Originally Published in the February issue of MILITARY AEROSPACE & ELECTRONICS

Today’s powerful server-grade microprocessors are pushing systems designers toward ever-faster data throughput, innovative power control and thermal management, industry standards like SOSA, CMOSS, and HOST.

 

All of these increasing speeds, however, mean generating a growing amount of system heat, which puts pressure on design engineers to devise innovative means of electronics cooling. “The power consumption in these systems is getting much higher — to the point where the P-zero connector at these lower voltages can only handle so much,” points out Pixus’s Moll. “Power limitation right now is an issue that is coming up, and sometimes we have to put in a special connector to get enough power on the backplane.”

 

GMS’s Ciufo characterizes the thermal management challenges that today’s embedded computing designers face. “Processors are consuming more and more power, and generating more and more heat. We have been introducing rackmount servers, starting with the Intel Xeon E5, and upgraded to the new Intel Xeon Scalable processor and second-generation Scalable processors.”

 

Yet the heat continues to rise, he says. “Intel’s latest and greatest middle-of-the-road 20- and 24-core Scalable processors consume 150 Watts per processor,” Ciufo continues. “Our systems have two to four processors, so easily it can get to 300 Watts for only two processors and 600 Watts for four processors.”

 

Still, today’s leading-edge embedded computing systems rely on more than just processors. “Add two artificial intelligence cards from nVidia is another 500 Watts — 250 Watts per processor. These systems easily are consuming in excess of 2,000 Watts and more. No longer is it trivial to air-cool rackmount servers, so we have stepped-up or game.”

 

It’s not just GMS that must step-up its game, but also every other high-performance embedded computing designer who seeks to serve this market. “Cooling is kind of the tool kit,” says David Jedynak, chief technology officer at the Curtiss-Wright Corp. Defense Solutions division in Ashburn, Va. “There are no new physics to magically cool things. At the chassis level, the thermal design can be very focused for the type of platform, but on the board there are only a few ways we can cool them.”

 

For GMS air-cooled chassis and enclosures “we have taken a page out of the VITA 48 playbook, to make sure we are doing managed air flow — essentially to make certain we can cool systems that in excess of 2,000 Watts,” Ciufo says. “We are doing computational fluid dynamics to make sure the air moves where it needs to, and moves the heat out of the back of the chassis.”

 

GMS designers are experimenting with different kinds of heat sinks that blend conduction and forced-air cooling. “We have developed more exotic materials for heat sinks,” Ciufo says. “We used a combination of active cooling in our heat sinks themselves to make sure they are as efficient as possible. We also hired new engineers to use not just metallic alloy, but also other elements to improve the amount of heat flux you can use to get out the heat. Many of our ATR chassis are conductively cooled from the card to the chassis wall, and then convection cooled along the edge of the chassis.”.

 

In this Military & Aerospace Electronics article, GMS CTO Chris Ciufo explains GMS' approach to thermal management.

 

By John Keller
Editor-in-Chief, Military & Aerospace Electronics