Live data from Hacker News

The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

anandtech.com

71–80 of 266 posts

Re: The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

#71

This line is the kicker for me: "Intel’s current Xeon offering simply isn’t competitive in any way or form at this moment in time." The rest of the paragraph twists the knife more, and declares there to be an open market on server hardware for AMD and Ampere to fight it out. With application stacks being what they are and optimizations for Xeon so deeply integrated, it's unlikely that folks running private datacenter…

that’s a statement refuted by the article in which it appears, though. The Ampere system is clearly not suited to latency-sensitive application services, unless you’re just looking to double your latency for no reason. If you looked at these specjbb results because you’re shopping for Java backend servers then you’re buying Xeon yet again.

Re: The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

#72
post #68

This line is the kicker for me: "Intel’s current Xeon offering simply isn’t competitive in any way or form at this moment in time." The rest of the paragraph twists the knife more, and declares there to be an open market on server hardware for AMD and Ampere to fight it out. With application stacks being what they are and optimizations for Xeon so deeply integrated, it's unlikely that folks running private datacenter…

I can imagine that every cloud hosting provider expands their fleet of machines, and also retires and replaces some of it. If the expansion and replacement will mostly involve AMD and ARM-based solutions, I won't be surprised. Of course Xeons are going to stay for a relatively long time, because some customers are deeply invested in them, e.g. relying on AVX512 heavily, etc. I also don't think that Intel is going to…

I think the big issues for Intel will be avoiding getting into a vicious circle of uncompetitive performance -> lower margins -> less investment in process /fabs -> uncompetitive performance

In part AMD broke out of this by using TSMC, but Intel relying on (not just using) TSMC would be a huge step.

Re: The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

#73

This line is the kicker for me: "Intel’s current Xeon offering simply isn’t competitive in any way or form at this moment in time." The rest of the paragraph twists the knife more, and declares there to be an open market on server hardware for AMD and Ampere to fight it out. With application stacks being what they are and optimizations for Xeon so deeply integrated, it's unlikely that folks running private datacenter…

> "Intel’s current Xeon offering simply isn’t competitive in any way or form at this moment in time."

What? That makes so little sense. Is this chip x86 or something? Because the architecture is a pretty big deal for me when I decide which chip to invest in.

Re: The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

#74
post #56

Earlier quoted context omitted.

but really, in what way would 32 cores become useful in a home server, except running a bunch of VMs that do nothing or donate money in terms of electricity to folding@home

Beefy CPU, couple TB of HDDs and an SSD to boot off of: (comfortably) host your own private cloud. And probably a instance for your friends & family too if you're feeling adventurous. Via authenticated Tor Hidden Services, for optimum easy-safe ratio. Now the value of a private cloud is a subjective question, true, but personal experience suggests having it right there on home lan encourages practical use, especially…

all of that could be trivially done on a raspberry pi for $30 instead of a massively overspeced ARM server for $10000

Re: The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

#75

This line is the kicker for me: "Intel’s current Xeon offering simply isn’t competitive in any way or form at this moment in time." The rest of the paragraph twists the knife more, and declares there to be an open market on server hardware for AMD and Ampere to fight it out. With application stacks being what they are and optimizations for Xeon so deeply integrated, it's unlikely that folks running private datacenter…

AMD should see the writing on the wall and start designing ARM CPUs.

I think that would be very exciting because they have the engineering expertise and talent necessary to compete against Apple's M1. I don't know of any other company that can do that (besides Intel, but you know...).

Additionally, there is a huge hole open right now for ARM on everything else that is not an Apple product. Apple normalized ARM on consumer's desktops/laptops: they brought down the "subjective" barrier against ARM, which is much harder to break sometimes than the technological one.

But ARM CPUs for everything else that is not an Apple product is up for grabs. The M1 is extremely competitive but it will never leave Apple's realm (unless Apple wants to start competing on the CPU business, but I see that going very strongly against Apple's consumer focus).

It is incredible, looking at it in retrospective, but Apple opened up (or normalized, if you will) a whole new market for ARM products.

Re: The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

#76
post #47

Earlier quoted context omitted.

but really, in what way would 32 cores become useful in a home server, except running a bunch of VMs that do nothing or donate money in terms of electricity to folding@home

Since it is ARM, running them at 0% would use little electricity :)

I assume because of leakage current the basic power usage at low-load isn't all that different for the same amount of chip real estate

Re: The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

#77

This line is the kicker for me: "Intel’s current Xeon offering simply isn’t competitive in any way or form at this moment in time." The rest of the paragraph twists the knife more, and declares there to be an open market on server hardware for AMD and Ampere to fight it out. With application stacks being what they are and optimizations for Xeon so deeply integrated, it's unlikely that folks running private datacenter…

I can't wait to see if the engineering teams from Cavium push Marvel back into the ring. Cavium was designing an Arm server to compete with Intel for years before they switched to HPC. The market is primed. Intel is in serious trouble (I know, people have been saying this since NexGen in 1995, but for real real this time!)

Re: The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

#78
post #4

I love seeing those alternatives to x86, Intel an AMD showing up. After long years of nothing exciting on the CPU side until AMD released Zen, now the M1 from Apple and this, hardware feels exciting again. I also really like the fact that Ampere cpus are not something that will only live in one ecosystem (like the graviton that will, i guess, never exists outside of AWS or the M1 outside of Apple products). I really…

I'm just blown away that Apple gets the credit for the Arm switch, but I'll take it.

Re: The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

#79
post #45

Earlier quoted context omitted.

Linux has been running on 2000+ core machines for longer than a decade (at least): https://unix.stackexchange.com/questions/4507/how-many-cores...

The last time discussions about Linux and SMP were hip, its competitors were Windows 2000 and 2003 :-D

Imagine a Beowulf cluster of these things!

Re: The Ampere Altra Review: 2x 80 Cores Arm Server Performance Monster

#80

So how are the cores interconnected? As ARM server designs love to throw out tens to hundreds of cores which sounds good on paper, but are not really useful to lots of workloads as they have a shit interconnect or lots of NUMA nodes. Interconnect is key, e.g., see the improvements between zen2 and zen3, especially for the `4 An anecdote: I got some hands on a box with a Cavium ThunderX2 96 logical core processor abou…

This is a serious problem. On-die interconnect requires insane bandwidth and lots of caches. We're talking TB/s to keep cores fed. This is significant challenge (source: I worked on Intel Xeon Phi) and at the time I worked on it in 2007 (pre-name-change), there was not a lot of empirical data, just lots of simulations, guard-bands and finger crossing.
Post reply on HN