Welcome back to my project.
This is Eric. Now we can see microcontrollers based on RISC-V
all around us. Several companies have now started producing
and selling development boards and CHIPs, and many are looking forward to it. When I was a student, I remember studying
CISC and RISC architecture. In the case of RISC, small logic is sufficient,
but in the case of CISC, large logic is required due to the complexity of the processes that
run for each instruction. Anyway, the era of RISC-V, which is more advanced
in RISC, has arrived. The advantage of RISC-V is designed to reduce
clock cycles per instruction by simplifying the instruction set. So it operates with low power compared to
other architectures. Also, it's the use of an open-source instruction
set. This means that RISC-V is license-free and
royalty-free. So, it can be used by anyone to learn, develop,
manufacture, or sell RISC-V chips and software. Because of these advantages, many MCUs based
on RISC-V will be launched sooner or later. Okay. So far so good. What I'm curious about is this. Assuming that the same work is done at the
same clock speed, the question is whether RISC-V has better
efficiency. ESP32 is already a low-power processor and
can reduce a lot of power consumption by adjusting the clock speed or using Sleep Mode. What difference does it make in actual use? Here is my ESP32-C6. Compared to other MCUs from Espressif,
this supports 802.15.4 radio connectivity, enabling Thread and Zigbee communication. It also supports WiFi6 and allows you to make
Matter accessories. It's like an all-in-one device. Cool. In addition, if only WiFi 5GHz was supported,
it could have been a better MCU, maybe we can look forward to the next version. Anyway, first of all, this is a single-core
MCU based on RISC-V, not Xtensa. It has two USB-C ports here. It says ESP32-C6 Sample Chip Version 0.0. Perhaps this board will be upgraded in some
cases. Also, this board has a jumper on it. According to the document, it's used for current
measurement. Here's ESP32-S3 DevKit-C. This dev kit is based on Xtensa architecture
and works as a dual-core. If you are starting a project for the first
time, I recommend this S3 dev kit. Most projects can be started with this. What will be the performance and heating result
of doing the same task at the same clock speed with S3 and C6? It will add and subtract integer numbers and
double numbers a million times. As you know, the S3 can run up to 240Mhz,
but I've lowered it to 160Mhz here. So both MCUs have the same clock speed. The result is very interesting. This is the amount of time spent calculating,
so the less the better. I thought the S3 would be overwhelmingly faster
in every way. But the results, as you can see, are fast
only for real number calculations. This is a natural result. S3 supports single-precision floating-point
unit. Since the FPU can be accelerated at the hardware
level, the S3 is overwhelmingly faster in this area. However, the C6 shows faster performance at
the same clock speed for general integer calculations. That's pretty cool. This time, I measured the temperature in the
idle state. S3 is on the left, C6 is on the right. The hottest part on each board is the regulator,
so don't worry about it. In the IDLE state, the temperature of the
MCU seems to be around 20, 21 degrees Celsius. It doesn't differ very much. Let's take a look at the temperature change
when the MCU is fully loaded and working. The use of WIFI consumes a lot of power. Looking at the datasheet, using WIFI consumes
a lot of currents. The peak power consumption of C6 is higher
for the same 802.11b mode. Instead, you can see that the peak power consumption
is the smallest in AX mode. If possible, it would be best to use WIFI
in AX mode in C6. In this test, the MCU continuously transmits
1500 bytes of data to the local server through UDP. This can be a heavy task for the MCU, and
over time it will heat up itself. Let's see how far the temperature can go in
extreme operation. They are both running at the same clock speed
and both seem to start at similar temperatures. They are still working on sending data to
the server via WIFI, so it will get hotter and hotter. Currently, both are connected in 802.11n mode. Let me play this video more quickly. Let's look at current consumption for a moment. I wish I could draw a graph of power consumption,
but I don't have such a tool yet. Sorry about that. I couldn't find any significant difference
while looking at this. Both devices have similar peak power consumption
and heating is also similar. If I compare it to X86 or ARM, I think RISC-V
will show pretty good power efficiency. However, my short conclusion is that RISC-V's
MCU is not much better than general low-powerMCU. I think ESP32 is already showing good performance
while operating at low power. This is a test using two ESP32-C6,
one for a Light device and the other for a Switch device. Simply, the built-in LED of the light device
can be turned on and off via the switch device. This demo example is included in ESP-IDF version
5.1, you can learn the minimal part about connecting two Zigbees. Today's video ends here. Thank you for watching. See you on the next project.
