4 ms·
In the long run we'll only use FPGAs and soft cores. In 2050 if your old 2035 Ford needs a new ECU you'll just stuck a somewhat newer and larger FPGA on it, an
by VLM 5y ago
In the long run we'll only use FPGAs and soft cores.
In 2050 if your old 2035 Ford needs a new ECU you'll just stuck a somewhat newer and larger FPGA on it, and if the 2035 ECU needs three hardware I2C bus with clock stretching, one of which bus has to do 10 bit I2C addrs and it also needs three hardware PWM pins with 11 bit resolution and two 8051 cores running at 5.25 MHz each and two CANBUS, you (or more likely Ford) will just compile the brand new FPGA to have exactly and precisely all that stuff and it'll work fine.
- noir_lord 5y agoI think you are probably right, economies of scale mean that using a ridiculously powerful modern process to emulate an old processor often makes sense. The C64 mini runs a modern(ish) arm processor and an emulator to pretend to be a C64 - that's a machine that is literal multiple orders of magnitude more powerful than the original system.
- bsder 5y ago> In the long run we'll only use FPGAs and soft cores. People have been saying this for decades and it is no truer today than it was then. FPGAs and soft cores lose on cost and power--which are both ferociously tracked engineering goals when you have real volumes. To a first and second order, if all FPGAs suddenly disappeared, nobody would care. Networking and test equipment would get somewhat more expensive and double their lead times. Nobody else would notice.
- pas 5y agoIt could make sense if longevity (serviceability, "right to repair") were a design goal.
- kazinator 5y agoWould a soft core which is 10 times smaller than the original chip being emulated always lose on power? What if it's clocked down so that the emulated chip doesn't run much faster than the original? In automotive, is power so ferociously tracked? Most of the energy in a car is to get it moving, and even running the cabin fan requires more wattage than a reasonable control system.
- bsder 5y ago> Would a soft core which is 10 times smaller than the original chip being emulated always lose on power? If that were even possible (which it probably isn't--the fundamental "building block" of an FPGA is easily 10x the size of a similar fully custom gate), the soft core would still likely lose. Power in CMOS is proportional to C Vdd^2 f. Vdd hasn't moved in a while--so you've lost the biggest hammer in terms of power consumption. And C goes up when you change nodes--not down. And we're holding f constant according to your assumptions. Furthermore, FPGAs are general purpose so they need things like clock spines that are optimized to work at their max frequency rather than at whatever frequency you chose. You burn a remarkable amount of power and area so that the chips can operate at say 400MHz even if you never run them faster than 8Mhz. > In automotive, is power so ferociously tracked? Not as such--automotive is generally more interested in whether your chip can survive power spikes. However, automotive is super sensitive to cost which is all about chip area. > Most of the energy in a car is to get it moving While true of ICE cars, that is less true than you might think with electric cars and regenerative braking. I can go for a very long time on 20mph roads between charges but the moment I get on a 65mph freeway, my battery gets sucked dry. I think atmospheric drag is something like to the third power of velocity--it's a really huge increase when you go even slightly faster. However, as you point out, climate control is the real killer of energy--running AC or heating is power hungry. In addition, the self-driving car folks discovered that running all that vision and radar and then processing it sucked down a remarkable amount of juice. Computes and sensors aren't exactly free.