Slashdot Mirror
On-Chip Liquid Cooling Permits Smaller Devices With No Heatsinks Or Fans
An anonymous reader writes: DARPA-funded research into on-chip liquid cooling has resulted in a field-programmable gate array (FPGA) liquid-cooled device that can operate at 24 degrees Celsius, versus 60 degrees Celsius for an equivalent air-cooled device. The cooling fluid resides only nanometers from the heat it must address, and operates so efficiently as to offer potential to stack CPUs and GPUs using copper columns, as well as dispensing with heat-sinks and fan systems. With those components removed, the system can facilitate far more compact designs than are currently feasible.
I can hear Jony Ive yell: "thinnnnnner...!!
the article states that the tech was add to a standard chip and now the chip - can operate at 24 C vs 60 C for the air cooled.
Note it a chip that needs to operate at the lower temperature, so this should translate to a performance gain.
...there still is a form of exterior cooling, it's just now the interface between the case's liquid cooling system interfaces with the IC packaging rather than with an exterior heatsink module that's in contact with the packaging.
This is not a cooling system integrated into the chip directly without an exterior component.
I see good and bad. Good, packaging becomes smaller so the processor can fit into smaller cases, and now there's no need for all of the mounting bosses for the traditional heatsink. Bad, the interface between the cooling system and the chip will undoubtedly be more fragile than between a cooling system and a large (relatively speaking) metal heatsink, and if there's a problem in the cooling passages on the chip there is no inexpensive method to replace the cooling portion if it's clogged-up.
We'll have to see how well this operates in the wild. If a lot of cooling system pressure loss and leaking occurs where the tubing interfaces with the chip then this won't be so good. If it manages to not leak and not plug-up then this could be a nice evolutionary step.
Do not look into laser with remaining eye.
Liquid cooling, no matter how efficient, still requires you to dump the waste heat somewhere. You don't magically get to just seal up the vents in the case because "liquid!".
That said, yes, this counts as a very cool (no pun intended) step forward, and will vastly improve the number of transistors we can pack into an arbitrary sized box - But make no mistake, that "savings" comes at the cost of needing an external radiator.
TANSTAAFL.
DARPA-funded research into on-chip liquid cooling has resulted in a field-programmable gate array (FPGA) liquid-cooled device that can operate at 24 degrees Celsius, versus 60 degrees Celsius for an equivalent air-cooled device.
So do most FPGAs need an external heater to get them up to 60C before they'll operate, or don't they all work at 24C?
Or do they mean that this one WILL only REACH 24C WHILE running?
If it is liquid as mercury then it can make corrosion to the metals or semiconductors. It is a bad notice of using liquid in the system.
And the liquid can be evaporized until no liquid here.
...the heat it must address.
Except the article says it is a few hundred microns. Not much different I see here from work we did at IBM around 2001 or others have done before, even comercial ventures (Cooligy, etc.).
It seems to still require an external pump, and liquid cooling didn't seem to take off yet except among hardcore overclocking enthusiasts. It's complicated, messy, and can fail in ways that are much worse than air cooling.
And what happens if those tiny channels erode or get clogged?
Or perhaps this is supposed to be paired with an OEM system intended to be maintenance free to solve such problems?
The article unfortunately is short on useful information.
Where does the heat go? Is it magically carried away by the liquid into some sort of wormhole?
Reading that the cooling lines are only "nanometers" away from the transistors really seemed like a significant advance in combining microfluidic and microchip processing. I mean, only thin film processing generally produces submicron uniform films. However the rather short article mentions nanometers NOWHERE. Instead, it says the cooling is "just a few hundred microns away from the transistors" Considering that an entire silicon wafer is generally ~300-800microns thick, that's much less impressive.
Moreover, the issue of heat dissipation still remains. A 100W chip still needs advect 100W of heat through the cooling line, and in microfluidics the small volumes would require huge fluid velocities and high pressures to get sufficient cooling.
Article says "We have eliminated the heat sink atop the silicon die by moving liquid cooling just a few hundred microns away from the transistors,”. Aka, more than 200000 nanometers away. The summery is rather mislearing when its talking about things being on the nanometer scale here. Liquids don't work well at the nanometer scale.
Who wrote this nonsense? Nothing "addresses" heat. Americans...
Do you want Terminators? Because this is how you get Terminators.
wasnt IBM talking about this years and years ago?
http://www.cnet.com/news/ibm-to-cool-layered-chips-with-water/