Is there any MCU that can perform AES at fill USB HS rate?

[Miyuki]

Hi folks,
I know many MCUs have AES or other cipher accelerators, but it is rather hard to find their performance.
So are there some, that can provide performance at rates at least close to USB High speed?

[brucehoult]

If I understand this page correctly, a 600 MHz CM7 can do AES256 at 384 Mbps (400 cycles per 256 bits) without any special AES instructions or accelerator at all.

https://github.com/jnk0le/cortexm-AES

@jnk0le

[std]

Hi folks,
I know many MCUs have AES or other cipher accelerators, but it is rather hard to find their performance.
So are there some, that can provide performance at rates at least close to USB High speed?

Here AES cycle accurate figures for stm32h7 and other modern STM32's
https://wiki.st.com/stm32mcu/wiki/Security:STM32H7_Series:_Cryptographic_Library_Performance

upd:
Now I took a closer look, they give numbers for the H743, but AES hardware peripheral not announced for it. The H753 has AES peripheral.
The numbers at 400Mhz rougly: 6.4Mb/s software AES-256, 48.3Mb/s hardware. USB high-speed 480Mbits (60Mb/s) is only theoretical limit. In practice, however, USB 2.0 achieve ~43Mb/s (for some good conditions, more real number ~33Mb/s). So, hardware AES of the STM32H753 @ 400Mhz will encrypt/decrypt on the fly.

[std]

Tried initializing CRYP peripheral, tried AES, encoded and decoded the data. All is OK.
The peripheral works fine on the STM32H743, although according to the ST datasheet it is not there.

Do you see the gopher?
- No
- And I don’t see. And he is.

[eliocor]

according to this post:
https://www.mikrocontroller.net/topic/459746
the following CPUs mount the same chip die: despite the different name/families, their functionalities/memory are the same!
DIE450  STM32H742A(G-I)Ix
DIE450  STM32H742B(G-I)Tx
DIE450  STM32H742I(G-I)Kx
DIE450  STM32H742I(G-I)Tx
DIE450  STM32H742V(G-I)Hx
DIE450  STM32H742V(G-I)Tx
DIE450  STM32H742X(G-I)Hx
DIE450  STM32H742Z(G-I)Tx
DIE450  STM32H743A(G-I)Ix
DIE450  STM32H743BGTx
DIE450  STM32H743BITx
DIE450  STM32H743IGKx
DIE450  STM32H743IGTx
DIE450  STM32H743IIKx
DIE450  STM32H743IITx
DIE450  STM32H743V(G-I)Hx
DIE450  STM32H743VGTx
DIE450  STM32H743VITx
DIE450  STM32H743XGHx
DIE450  STM32H743XIHx
DIE450  STM32H743ZGTx
DIE450  STM32H743ZITx
DIE450  STM32H745BGTx
DIE450  STM32H745BITx
DIE450  STM32H745IGKx
DIE450  STM32H745IGTx
DIE450  STM32H745IIKx
DIE450  STM32H745IITx
DIE450  STM32H745XGHx
DIE450  STM32H745XIHx
DIE450  STM32H745ZGTx
DIE450  STM32H745ZITx
DIE450  STM32H747A(G-I)Ix
DIE450  STM32H747BGTx
DIE450  STM32H747BITx
DIE450  STM32H747IGTx
DIE450  STM32H747IITx
DIE450  STM32H747XGHx
DIE450  STM32H747XIHx
DIE450  STM32H747ZIYx
DIE450  STM32H750IBKx
DIE450  STM32H750IBTx
DIE450  STM32H750VBTx
DIE450  STM32H750XBHx
DIE450  STM32H750ZBTx
DIE450  STM32H753AIIx
DIE450  STM32H753BITx
DIE450  STM32H753IIKx
DIE450  STM32H753IITx
DIE450  STM32H753VIHx
DIE450  STM32H753VITx
DIE450  STM32H753XIHx
DIE450  STM32H753ZITx
DIE450  STM32H755BITx
DIE450  STM32H755IIKx
DIE450  STM32H755IITx
DIE450  STM32H755XIHx
DIE450  STM32H755ZITx
DIE450  STM32H757AIIx
DIE450  STM32H757BITx
DIE450  STM32H757IITx
DIE450  STM32H757XIHx
DIE450  STM32H757ZIYx

to get the full list (attached) I went in the following directory (STM32CubeIDE) [yours can be different!]
C:\ST\STM32CubeIDE_1.13.1\STM32CubeIDE\plugins\com.st.stm32cube.common.mx_6.10.0.202311202037\db\mcu
and gave the command:

Code: [Select]

grep -o "DIE..." STM* | sed -e "s/\(.*\).xml:\(DIE.*\)/\2  \1/" | sort

[Berni]

So even tho they claim they have less Flash, RAM, peripherals..etc they are actually are actually all there if you try to use them?

I would have thought they would have some secret e-fuses they blow on the die to disable those if they use the same die pattern.

[eliocor]

to my knowledge ST does not use any method to differentiate their products (with the same die) except for a different name/description
Just as an example, during the chip shortage, I used some STM32L431 (no USB) to replace some STM32L433 (with USB) without any problem.
I have seen the same behavior with some TI H-bridges....

[tru]

..
DIE450  STM32H747A(G-I)Ix
DIE450  STM32H747BGTx
DIE450  STM32H747BITx
DIE450  STM32H747IGTx
DIE450  STM32H747IITx
DIE450  STM32H747XGHx
DIE450  STM32H747XIHx
DIE450  STM32H747ZIYx
..
DIE450  STM32H757AIIx
DIE450  STM32H757BITx
DIE450  STM32H757IITx
DIE450  STM32H757XIHx
DIE450  STM32H757ZIYx

to get the full list (attached) I went in the following directory (STM32CubeIDE) [yours can be different!]
C:\ST\STM32CubeIDE_1.13.1\STM32CubeIDE\plugins\com.st.stm32cube.common.mx_6.10.0.202311202037\db\mcu
and gave the command:

Code: [Select]

grep -o "DIE..." STM* | sed -e "s/\(.*\).xml:\(DIE.*\)/\2  \1/" | sort

Interesting, but I think the STM32H747/757 series are dual cores (M4+M7), while the others are single M7 core.  I do have a ST Nucleo 144 STM32H753ZI, hmm.

[jnk0le]

If I understand this page correctly, a 600 MHz CM7 can do AES256 at 384 Mbps (400 cycles per 256 bits) without any special AES instructions or accelerator at all.

https://github.com/jnk0le/cortexm-AES

@jnk0le

192mbps as the aes does only 16 bytes of data at a time.


CTR32 mode can go up to ~26MiB/s @ 400MHz or ~66MiB/s @ 1GHz (RT1170). Requires at least 4KiB of data to get to full speed.