Ya Want Some More Failure On Those Chips?

Posted this pet peeve on the video, copying here: Over the last year and a half with automotive chips in short supply there have been a lot of tech-bros snarking about vehicle makers using dinosaur chips. Fran, you have pointed out exactly why OEMs use mostly 16-28nm semiconductors. Whether it is electric power steering, airbag triggers, traction control, or any of the countless safety, operational, or convenience features found on a modern car the chips face a much harsher environment and a much higher standard of reliability than any other product a consumer is likely to own. Each chip must meet ISO 26262 The testing process is arduous and costly. A few low volume OEMs did choose to simply swap in similar parts from Mouser or wherever or maybe even simply delete safety features all together (ahem, Radar. You know who). OEMs who have had the joy of engaging with the US tort system knew better. Ford, for instance, built and warehoused tens of thousands of pickups waiting for one sub-$1 current regulator for their ADAS system, even though an uncertified one may have 'worked'. TLDR; A car will outlast a laptop running in that car. A laptop (probably) won't kill you when it crashes

Bill Kerr

TTL!

Fran Blanche

transistors are getting down towards the quantum instability realm, and the fix is now redundancy. The quantum computers will be using that realm to function using error correction to continue the increase in computer power.

In future all devices must exclusively use series 4000 cmos chips, or better still, valves (tubes)!

David Peaker

Newly-designed avionics are always based on the evolution of previously proven processors and memory devices. Believe me, your smart phone has processing power thousands of times more capable than all of the computers in a 747-400. And those processors in the autopilot and sensors are trusted to literally land an aircraft by itself with zero visibility just as long as the airport has the proper ground equipment to support it. Three computers are always comparing 3 sets of sensors, and each other, constantly voting in case something deviates from normal. The "odd man out" is disregarded and the aircraft continues to autoland. The development and incredible amount of testing required, always keep avionics far behind the cutting edge of processor technology. Just because a black box can be designed with a processor many times more capable than is necessary is NOT good practice. The heat, EMI/RFI characteristics and dozens of other concerns about unproven processors makes them unattractive for new design. Also, in the three separate autopilot processors, there is typically dual channels of processing happening where one processor is running the software written in ADA and the other processor running the same program in C++. Everything is done to protect the integrity of the hardware as a failure in a critical phase of flight could be catastrophic. When a black box is being tested, it is subjected to vibration, temperatures and EMI/RFI energy that would destroy a smart phone. They were worried about this "flipped bit" problem as far back as the early 1990s since there is a whole lot less atmosphere at 35,000ft and some random energy burst from the sun will hit the aircraft harder than something at sea level. To make matters worse these days, there are a lot of recycled, or just plain counterfeit, chips out on the open market. Detection of bogus parts is taken extremely seriously and the company will only buy from suppliers who have gone through an extensive vetting process and a proven track record of not supplying questionable ICs. So, for the industries that make life-critical equipment, I can only hope all avionics manufacturers are as careful. We can't afford another Boeing "MCAS" disaster. To those that wonder what the "MCAS" system processor looks like, you won't find a picture. It's because it was software added to the existing flight control computers. It was an ill-conceived Boeing-designed software change, and not some special "box", or hardware, that was at fault. What would happen if the actual flight computer processors suffered these "flipping bit" problems? It would be disastrous. Electronics manufacturers have a lot of supply chain problems out there and OEM processors malfunctioning is simply not acceptable.

Matt Wietlispach

Crazy! AMD's Milan-X Delivers 1.5GB of L3 Cache to EPYC Servers ( https://www.youtube.com/watch?v=oUugk0INTRU ) I am waiting on a 4 Epic server Board to start Building, and I limit it too 2 Units..

Michael

This maximum density transistor problem can be a nuisance on a smart phone or server, but could be lethal if it's associated with a life-critical system like an aircraft autopilot or as part of a medical device feeding oxygen/fluids to a patient.

Matt Wietlispach