Alibaba Reports Their XT910 RISC-V Core To Be Faster Than An Arm Cortex-A73

[cipri]

As much as I like seeing RISC-V progressing - Keep your fingers off of AliBaba. You should not trust Chinese companies.

Really? Check this wikipedia page:

https://en.wikipedia.org/wiki/Crypto_AG

[kuco]

As much as I like seeing RISC-V progressing - Keep your fingers off of AliBaba. You should not trust Chinese companies.

true. same for american companies. if you care about data security, dont trust american and chinese companies.

[CochainComplex]

FPGA X910 already deployed in Alibaba cloud

but this is like the amazon chip. It is used by them. What I'm looking for is a device which I can use at home (not remote).

[starshipeleven]

true. same for american companies. if you care about data security, dont trust american and chinese companies.

I like how you kids assume that companies of any other nationality are somehow better

[kuco]

I like how you kids assume that companies of any other nationality are somehow better

I dont

[PerformanceExpert]

Thanks for citing the actual paper!

FWIW I think that their "register + register addressing mode" is a good idea for the instruction set ...

But yes, they are absolutely doing nonstandard things to reach the perf of Cortex-A73, whereas I was hoping that the speed would be coming from the micro architecture ...

And oh yes, there's a reason why benchmarks such as SPEC get updated (they use SPEC 2006). Not as much because they are not as representative, but because people (compiler writers, chip designers, etc) get better and better at cheating them .

Yes register indexing and other powerful addressing modes are useful, and many RISC-V designs appear to add something similar. It's really weird this is still not standardized...

If you use the same compiler and options then you can make a fair comparison. I agree SPEC2017 would be more recent but at least they show SPEC results which is far better than quoting CoreMark or Dhrystone results!

[vladpetric]

Yes register indexing and other powerful addressing modes are useful, and many RISC-V designs appear to add something similar. It's really weird this is still not standardized...

If you use the same compiler and options then you can make a fair comparison. I agree SPEC2017 would be more recent but at least they show SPEC results which is far better than quoting CoreMark or Dhrystone results!

As an aside, I don't think ISCA would let you publish a paper with Dhrystone/CoreMark numbers ... Though to be clear, the selection and reviewing process is hardly perfect.

[PerformanceExpert]

To me, it seems the second Iteration of this processor, could be a ARM killer..

Even if the next generation beats Cortex-A73, remember that is a 4 year old design. Recent Arm cores are more than twice as fast, and improving yearly, so it would be difficult to catch up.

[mbello]

The success story of x86 platform is all about availability, affordability and openness (although in 1980 being open had a different meaning).

These 3 success factors were important in 1980 and are even more important in 2020 with the rise of Linux and OSS.

If Alibaba does not focus on making this arch available and affordable, it will not grow past a niche play.

[t.s.]

Chinese companies are not known for "giving back" or being team players in any sense of the word. Seeing as its a server chip, I doubt any consumer hardware will ever come of it.

Same with western companies. Although, yes, Chinese companies is in another level.