Intel is running rings around AMD and Arm at the edge
What will it take to loosen the x86 giant's edge stranglehold?
Analysis Supermicro launched a wave of edge appliances using Intel's newly refreshed Xeon-D processors last week. The launch itself was nothing to write home about, but a thought occurred: with all the hype surrounding the outer reaches of computing that we call the edge, you'd think there would be more competition from chipmakers in this arena.
So where are all the AMD and Arm-based edge appliances?
A glance through the catalogs of the major OEMs – Dell, HPE, Lenovo, Inspur, Supermicro – returned plenty of results for AMD servers, but few, if any, validated for edge deployments. In fact, Supermicro was the only one of the five vendors that even offered an AMD-based edge appliance – which used an ageing Epyc processor. Hardly a great showing from AMD. Meanwhile, just one appliance from Inspur used an Arm-based chip from Nvidia.
Intel's strength at the edge is by no means surprising. While AMD has grown by leaps and bounds in recent years and Arm processor vendors like Ampere have made inroads into the cloud and colocation datacenters – surely to the dismay of Intel execs – it's easy to forget that Intel still controls the lion's share of the market. At least, for now.
At a time when AMD holds a considerable core count and efficiency advantage in the datacenter and is competing neck and neck with Intel in the consumer space, it just feels strange that the company wouldn't have a more compelling showing at the edge, where performance per watt is the name of the game. In light of Apple's M-series chips, the same is true of Arm.
An edge case for every edge
Something to keep in mind is that the edge isn't one thing. The fact the term gets tossed around to describe just about any compute application that lives outside the cloud or private datacenter doesn't help.
There's the near edge, the far edge, the telco edge. Appliances deployed at each of these have to contend with wildly different operating conditions, power envelopes, and maintenance regimes.
For example, a stadium might rely on near-edge compute to process large volumes of streaming video and telemetry at ultra-low latency. In this scenario the application requirements are more likely to dictate whether an OEM specs an Intel, AMD, or Ampere processor. But as you move out to the far edge or telco edge, things get a bit more complicated.
Systems in these environments may be left unattended for weeks or months at a time, operate at temperature extremes, or have limited power budgets. Supermicro's new systems, unveiled this week, are made with many of these considerations in mind.
The systems are designed with IP65 dust and water resistance, can be deployed in harsh environments, and are specced to operate in temperatures as frigid as -40°C and as high as 85°C. Suffice to say these aren't your average rack-mount servers.
But none of these conditions would prevent AMD or Arm chipmakers from designing chips for these environments. The power efficiency offered by many of AMD's latest chips surely would make it a shoo-in for these kind of edge deployments. Yet they're seemingly nowhere to be found.
- Intel demands $625m in interest from Europe on overturned antitrust fine
- AMD refreshes Ryzen Embedded line with R2000 series
- AMD's AI strategy comes into view with Xilinx, GPU, software plans
- Arm, Microsoft at pains to say this CPU arch can be trusted with real server work
Have AMD's embedded processors lost their edge?
It's not like there aren't any AMD-based edge appliances – they do exist. AMD has an entire page on its site dedicated to its partners' edge appliances.
But a close inspection reveals that most of these systems are either running on consumer-grade Ryzen processors or are so woefully outdated that anyone in their right mind – and with a big enough wallet – would opt for Intel's Xeon-D processors.
AMD hasn't updated its Epyc Embedded family of processors in nearly four years. The Epyc 3251 processors powering both of Supermicro's AMD-based edge appliances are based on an ageing 14nm manufacturing process.
The chipmaker's Ryzen Embedded processor line, which targets IoT and low-power edge applications, has had an update more recently – last week in fact. But don't get too excited. The company's newly minted R2000-series processors are based on the same architecture as AMD's Epyc Embedded chips, but use a 12nm manufacturing process instead.
The edge and IoT is probably a relatively low volume market for AMD, and validation for these environments takes time and can be costly – but at this point why bother? It looks very much like AMD just had a bunch of 2000-series Ryzen processor dies sitting around, so why not deploy them? At a time when supply chain constraints mean customers will take whatever they can get their hands on, it's not a bad idea.
By comparison, Intel's Xeon-D 1700 and 2700-series processors, announced in February, offer up to 20 cores and share the same 10nm manufacturing process and architecture as its more powerful Ice Lake Xeon Scalable processors from 2021.
In AMD's defense, its chips are cheaper – at least on paper. The Epyc Embedded 3251 launched at an MSRP of around $300, while Intel's octacore D-1732TE has a recommended price of $663, according to the chipmaker's documentation. But it's hard to draw a value comparison without without pitting them against each other in real-world testing.
That's beyond the scope this piece. But suffice to say, our money would be on Intel, at least in this arena.
Arm's efficiency edge
But enough about AMD. What about Arm? Where are the Arm edge appliances?
Let's get something out of the way: Arm will absolutely dominate far-edge compute applications, and not even by a little bit. You only need to look at the performance offered by Apple's M-series processors to see why. These chips manage desktop-class performance at a fraction of the power required by their x86 contemporaries.
With that said, any application of Arm processors outside of consumer electronics and IoT is a weird subject right now. There is a whole class of industrial IoT products running on what is essentially a Raspberry Pi. But the list of Arm chips that can compete head on with Intel or AMD is much shorter.
Ampere, a front runner in the effort to bring Arm to the datacenter, has seen early success in cloud and near-edge applications. Cloudflare swapped out AMD's second-gen Epyc processors in favor of Ampere's Altra chips in its edge datacenters last year. The DNS giant, best known for periodically breaking the internet, claims the chips delivered 57 percent better performance per watt than AMD.
Another Arm vendor with its eyes on the edge is Nvidia. At Computex, the chipmaker opened its Jetson AGX Orin platform – which features an Ampere-series GPU (not to be confused with the aforementioned chipmaker of the same name) and 12 Arm Cortex-A78AE – to system integrators for edge and embedded AI use cases.
More than 30 vendors have said they plan to launch products based on the 60W single-board computer. And for customers that need to do a lot of inferencing at the edge on a limited power budget, it wouldn't be surprising to see appliances in this vein become the de facto standard.
Intel's lead will be short lived
So, to answer the question, where are all the AMD and Arm edge appliances? Well, they do exist and some of them are quite promising. But for the moment, there's a pretty wide gulf at the edge's two extremes.
At the near edge, it's arguably a toss up between the major chipmakers. Here, application requirements are going to have a bigger impact on which wins out. Meanwhile, at the farthest reaches of the edge, verging on IoT territory, the battle is far from settled.
In between these two extremes, Intel has been surprisingly successful at carving out a niche with its Xeon-D processors. For the discerning OEM looking to satisfy the lowest common denominator, Intel sure looks like the safe bet, and this is reflected in the sheer number of Xeon-based edge appliances on the market today.
How long Intel can maintain this lead remains to be seen. Arm's Neoverse N2 core architecture, unveiled last year, specifically targets edge applications. As more chipmakers embrace these core designs, we'll see a wave of low-power Arm appliances begin chipping away at Intel's edge stronghold.
Meanwhile, we could soon see a refresh to AMD's Epyc Embedded processor line. At AMD's financial analyst day earlier this month, the company teased a new Epyc processor family, called Siena, that it says is optimized for "intelligent edge and communications deployments." That sounds an awful lot like a Xeon-D competitor. ®