Hacker News new | ask | show | jobs
by tasty_freeze 720 days ago
I can understand how the material cost is low for such a tiny chip -- it said 337,000 of them fit on a wafer, but testing time must cost more than 1.5 cents per die.
3 comments

Everything is automated. Testing a simple micro with few pins can be extremely fast.

Look up some modern factory tour videos. The RGB LED factory videos are interesting to watch. You can see how the bulk of the work is completely automated. Once the machines are set up, you let them run and collect the output.

My guess is that testing complexity is a function of the number of gates, so it doesnt matter too much if they are on a big die or many small dies.

Also testing is an extremly paralell process, scaling with the number of connections you can connect to the tested wafer, and recent technology(from testing hbm ram) can really scale that:

https://www.formfactor.com/applications/high-volume-test-on-...

They probably don't test. Yield is likely high enough that failures are the responsibility of the customer.

Honestly, this is where I'd love RISC-V to be making inroads. It would be really nice to have a super simple processor with a standard architecture and tooling.

No, they _are_ tested during manufacturing.

> Honestly, this is where I'd love RISC-V to be making inroads

A 32-bit CPU is way too big for this price point still.

> A 32-bit CPU is way too big for this price point still.

Not sure what you mean by "this price point" but CH32V003 is RISC-V and around 10 cents. It is only 6 times more expensive.

Yeah, that's a savage bastard too. 48MHz, DMA, UART, I2C, SPI, a sleep mode ... 16KB flash and 2KB ram. Certainly enough to run something with a UI, probably enough to implement a basic software defined radio, could probably do some neat tricks when buried inside a USB cable.

I mean, clearly we're at the "splash the cash" end of the Chinese processor market here but at least you get good bang for buck.

With small chips like this do you think they would test at a high sampling level?
> It would be really nice to have a super simple processor with a standard architecture and tooling.

Pragmatically speaking, that architecture already exists -- it's the 8051. (This part is a PIC clone, but 8051 is also extremely common.)

RISC-V has promise for larger microcontrollers, but really tiny ones like these (48 bytes RAM, 1KB ROM) are better off with an 8-bit architecture.

Ya...8051s are literally everywhere, and there's a vast ecosystem of code and programmer talent to draw on.
Yes but... modern 8051 has grossly different peripherals / drivers than the original.

Most of your code / effort will be in getting these peripherals to work. A ton of microcontroller programming is a simple if(voltage>0.5V) then {setup peripherals to do Y}.

In fact, voltage > 0.5V is a peripheral (aka: the Analog Comparator + DAC loopback) on most uCs today. So in practice its:

"If (analog comparator) then {do Y}".

Except "if analog comparator" is itself an interrupt that you put into the interrupt table, so that the processor can sleep for minimum power usage. So really its all just configuration of interrupt table handlers, configuration of various peripherals, and then a massive amounts of sleep.

The first link in the article is to a .10 RISCV part. https://zeptobars.com/en/read/Nyquest-Technology-NY8A051H-80...

The RPI2040 is under a buck in bulk (the chip).

The Pine Ox64, The Milk-V duo and the RPI packages are all interesting products (at a sub 10 bucks each).

The MILK-V duo S is very appealing (form factor wise). If they ever mange to get POE/HDMI in version out it will be a killer product (even more so if they get all that into a sub 20 buck price point).