Mini Thermal Analysis of Avalon

[Aseras]

I said I would do this when my units came in.

Here they are. This is what they look like sitting on my lab bench.
The fans are running. When everything is ok, it's blue. When something is wrong, its red.


You can notice that they are notched to allow stacking with each other. A very nice touch. Everyone should have more than one!

Thermal Imaging. All temperatures are Celsius.
It should be noted that I put the top back on after each image for a while to let the units try to be operated as briefly as possible with the cover removed so as not to skew the results.

Here's the same view in IR



Here's the top unit with the cover removed



The next two images are of the card next to the powersupply. The other two cards are hard to images because of how close they are.




Another view from the front with the top removed.


A closeup on the "Hot" part of the 3rd module


A view from the back side of the case looking in.


Here's the money shot. You can see the temperature gradient along the heatsink The front is cool, the back is warmer. The further from the intake the warmer the cards.
It's also worth noting that all 3 cards have similar temperatures. Even the 3rd card that's exposed it only slightly hotter in the bottom corner.
The 1st card wedged up by the powersupply is also cooled very well.


and lastly, just for fun

This is just a quick review. I'll go much more in depth later. I'm getting ready for some heavy travel shortly. At some point I'd like to do a full tear down and take some xray of the boards, when scheduling and equipment permits.

Thanks to team Avalon for the new toys!

[miter_myles]

That's pretty cool!   

[farproc]

Thanks a lot.

[Sitarow]

Nice review, Looking forward to seeing the X-ray images.

[cchan]

really cool!

[goxed]

hawt

[dogie]

You're not using a thermal camera right if you've not taken it to the bathroom yet.

Priorities man, priorities.

[Entropy-uc]

Fascinating stuff.

I would change red is wrong, to red is hot!

58 C is high but could be just fine.  What matters is T_j the temperature at the transistor junction.  This needs to stay below the design temperature or you will have reliability problems in the long term.  I'm not sure what TSMC's T_j specification is for their 110 nm process, but it is probably around 75 C.

You can calculate T_j from what you have there with the thermal resistance R_jc and the power being dissapated in the device.  JC being junction to case.  It should be possible to look up an Rjc value for the heat sink used here.

Clearly there is some room for optimization though.

[S M I L Y]

Very cool...  I mean Hot!

[Dalkore]

Very helpful.  Thank You.

[Wayne_Chang]

Interesting.
As graphic chip normally works under 80+ degrees, I think this level temperature is nothing to worry about.

[uuidman]

Nice pictures, this confirmes that the design with many small chips and a big passive heatsink was wise. But wouldnt a push-pull work even better ?
Now the positive preasure pushes warm air out of the box, some front fans working pushing air out, would suck in cold air through the holes in the box also ?

[Aseras]

Fascinating stuff.

I would change red is wrong, to red is hot!

58 C is high but could be just fine.  What matters is T_j the temperature at the transistor junction.  This needs to stay below the design temperature or you will have reliability problems in the long term.  I'm not sure what TSMC's T_j specification is for their 110 nm process, but it is probably around 75 C.

You can calculate T_j from what you have there with the thermal resistance R_jc and the power being dissapated in the device.  JC being junction to case.  It should be possible to look up an Rjc value for the heat sink used here.

Clearly there is some room for optimization though.

red means something. Not really sure, lol. It may have been nice to get a manual and some technical documents ( hint Yifu and Dr Zhang ) when BFGminer flash bricked unit one it was red the whole time until i put it in recovery.

[Aseras]

Nice pictures, this confirmes that the design with many small chips and a big passive heatsink was wise. But wouldnt a push-pull work even better ?
Now the positive preasure pushes warm air out of the box, some front fans working pushing air out, would suck in cold air through the holes in the box also ?

for the heatsink units yes, but Dr Zhang says this will cause the powersupply to overheat.

guess what i'm going to test soon?

[Big Time Coin]

thanks for the pics -looks like they are well designed with very good heat distribution and dissipation.