Mandelbrot benchmark

Page last updated 26 Dec 2015, by Jack Berkhout. 2 replies

Dhrystone Benchmark Program C/1 12/01/84

1. DISCO-F746NG       (STM32F746NG ARM Cortex-M7 MCU with FPU @ 216MHz)     :1243666 dhrystones/second
2. ST Nucleo F446RE   (STM32F446RET6 ARM Cortex-M4 MCU with FPU @ 180MHz)   : 737467 dhrystones/second
3. NXP LPCXpresso4337 (LPC4337 ARM Cortex-M4 MCU with FPU @ 204MHz)         : 725736 dhrystones/second
4. DISCO-F469NI       (STM32F469NIH6 ARM Cortex-M7 MCU with FPU @ 180MHz)   : 670980 dhrystones/second
5. Seeed Arch Max     (STM32F407VET6 ARM Cortex-M4 MCU with FPU @ 168MHz)   : 667266 dhrystones/second
6. ST Nucleo F411RE   (STM32F411RET6 ARM Cortex-M4 MCU with FPU @ 100MHz)   : 399424 dhrystones/second
7. ST Nucleo F401RE   (STM32F411RET6 ARM Cortex-M4 MCU with FPU @ 84MHz)    : 349667 dhrystones/second
8. Seeeduino-Arch-Pro (LPC1768 ARM Cortex-M3 MCU @ 96MHz)                   : 325093 dhrystones/second

PUB_RA8875_Mandelbrot @ 800x480

1. Seeed Arch Max     (STM32F407VET6 ARM Cortex-M4 MCU with FPU @ 168MHz)   : 15.708 Sec
2. NXP LPCXpresso4337 (LPC4337 ARM Cortex-M4 MCU with FPU @ 204MHz)         : 16.121 Sec
3. DISCO-F746NG       (STM32F746NG ARM Cortex-M7 MCU with FPU @ 216MHz)     : 33.017 Sec
4. ST Nucleo F446RE   (STM32F446RET6 ARM Cortex-M4 MCU with FPU @ 180MHz)   : 35.672 Sec
5. DISCO-F469NI       (STM32F469NIH6 ARM Cortex-M7 MCU with FPU @ 180MHz)   : 36.895 Sec
6. ST Nucleo F411RE   (STM32F411RET6 ARM Cortex-M4 MCU with FPU @ 100MHz)   : 60.700 Sec
7. Seeeduino-Arch-Pro (LPC1768 ARM Cortex-M3 MCU @ 96MHz)                   : 65.132 Sec
8. ST Nucleo F401RE   (STM32F411RET6 ARM Cortex-M4 MCU with FPU @ 84MHz)    : 71.478 Sec

A display from buydisplay.com was used (7 inch LCD Module wOptional Capacitive Touch Screen Panel,I2C_SPI), except for the DISCO boards, which have their own graphical displays. This may influence the comparison, as plotting pixels might take different efforts.

2 comments on Mandelbrot benchmark:

Ken Boak
21 Nov 2015

Jack,

That's unusual.

The STM32F746 Cortex M7 should whip the '407 by a factor of at least two. There must be something else going on - definitely worth further investigation.

Ken

Jack Berkhout
26 Dec 2015

Hi Ken, I was wondering about this as well, but can't find a clue, yet. I have added Drystones benchmarks. Regards, Jack.

Please log in to post comments.