[crayonshin]

Sound advice for sure, thanks! Still, I'm curious to see if anybody else has actually tried this. You know, just in case I get the urge to break something in my lab down the road.

 

[TeeJayHoward]

Different brands of thermal pastes conduct heat differently. In the real world, you may see a 1-2*C drop in temperature going from generic to high-end thermal paste if you operate under the assumption that the initial HSF combo was installed correctly. Most modern processors have thermal paste or even solder between the processor die and the IHS. This, combined with the superior factory solutions available, has eliminated the good-old-days of seeing 5-10* drops by switching to a different paste. If you're feeling super adventurous, you can razor off the integrated heat spreader and connect the heatsink/fan directly to the die. You run a high risk of ruining the CPU, chopping up your fingers, and the end result not working right due to a poor fit. I wouldn't recommend it, but you might see as high as 5*C temperature difference.

Back in the real world, there's no real advantage to swapping CPU paste. And it appears that the heatsink is removable from the picture found on google:

Edit: Looks like the picture won't load. (Because it contains the word "naked" in the URL maybe?) Here's the link:

http://images.anandtech.com/doci/9185/XeonD_naked.png

 

[Evan]

Not the kind of surgery I would feel comfortable with on a brand new $900 board

 

[pgh5278]

Interesting the CPU is marked Intel confidential on a commercial product?

 

[Evan]

Interesting the CPU is marked Intel confidential on a commercial product?

Probably an ES chip, also why somebody was willing to pull or apart.

 

[crayonshin]

I found that same picture shortly after I started this thread. I did note the "confidential" on the CPU but the important thing to me was that the CPU looked liked it had a nice thin, even layer of paste applied before the heatsink was removed. I definitely agree with TeeJayHoward that it's not worth the risk and likely will not provide any benefit. I won't be trying it. I just thought the idle temperature, which which is in the mid 40's C seemed a bit high for a 35w TDP part. I guess I'm too used to CPUs with huge heatsink whereas this is an SoC with other stuff packed into the package using a tiny heatsink.

 

[turskey]

Hello all! I've recently put together a SuperMicro X10SDV-6C+-TLN4F inside a CSE-721TQ-250B Super-chassis. Thanks!

This is slightly off topic, but I was wondering if it might be possible to know the clearance between the motherboard and the drive cage in that chassis? What is the maximum height of CPU cooler that could go in it? I am thinking of getting this same chassis.

 

[wazoo42]

I have two of the D-1540 models and one D-1520. I pulled the HSF off of one of the D-1540 models and replaced the TIM with AS5. I talked with tech support about cooling and taking off the HSF and they recommended I not do that. I did not have any problems, though it was stuck on the CPU rather strongly. I ended up twisting while pulling to get the TIM to let go.

I did not see any sort of change in temps (a few degrees worse). I then lapped the HSF base b/c it was bowed and had a fairly rough surface (progression from 400 grit to 3000 grit). This was still worse than the stock. I then replaced the fan with an 80 mm Sunon model, which lowered the full-load temps a few degrees below the starting temps. Lastly, I purchased some custom heatsinks, which work well in the one test, lowering my full-load temps from 60-65 C to 48-51 C.

 

[PigLover]

@wazoo42, do you have a link or other info on the custom heatsinks you used? I'm now running 5 Supermciro Xeon-D systems and the temps on all of them are horrible. I've gotten them to tolerable levels with some simple fan ducting to increase airflow, but they still see crazy teams for such a little board. And as "summertime ambient" temperatures take over my unconditioned lab space it's a bit worrisome.

 

[wazoo42]

@PigLover, the heatsinks were manufactured by Alpha Novatech. Here is a link to a similar article, and I posted pics of the heatsinks and a longer discussion of the results down in the comments. I have an extra 6 of these if you are interested (buying just a few is rather pricey, so I opted to get extra). Similar to the article, I plan to make a fan adapter that clips on to the heatsink, but that is still 2-4 weeks away.

CPU and chassis fan replacements can reduce Supermicro SYS-5028D-TN4T mini-tower noise and increase M.2 SSD cooling | TinkerTry IT @ Home

 

[max88]

I had to remove heatsink to remove screws that holds fan bracket, in order to put on a Noctua CPU fan in Patrick's article. Arctic MX-2 was the first replacement thermal paste. There is not much difference. Then tried Arctic Silver 5, which seems to be 1C better, could be just wrong, as ambient temperature may have changed between readings (the box is beside an open window).

Stock fan cools much better, full load CPU temperature is at 71C, while Noctua fan puts in at 83C. That's quite a difference. Idle CPU temp is at 40C with stock fan, while at 48C with Noctua.

After playing around with, and finally removed, air flow deflector that comes with SYS-5028D-TN4T, and a second Noctua fan pointing towards the RAM slots, full load CPU temperature stays at 75C, and idle temp stays at 45C. With much lower noise!

Fan mode was set to standard in all cases.

 

[max88]

@PigLover, the heatsinks were manufactured by Alpha Novatech. Here is a link to a similar article, and I posted pics of the heatsinks and a longer discussion of the results down in the comments.

Is this a straight replacement? What's the part number?