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

[phoronix]

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

While the talk in recent weeks has been about the performance of Apple's M1 ARM chip and then rumors there might be a 32 core chip in the pipe, there is already something much stronger: Ampere Altra has begun shipping and its flagship 80-core SoC with up to two sockets per server can easily take on the AMD EPYC 7742 "Rome" and Intel Xeon Platinum 8280 "Cascade Lake" performance across a variety of workloads. Here is our initial look at the Ampere Altra performance on Linux in our independent performance benchmarks.

http://www.phoronix.com/vr.php?view=29773

[Volta]

This is performance and scalability Apple will never achieve.

[elatllat]

What's the pricing like though?

[pracedru]

Very impressive.
ARM is beginning to kick ass.
Can't wait to get a laptop with 256 cores
The future is great

[lejeczek]

Who made the server, make & model? Is it Gigabyte's?

[Jumbotron]

This is performance and scalability Apple will never achieve.

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.

[Jumbotron]

Very impressive.
ARM is beginning to kick ass.
Can't wait to get a laptop with 256 cores
The future is great

Amen and again I say....AMEN !!

[baryluk]

Very impressive. It lacks a bit in single threaded performance, but otherwise very usable system. Considering their previous systems were really competitive in terms of price, I think this one will be a hit for a lot of applications, especially considering it IO connectivity, and good power efficiency.

Michael, hope you can make more single threaded benchmarks. I.e. scimark 2 (C version), SPECint / SPECfp if you have access to them, etc.

[AmericanLocomotive]

Michael,

You mentioned having an "up-to-date" compiler is very important. I was wondering if you could have ran the tests with the Intel and AMD software being compiled for march=native (or at least znver2/skylake), since obviously the N1 compiler is going to be highly optimized already.

[Jumbotron]

Very impressive. It lacks a bit in single threaded performance, but otherwise very usable system. Considering their previous systems were really competitive in terms of price, I think this one will be a hit for a lot of applications, especially considering it IO connectivity, and good power efficiency.

Michael, hope you can make more single threaded benchmarks. I.e. scimark 2 (C version), SPECint / SPECfp if you have access to them, etc.

You can never ever never ever.....ever....undervalue power efficiency going forward. It is a HUGE component of the cost of doing business for the big cloud players....Google, Amazon, Facebook, Microsoft, Oracle, etc. Also for the Supercomputer world. Billions of billions of watts of power are used every day, 24/7 by all of these guys and they have to pay for it....and increasingly find other means of powering these compute monsters that don't rely on fossil fuels. That costs them 100s of millions of dollars per year. That is a direct drain on profit. ANYTHING these guys can do to decrease costs but still give them the compute power needed for the tasks at hand is job number 1.

ARM is the solution for this. And they increasingly know this.

x86 has now been shown to be a complete disaster and failure at scaling small while increasing compute power per watt.

x86 is now showing to be a failure at scaling up compute power without becoming the main cost of your power bill.

And x86 has been shown to be a failure in incorporating AI and ML into the CPU wafer itself.

ARM has conquered all three of the above metrics.

The Age of ARM is here. x86 is now legacy.