Replacing Silicon With Gallium Nitride In Chips Could Reduce Energy Use By 20%
Mickeycaskill writes: Cambridge Electronics Inc (CEI), formed of researchers from the Massachusetts Institute of Technology (MIT), claim semiconductors made of gallium nitride (GaN) could reduce the power consumption of data centers and consumer electronics by 20 percent by 2025. CEI has revealed a range of GaN transistors and power electronic circuits that have just one tenth of the resistance of silicon, resulting in much higher energy efficiency. The company claims to have overcome previous barriers to adoption such as safety concerns and expense through new manufacturing techniques. "Basically, we are fabricating our advanced GaN transistors and circuits in conventional silicon foundries, at the cost of silicon. The cost is the same, but the performance of the new devices is 100 times better," Cambridge Electronics researcher Bin Lu said.
What's Gallium Nitrade?
This suggests less heat which means that chips could be packed in more densely, need smaller heat sinks. This could also remove some of the heat issues associated with multi layer chips.
A win in many ways!
I remember back in the 80s that light meters in cameras used to use Silicon (SPD - Silicon Photo Diode), but then they all started using Gallium Arsenide (GASP - Gallium Arsenide Photo Diode), as it reacted faster (presumably because of the lower resistance).
There was even talk back then about making Gallium based semi-conductors, for the same reason.
Good to see it coming to fruition
"She's furniture with a pulse"
Efficient Power Conversion (EPC)
GaN Systems
Transphorm
Panasonic
Infineon
Disclaimer: I work for one of the listed companies. We welcome new members to the GaN club!
I apologize to the ones that I missed.
It's also used in LED's, plus military applications like active phased array radar systems. This breakthrough will make the LED market cost plummet, plus bring the modern radar systems cost down even lower to where other gov't agencies like NOAA and even upper crust civilian markets to own the radar for their own uses.
First rule of holes; When in one, stop digging.
Dihydrogen Monoxide kills hundreds of times more people every year than Sodium Chloride does... If you die from Sodium Chloride, to the police categorize it as assault?
I've abandoned my search for truth; now I'm just looking for some useful delusions.
TFA is a bit light on details, but (having heard of GaN before), it is good at handling large voltages/currents, and they are probably talking about more efficient power supplies (saving 20%, apparently), not replacing Si in logic chips. Or maybe integrating power conversion onto processor die itself, but the latter is still made of good old CMOS. Currently, from what I've heard, a good chunk of pins on your processor are used to supply power -- if you think of it, 30W processor with 3V bias needs to get 10A of current.
Paul B.
GaAs and GaN have some real advantages over silicon, but price and density has not been their strong suit. Silicon is cheap per unit of area, and is compatible with copper metallization. GaAs in particular has always been sensitive to copper contamination, hence the use of gold for most of the interconnects. GaN has very fast switching speeds while handling an order of magnitude more voltage than GaAs (~10x the breakdown for the same Ft). So for a ~100 GHz Ft silicon can handle just a volt or so (think 65 nm nodes and smaller to get those speeds), GaAs can handle about ~5V, and GaN is more like 40V.
The real story here is the ability to run GaN with existing silicon equipment with silicon type costs, rather than the relatively low density fab equipment GaAs and GaN are usually processed on (usually the repurposed cast-offs from old silicon fabs).
GaN by its nature operates at high voltages, like 15V minimum, so don't expect a GaN based processor any time in the near future. Instead it sounds like they are claiming a 100x improvement in the Ron*Coff figure of merit for power switching devices. The payoff would be smaller and more efficient power electronics, such as the mentioned laptop supply brick, electric car power electronics, etc. Don't expect this has any bearing on the plateau of Moore's Law.
Are you sure? http://www.digikey.com/product...
Switching power supply efficiencies are typically in the range of 60-80% depending on load, configuration etc. Typically, slightly more than half of the "wasted" power is in the switches, and about half of that is switching the gates themselves if you're trying to go much above 5 MHz (as you go higher and higher frequency, you burn more and more energy charging/discharging the gate of the switch and your efficiency drops, which is why you don't typically see non-integrated switching power supplies above 5-10 MHz). So even best case, completely eliminating the resistance of the switch will only buy you a 20% savings in energy, and even then that's assuming a power supply that's not well tuned to its load.
GaN is neat stuff, but gallium is nasty and can be both dangerous and touchy to process. It's been "the next new thing" for switching power supplies now for at least 5 years, but this is the first time I've seen someone actually announce general purpose devices using it.