Worth remembering also that there are many variants of the ESP32, with dual core variants with hardware floating point, and far more GPIOs than the C6, although the C6 has the most recent BlueTooth support. I can't imagine Silabs will be able to compete on price, so they will have to exceed the feature sets and power consumption at least. As for the IDE - I think this is one of Espressif's weak points. Their own IDF, and VScode extensions are a complete nightmare to get working (on Windows at least), and their internal team often break compatibility or introduce bugs in libraries, then deny they did anything, then claim people are just "using it wrong", then finally fix what they've broken. The Arduino extention (not mentioned in this video for some reason) is actually the easiest way to work with their devices. I use this with my own custom boards. Silabs also have a VSCode plugin along with Simplicity Studio, which is not mentioned in this video.
Hi. You might be right about that. Most likely they are cheaper. Its just hard to find good documentation for the Bbouffalo Lab products. Do you know where to get decent datasheets?
Hi. In some of SiLabs processors you find a co-processor which they refer to as MVP - Matrix vector processor. Using the coprocessor allows for faster matrix and vector operations. And it would off-load the main processor.
Well I am not an expert in this field but I assume its limited to the size of the model fitting in the µC. I assume most µC based ML/AI implementations will use a trained model (trained on a powerful system) that is then put into the µC for things like pattern detection /anomaly detection to do things like gesture detection and so on. If I am not mistaken simple object recognition should be doable. Could be a nice topic to dive into at some point.
What was the point of this video. There is nothing comparing these things at all. Apples and oranges. You need to figure out more than what you can see from ONE SLIDE. One slide is all you showed us and it wasn't even a good slide. This is joke garbage content. Just because you can make a video and edit it doesn't mean you SHOULD.
Silicon labs is very expensive ... 80€ EFR32xG24 dev board vs 9€ for a ESP32-C6 dev board
From the USB to serial chip in old esp8266 we already know silicon labs is a joke and worse then Chinese company
Worth remembering also that there are many variants of the ESP32, with dual core variants with hardware floating point, and far more GPIOs than the C6, although the C6 has the most recent BlueTooth support. I can't imagine Silabs will be able to compete on price, so they will have to exceed the feature sets and power consumption at least. As for the IDE - I think this is one of Espressif's weak points. Their own IDF, and VScode extensions are a complete nightmare to get working (on Windows at least), and their internal team often break compatibility or introduce bugs in libraries, then deny they did anything, then claim people are just "using it wrong", then finally fix what they've broken. The Arduino extention (not mentioned in this video for some reason) is actually the easiest way to work with their devices. I use this with my own custom boards.
Silabs also have a VSCode plugin along with Simplicity Studio, which is not mentioned in this video.
looks like esp c6 is the killer
For those who use Arduino, the latest ESP32 core now includes the C6.
Do you know who i can use zigbee with arduino ide?
bl616/618 also has 2.4ghz wifi-6. Already available and probably cheaper than Silicon labs one.
Hi. You might be right about that. Most likely they are cheaper. Its just hard to find good documentation for the Bbouffalo Lab products. Do you know where to get decent datasheets?
WTF is an AI and machine lerning accelerator? In an MCU, I mean.
Hi. In some of SiLabs processors you find a co-processor which they refer to as MVP - Matrix vector processor. Using the coprocessor allows for faster matrix and vector operations. And it would off-load the main processor.
@@TheEmbeddedDude What are the applications, though?
Face and object recognition?
Well I am not an expert in this field but I assume its limited to the size of the model fitting in the µC. I assume most µC based ML/AI implementations will use a trained model (trained on a powerful system) that is then put into the µC for things like pattern detection /anomaly detection to do things like gesture detection and so on. If I am not mistaken simple object recognition should be doable. Could be a nice topic to dive into at some point.
Futur will be RISC-V based ...
What was the point of this video. There is nothing comparing these things at all. Apples and oranges. You need to figure out more than what you can see from ONE SLIDE. One slide is all you showed us and it wasn't even a good slide. This is joke garbage content. Just because you can make a video and edit it doesn't mean you SHOULD.