Linux 6.10 Makes AES-XTS Disk/File Encryption Much Faster For Modern Intel/AMD CPUs

[phoronix]

Phoronix: Linux 6.10 Makes AES-XTS Disk/File Encryption Much Faster For Modern Intel/AMD CPUs

The work written about one month ago on Phoronix for much faster AES-XTS on modern Intel/AMD CPUs for speeding up disk and file encryption by as much as 155% with AMD Zen 4 CPUs has been submitted for Linux 6.10! As expected, this work providing new AES-XTS implementations for modern x86_64 processors is going into Linux 6.10 as part of the crypto subsystem updates...

https://www.phoronix.com/news/Linux-6.10-Crypto

[Danny3]

At what cost?
Again the performance increase by taking shortcuts that will be discovered and mitigated in the years to come?

[Jumbotron]

Whoah, what happened to AMD Zen 2 ? The gains just dropped significantly from Zen 1 ?

[Anux]

At what cost?
Again the performance increase by taking shortcuts that will be discovered and mitigated in the years to come?

What do you mean by that? This is about making the exact same calculations just with AVX instead of SSE.

Whoah, what happened to AMD Zen 2 ? The gains just dropped significantly from Zen 1 ?

Probably Zen 2 was already well optimized.

[mathieu.cossette]

Awesome!

[Chugworth]

That's a good chart to complain about Intel's code name system. Look at AMD for example: Processor. Generation. Done.

Intel on the other hand... Which came first, Cascade Lake or Sapphire Rapids? How do you even remember all the various lakes? I don't know how many times over the years I've had to do web searches to see what processor the stupid code name correlates to.

I know AMD's is not a perfect system, as there are some cases where the chip generation doesn't neatly align with the marketing number. That sucks, but it's still easier to keep up with than remembering all the different lake names.

[chuckula]

That's a good chart to complain about Intel's code name system. Look at AMD for example: Processor. Generation. Done.

Intel on the other hand... Which came first, Cascade Lake or Sapphire Rapids? How do you even remember all the various lakes? I don't know how many times over the years I've had to do web searches to see what processor the stupid code name correlates to.

I know AMD's is not a perfect system, as there are some cases where the chip generation doesn't neatly align with the marketing number. That sucks, but it's still easier to keep up with than remembering all the different lake names.

The fact that Larabel used Intel internal codenames isn't Intel's fault. He could have just as easily used Core generation numbers and Intel's families would be just as easy to distinguish.

Conversely, if he had used AMD's internal code names of Summit Ridge, Matisse, Breckenridge, and Genoa you wouldn't be jumping up and down giving AMD hi-fives for simple and logical naming schemes either.

[Chugworth]

The fact that Larabel used Intel internal codenames isn't Intel's fault. He could have just as easily used Core generation numbers and Intel's families would be just as easy to distinguish.

Conversely, if he had used AMD's internal code names of Summit Ridge, Matisse, Breckenridge, and Genoa you wouldn't be jumping up and down giving AMD hi-fives for simple and logical naming schemes either.

Well for some reason it seems to be a common trend among tech sites to refer to the Intel processors by their code names. I don't see that as much with AMD. In fact, I hardly ever see them referred to by their code names. "10th gen Core" would make much more sense to me than whatever lake was its code name.

[fadedmaple]

The fact that Larabel used Intel internal codenames isn't Intel's fault. He could have just as easily used Core generation numbers and Intel's families would be just as easy to distinguish.

Conversely, if he had used AMD's internal code names of Summit Ridge, Matisse, Breckenridge, and Genoa you wouldn't be jumping up and down giving AMD hi-fives for simple and logical naming schemes either.

Core generation is a disaster, and the desktop and laptop can even be different architectures and instruction sets.

[streblo]

This is awesome, but for my system at least my

cryptsetup benchmark

result for aes-xts still exceeds my read and write speeds.