What would you like to see in The Next Chip?

[chromebin]

Here's an ESP32 die shot. https://s.zeptobars.com/esp32-HD.jpg.
Just for fun, can anyone spot the two CPU's.

Just took a look. Sprite was not joking, holly s!

Maybe a higher density lithography process could be used instead of 40 nm (32, 28, 22 nm) to keep the same die size and have more RAM ...

The die was quite a surprise with the 2 cores (almost in the middle of the 2nd quadrant) taking up so little space (1%)! So instead of more RAM we can more easily have more cores..

I bet RAM usage would be significantly reduced when using the Rust programming language, combined with a run-to-completion scheduler. Such schedulers lead to significantly better code (no races guaranteed), less overhead and a single stack model (which much reduces RAM usage). Rust also brings better language abstractions and better tooling, and easily integrates with legacy C code.

Rust is available for RISC-V, of which Expressif is already a member, so it looks my main request (besides a better ADC) will be realized!!!

[ESP_Sprite]

neltnerb: I2S normally doesn't do that, but we extended our implementation of it a bit; we decided having parallel input/output paths could be quite convenient when combined with the DMA and sample rate settings I2S already had.

chromebin: Note that adding more CPUs is somewhat more involved than just copy/paste'ing the cores... if you, for example, don't do some smart tricks with the memory buses, four CPUs will be mostly waiting for the memory to be available and not be much faster than a two-core system. There are other things to consider as well, like cache consistency etc, making life slightly more difficult for a multi-core system.

[hassan789]

Hardware support to enable safe-upgrade of the bootloader. Perhaps something like a small bit of dual-bank flash.

[dave-in-nj]

I may go against the mainstream.
I would like to see that the platform make a commitment to a split. A micro-controller on one, a SBC the other.
the 32 is closer to a SBC than anything I have seen before in the micro-controller environment

I would like to something between the 8266 and the 32. Micro-controller side.
the 32 is too large and too complicated. I still have not seen anyone answer the question of the difference between Vspi and Hspi....
as you can tell, I am a veriest tyro, but I believe that 99% of the users are too.

I would like 10 or more GPIO on the 8266,
add the ultra low power engine
leave the SPI and I2C as direct, the 10 new pins to be GPIO with ADC
add the RTC

I have a choice, use the 8266 and a port expander/shift register/ADC to get the extra pins I need
use the 32 to get the pins
use an Arduino board to get the GPIO pins. (second board)

as for memory, in the micro-controller world, few people ever need more than the Arduino offers.
in the SBC world, there can never be enough flash.
just add pads. let the user add the chip(s) they want

in the microcontroller world for battery power LOWER the clock speed. the Arduino is at 16, go to 40, stop with the 3 digit speeds. it winds up like the Rpi that over heats, cannot be battery powered and burns out SD cards so cannot be relied upon for more than a toy. An Embedded Rpi is a silly concept. An Embedded Arduino is a solid platform.


In my opinion, the ARDUINO is out-dated. the 20ish I/O pins are great, but no wifi and only 10 bit ADC.
no LAN, no RTC
the 30-pin-8266 could be the next mainstream board for beginners/hobbiests.

====================

second suggestion. the WROOM32 module is fantastic, but it seems that there are so many different dev-boards that I cannot get 2 that are the same. create a form-factor that is factory authorized so the physical dimensions and the pin locations are fixed.
three versions, the dual row compact, the 30 pin I/O and the Aircraft carrier version.

There are those of us who want to create a product using a board that is guaranteed to be made for the next 3-5 years. Like Ubuntu or Apple does. there is always room for people to make their own, but an official layout would be welcome.
I do not think there are any suppliers that can say they will off the same board in 6 months.
There were at least 12 ESP8266 boards, from 8 pins on up. and they were all made by that spec. The ESP-01 is still popular !

========================
#3) add LoRa
=======================
#4) add ESP-Fast as a new standard. no need to WiFi between boards. I have projects than need up-ward of 20 boards in a greenhouse. All less than 100 meters, usually less than 20.
monitor soil, turn on valves, have touch screen. all go back to a SBC for data storage and Web portal.
on a farm, you can have a handful in each building and some further out in the fields.

[woofy!]

I still have not seen anyone answer the question of the difference between Vspi and Hspi.....

https://esp32.com/viewtopic.php?t=1948#p9227

[Agree007]

LORA on board (or chip) would be great

[andre_teprom]

An archicteture improvement, supporting Linux OS would be nice, and would become ESP32 competitive with Raspberry PI.

[mikemoy]

"What would you like to see in The Next Chip?"
A somewhat dangerous question to ask, as most people are going to think of their own goals, and try to cram every known peripheral in and the kitchen sink in as well. Some things are better left as add on components because typically thats all they do, so it will be a hardened solution.

People need to keep in mind the more things you add the more memory your going to need to support it. I feel for the price of the ESP32 its pretty dam good the way it is. Where else you going to get a 240MHZ dual core chip for that price? Let alone Wi-Fi, Bluetooth and Ethernet capability. I am amazed that they were even able to cram all that into the chip already.

From what I have read here I see that the ram take up most of the die, and most of the cost of it as well. But honestly, I would think that people would not mind paying more for more internal ram, at least had another variant of the EPS32 with much more ram so people have the choice to buy that one.

On a side note, not topic related. I am constantly puzzled why this ESP32 is not the most talked about thing everywhere. IMHO its one of the best dam chips to hit the market. Remember when Arduino first came out it was on the lips of everyone? Yet all they did was take an off the shelf Atmel part and put it on a board, and created a ridiculously simple GUI to make it child's play for people to program it.

I have used netburner products for over 10 years now. They have a one click install that installs the compilers and Eclipse all in one shot and Eclipse is all pre set up to program their devices. Just pick the device your using and click either build or upload. Its just that simple. https://www.netburner.com/

If Espressif were to take the time to do the same, I can't see how anyone would use anything else.

[Vader_Mester]

For the RAM it is worth to increase it, so you wouldn't have to buy an external chip to get a bit more.
The 2 ballances out each other.

On the other hand, selling ESP32 capable SPI ram is a good business for Espressif, so that's another reason I guess.

Anyway, I feel that RAM is what keeps the ESP down. Brutal computing capability is not enough.
It's like a laptop with an i3 CPU but an IDE HDD. Fast CPU and slow HDD results in slow overall system.

[mikemoy]

It's like a laptop with an i3 CPU but an IDE HDD. Fast CPU and slow HDD results in slow overall system.

It's like a laptop with an i3 CPU, and 384k of ram. Fast CPU and little memory results in slow overall system.