Ampere Altra Performance Shows It Can Compete With - Or Even Outperform - AMD EPYC & Intel Xeon

[Mike Frett]

Well I think it's awesome. You can already buy little ARM based computers for like $60 and those little Android TV boxes are neat too. Great time to be alive if you like to play with new toys.

[Volta]

They already have with M1. Wait until they release their 16-32 Core CPU for the iMac and the desktop PowerMac.

The Age of ARM is here. X86 tech is now legacy.

You can get wildly ripped off on price but get good energy efficiency with Intel.

You can get better value and bang for your computing buck with AMD but you get boned on power draw.

Or you can get all of it with ARM.

I was talking about software. When comes to hardware it's obvious ARM is kicking ass right now.

[pal666]

You can get wildly ripped off on price but get good energy efficiency with Intel.

You can get better value and bang for your computing buck with AMD but you get boned on power draw.

Or you can get all of it with ARM.

or you can stop posting bullshit under article in which most power was drawn by xeon

[pal666]

The Age of ARM is here.

arm is neither modern nor well designed. it was "cheap and dirty" design from 80s, not much newer than x86, its main feature was "low number of transistors"

[coder]

arm is neither modern nor well designed. it was "cheap and dirty" design from 80s, not much newer than x86, its main feature was "low number of transistors"

Arm has come a long way, since those days. Unlike x86, they have not retained binary compatibility that far. Fujitsu's recent ARMv8-A.2 cores even dropped compatibility with ARMv7!

Now, if you want to talk about old ISAs and saving transistors, the 8088 (used in the original IBM PC) launched in 1979 and had just 29,000 transistors!

[coder]

In order for ARM to have any future beyond mobile devices then we need a desktop ARM machine so people can learn to work with it. I don't mean RPI or cell phones either. This is where I believe Nvidia will step up and bring ARM to the desktop

You already have at least 4 options:

1. Amazon AWS Graviton2 cloud instance.
2. Qualcomm 8cx or 7c-based laptops
3. Ampere eMAG 32-core workstation: https://www.anandtech.com/show/15733...64-workstation
4. Huawei's Kunpeng 920-based 24-core desktop: https://www.notebookcheck.net/Huawei....485582.0.html

[ldesnogu]

arm is neither modern nor well designed. it was "cheap and dirty" design from 80s, not much newer than x86, its main feature was "low number of transistors"

Did you miss Arm 64-bit announcement in 2011?

[AmericanLocomotive]

He specifies the compilation options in the image captions of the plots, but some of the tests, like the TNN deep learning benchmark, do not use -march=native. This should mean that x86 is only using SSE, instead of AVX2 or AVX-512. That would put them at a huge disadvantage.

Ah, I see that now. I don't understand why Michael was so inconsistent with his compiler options. Why would march=native only be used for a couple of tests, and not all of them? From what we saw in his last compiler test there was at least 5-25% more performance to be gained by using march=znver2 or skylake over x86-64.

I get that for "general computing", having highly-tuned code compiled for a specific architecture isn't realistic. But I feel that it is something enterprise users looking to eek out every bit of performance and efficiency would do?

[s_j_newbury]

The Age of ARM is here. X86 tech is now legacy.

x86 was always legacy. It's not a co-incidence you can run 8086/8 code on modern x86 CPUs. It was always the "unique selling point", virtually every other architecture has had ABI breaks where legacy cruft was removed. While it hasn't completely held it back, it has shaped the development and the ethos of the design. Remember, even Intel wanted to leave it behind via IA64.

[Michael]

Ah, I see that now. I don't understand why Michael was so inconsistent with his compiler options. Why would march=native only be used for a couple of tests, and not all of them? From what we saw in his last compiler test there was at least 5-25% more performance to be gained by using march=znver2 or skylake over x86-64.

I get that for "general computing", having highly-tuned code compiled for a specific architecture isn't realistic. But I feel that it is something enterprise users looking to eek out every bit of performance and efficiency would do?

The same upstream compiler flags were used for this testing. For CPU reviews do normally do -march=native but at least for mcpu/mtune neoverse-n1 on Ubuntu 20.10's GCC10 been seeing differing results in some cases actually performing worse than generic aarch64 code meanwhile on other compilers seeing better results. So that is why the defaults / no override was used for this testing but as said in the article will have a compiler comparison up soon that will look at the flags and compilers across the CPUs for those interested.