What are the differences between STM32L5 and U5 PKA?

[OliM]

Even though it is not really advertised as having changed a lot, it looks like the U5 increased the throughput of the PKA unit nearly 3 fold (upper L5, lower U5):

Is this really the case or did the way of obtaining those numbers change in between (e.g. by using different curves)?

I was very surprised when I saw those numbers because the marketing/summary material only mentioned hardening, not performance increase.

[Mohamed Aymen HZAMI]

Hello @Oliver Müller​ and welcome to the community,

Indeed, the version of the PKA in STM32U5 and L5 is not the same.

Mohamed Aymen

[OliM]

"Not the same" is a bit of an understatement considering from further research it also looks like the U5 variant uses two times the clockrate (no more mention about halved input clock on the core). So we are talking about less than 1/6th computation time compare to the L5.

[Mohamed Aymen HZAMI]

Hello @Oliver Müller​,

You can refer also to this migration guide

Mohamed Aymen

[OliM · Accepted solution]

It does not mention the performance differences I found on my own.

Instead it says:

"The STM32U585 and STM32L562 share almost the same PKA features but the STM32U585 embed two new features and three new computation operators. Registers are compatible except new bits added in STM32U585 to map the new features"

From a functional point of view it is quite similar, that's true. But the performance difference is big enough to have an influence on my security architecture. This might now become more secure thanks to the U5, so you as a company might want to communicate that improvement a bit more offensive.