MAA Posted May 22, 2019 Share Posted May 22, 2019 Hello, why did you remove benchmark for SHA1/SHA2? this benchmark could be hardware accelerated https://en.wikipedia.org/wiki/Intel_SHA_extensions Quote Link to comment Share on other sites More sharing options...
Fiery Posted May 22, 2019 Share Posted May 22, 2019 1 hour ago, MAA said: Hello, why did you remove benchmark for SHA1/SHA2? this benchmark could be hardware accelerated https://en.wikipedia.org/wiki/Intel_SHA_extensions As a hashing algorithm, SHA1 is obsolete. It's also redundant when you have a more modern implementation of SHA. But, to make sure not to leave it behind completely, you can still run the SHA1 benchmark from the GPGPU Benchmark panel (AIDA64 / main menu / Tools / GPGPU Benchmark / double-click on the box in the x64 column, third line from the bottom). Quote Link to comment Share on other sites More sharing options...
MAA Posted May 22, 2019 Author Share Posted May 22, 2019 >> As a hashing algorithm, SHA1 is obsolete. SHA1, but not SHA2. NIST does not currently plan to withdraw SHA2 or remove it from the revised Secure Hash Standard. Quote Link to comment Share on other sites More sharing options...
Fiery Posted May 23, 2019 Share Posted May 23, 2019 On 5/22/2019 at 3:32 PM, MAA said: >> As a hashing algorithm, SHA1 is obsolete. SHA1, but not SHA2. NIST does not currently plan to withdraw SHA2 or remove it from the revised Secure Hash Standard. AIDA64 has never implemented a SHA2 benchmark... Quote Link to comment Share on other sites More sharing options...
MAA Posted January 5, 2021 Author Share Posted January 5, 2021 But why did you do SHA3 instead of SHA2? sha2 is used much more often than sha3. also sha2 is hardware accelerated in CPU. sha3 NO. Quote Link to comment Share on other sites More sharing options...
Fiery Posted January 6, 2021 Share Posted January 6, 2021 23 hours ago, MAA said: But why did you do SHA3 instead of SHA2? sha2 is used much more often than sha3. also sha2 is hardware accelerated in CPU. sha3 NO. We've picked SHA3 over SHA2 exactly because it has no specific CPU acceleration. Quote Link to comment Share on other sites More sharing options...
MAA Posted January 6, 2021 Author Share Posted January 6, 2021 Very strange explanation. Why then do you use AES-benchmark? He also uses acceleration. Quote Link to comment Share on other sites More sharing options...
Fiery Posted January 6, 2021 Share Posted January 6, 2021 1 hour ago, MAA said: Very strange explanation. Why then do you use AES-benchmark? He also uses acceleration. When we came up with the AES benchmark, hardware acceleration wasn't available for it yet. Same thing with the SHA1: hardware acceleration was introduced after we designed and introduced that benchmark method too. Now it's the same story with SHA3: currently there's no hardware acceleration (at least not in x86 CPUs), but it's coming soon. As for the longevity and relevance of the AES benchmark, we're watching the market and will make a decision on its fate when the right time comes. Quote Link to comment Share on other sites More sharing options...
MAA Posted January 6, 2021 Author Share Posted January 6, 2021 But why don't you like "hardware acceleration"? Especially since this is not real hardware acceleration, but only extension to the x86 instruction set architecture. Quote Link to comment Share on other sites More sharing options...
Fiery Posted January 7, 2021 Share Posted January 7, 2021 10 hours ago, MAA said: But why don't you like "hardware acceleration"? Especially since this is not real hardware acceleration, but only extension to the x86 instruction set architecture. We prefer measuring actual number crunching performance using classic x86, x87 and SIMD instructions. If you compare the performance jump from going classic x86/x87 to e.g. SSE2, you can see that SIMD doesn't neary bring as huge of a benefit in terms of performance delta than SHA1 or AES instructions since the latters use hardware acceleration. So to us HW AES and HW SHA instructions are not in the same basket as SIMD, and we treat those separately. And as you can see, we're not against using HW AES or HW SHA instructions when they become available on new processors, but we do not specifically design a benchmark method around the idea of it being hardware accelerated or specifically + heavily optimized. And if you look back in the history of x86 processors, you can see that we haven't immediately dropped SHA1 benchmark when it became hardware accelerated too. Quite frankly, I don't see a reason to debate heavily on this topic when the both the SHA1 and SHA3 benchmarks are still available in AIDA64, albeit SHA1 is a bit tucked away in the GPGPU Benchmark panel. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.