ESP32-S3 vs Teensy 4.0
Xtensa LX7 (ESP32-S3) at 240 MHz against Cortex-M7 (i.MX RT1062) at 600 MHz, on real hardware, same C kernels, clocks verified.
ESP32-S3
Xtensa LX7 (ESP32-S3)
- Family
- Xtensa
- Board
- ESP32-S3 DevKit
- Clock
- 240 MHz
- float / double
- 32-bit / 64-bit
Teensy 4.0
Cortex-M7 (i.MX RT1062)
- Family
- Arm Cortex-M
- Board
- Teensy 4.0
- Clock
- 600 MHz
- float / double
- 32-bit / 64-bit
Typical gap
6.28×
Teensy 4.0 faster, geometric mean of 30 cells
Biggest win for ESP32-S3
0.32×
uint64_t divide
Biggest win for Teensy 4.0
30.4×
double divide
Clock for clock
2.51×
Teensy 4.0 needs fewer cycles per op (2.50× clock gap)
Operation by operation
Every cell
How many times faster ESP32-S3 is than Teensy 4.0, operations per second. Blue: ESP32-S3 ahead. Orange: Teensy 4.0 ahead.
| Add | Subtract | Multiply | Divide | Modulo | |
|---|---|---|---|---|---|
| uint8_t | 1/6.10× 12.7 M/s · 77.4 M/s | 1/5.93× 12.7 M/s · 75.3 M/s | 1/5.94× 12.7 M/s · 75.3 M/s | 1/4.56× 11.5 M/s · 52.3 M/s | 1/4.23× 10.9 M/s · 46.3 M/s |
| uint16_t | 1/5.93× 12.7 M/s · 75.3 M/s | 1/5.91× 12.7 M/s · 75 M/s | 1/5.40× 12.7 M/s · 68.6 M/s | 1/4.09× 10.1 M/s · 41.5 M/s | 1/3.86× 9.72 M/s · 37.5 M/s |
| uint32_t | 1/6.17× 13.4 M/s · 82.8 M/s | 1/6.39× 13.4 M/s · 85.7 M/s | 1/5.11× 13.4 M/s · 68.6 M/s | 1/3.59× 11 M/s · 39.3 M/s | 1/3.33× 10.5 M/s · 34.8 M/s |
| uint64_t | 1/8.02× 7.87 M/s · 63.2 M/s | 1/8.09× 7.81 M/s · 63.2 M/s | 1/8.03× 5.45 M/s · 43.7 M/s | 1/3.08× 1.8 M/s · 5.54 M/s | 1/3.18× 1.74 M/s · 5.54 M/s |
| float | 1/6.13× 12.7 M/s · 77.7 M/s | 1/6.13× 12.7 M/s · 77.7 M/s | 1/6.13× 12.7 M/s · 77.7 M/s | 1/7.77× 3.48 M/s · 27 M/s | 1/3.42× 1.95 M/s · 6.67 M/s |
| double | 1/18.4× 3.04 M/s · 56 M/s | 1/19.6× 2.79 M/s · 54.7 M/s | 1/21.9× 2 M/s · 43.7 M/s | 1/30.4× 516 k/s · 15.7 M/s | 1/3.96× 1.22 M/s · 4.84 M/s |