feature article
Subscribe Now

AMD Rolls Out the Big Iron

Opteron Chips Provide Big Muscle for Big Embedded Systems

True to my word, I’m highlighting a handful of x86 processors this week. Well, maybe “handful” isn’t appropriate; more like a bushel full. These things are large.

AMD just rolled out a trio of new Opteron chips for embedded systems. It’s not often that you hear “Opteron” and “embedded” in the same sentence, so some clarification is in order here. (Frankly, it’s not often that you hear Opteron mentioned at all, but that’s a different problem.) These are embedded processors in the sense that a Gulfstream V is a small jet. Sure, compared to a 777 it’s a little thing, but the GV is hardly pocket-sized. Smartphone makers need read no further.

If you’re still with us, you’re probably designing or programming systems for telecommunications, networking, medical devices, building infrastructure, or other similarly important, bulky, and valuable equipment. You want a manly processor for your manly gear. No girlie processors need apply here. Everybody grunt with me.

Ready? Let’s git ’er done. The new AMD Opteron parts are server-class processors with some add-ons and take-offs for embedded designers. They’re big and fast 64-bit multicore processors with all the bells and whistles you’d expect from a server or PC chip: built-in floating point, multithreading, three-level caches, MMX, SSE, SSE4, virtualization, overclocking… the works. AMD doesn’t pack on a lot of on-chip peripherals, however, preferring instead to offer high-speed links via two DDR3 memory channels and two HyperTransport 3 links. You can attach AMD’s official support chip and get SATA, PCIe, and all the other goodies that way.

If you’re filling out your AMD roster, the new chips are called Opteron 4386, 4376 HE, and 4310 EE. They slot in between a number of existing Opteron 43xx and 33xx devices, so don’t expect any huge differences. They’re 4- and 8-core processors running at anywhere from 2.2 GHz up to 3.8 GHz. All three have AMD’s “TurboCore” technology, meaning they can overclock themselves for short periods. The boost is considerable, too. The 4376 HE part, for example, is nominally spec’d for 2.6 GHz, but it can sprint at 3.6 GHz for a few seconds at a time. That’s a 1.0 GHz boost—not inconsiderable. Unlike old-school overclocking, however, this is both supported by the chip and controlled by it. That is, the parts will refuse to boost if they sense they’re getting hot. The whole TurboCore feature is based on heat dissipation, not circuit timing paths, so it’s all a matter of balancing performance against thermals. The silicon itself would happily run at the faster speed all day long, if only we could stop converting resistance to heat. Stupid physics.

So what’s the power dissipation really like? To be honest, I don’t even bother measuring the power on high-end processors anymore. It’s not because I can’t handle bench instruments, but just because the results are so application-dependant. AMD says the TDP (thermal design power envelope) for these guys ranges from 35W to 95W. Clearly, they’re for fan-cooled systems, not sealed boxes, but if you’re making an AC-powered product those aren’t terrible numbers. But they’re just ballpark estimates; YMMV. A lot depends on how much external I/O you’re doing. Wiggling those HyperTransport or DDR3 lines expends a lot of joules. And these chips, like all modern processors, cool down quickly if they’re not doing a lot of floating-point math or vector arithmetic. Cache hit rates, program branches, and even the number of decimal places in your floating-point code can all affect their power consumption. That’s why even the EEMBC benchmarks don’t attempt to measure power numbers. (More on EEMBC next month.)

Prices are in the $350–500 range, so we’re talking big iron here. Big branding irons, since the chips require large metal heat sinks and put out a lot of heat, as we’ve mentioned. But all that is fine if you’re building big boxes and want the utmost in performance. The new AMD parts are even socket-compatible with others in the same chip family, so this represents an up- and downgrade path if you were smart enough to design your board to accommodate varying power requirements.

Yeah, they’re big processors, but they’re designed for big jobs. Running Web servers, for example, or RAID farms, or telecom racks that have to pass NEBS testing. They’re not appropriate for handheld devices or pocket gadgets, but something has to power that wireless infrastructure, and I’m glad these chips are up to the job. 

Leave a Reply

featured blogs
Oct 21, 2020
We'€™re concluding the Online Training Deep Dive blog series, which has been taking the top 15 Online Training courses among students and professors and breaking them down into their different... [[ Click on the title to access the full blog on the Cadence Community site. ...
Oct 20, 2020
In 2020, mobile traffic has skyrocketed everywhere as our planet battles a pandemic. Samtec.com saw nearly double the mobile traffic in the first two quarters than it normally sees. While these levels have dropped off from their peaks in the spring, they have not returned to ...
Oct 19, 2020
Have you ever wondered if there may another world hidden behind the facade of the one we know and love? If so, would you like to go there for a visit?...
Oct 16, 2020
[From the last episode: We put together many of the ideas we'€™ve been describing to show the basics of how in-memory compute works.] I'€™m going to take a sec for some commentary before we continue with the last few steps of in-memory compute. The whole point of this web...

featured video

Better PPA with Innovus Mixed Placer Technology – Gigaplace XL

Sponsored by Cadence Design Systems

With the increase of on-chip storage elements, it has become extremely time consuming to come up with an optimized floorplan with manual methods. Innovus Implementation’s advanced multi-objective placement technology, GigaPlace XL, provides automation to optimize at scale, concurrent placement of macros, and standard cells for multiple objectives like timing, wirelength, congestion, and power. This technology provides an innovative way to address design productivity along with design quality improvements reducing weeks of manual floorplan time down to a few hours.

Click here for more information about Innovus Implementation System

featured paper

An engineer’s guide to autonomous and collaborative industrial robots

Sponsored by Texas Instruments

As robots are becoming more commonplace in factories, it is important that they become more intelligent, autonomous, safer and efficient. All of this is enabled with precise motor control, advanced sensing technologies and processing at the edge, all with robust real-time communication. In our e-book, an engineer’s guide to industrial robots, we take an in-depth look at the key technologies used in various robotic applications.

Click here to download the e-book

Featured Chalk Talk

Single Pair Ethernet

Sponsored by Mouser Electronics and Harting

Industry 4.0 brings serious demands on communication connections. Designers need to consider interoperability, processing, analytics, EMI reduction, field rates, communication protocols and much more. In this episode of Chalk Talk, Amelia Dalton chats with Piotr Polak and McKenzie Reed of Harting about using single-pair Ethernet for Industry 4.0.

Click here for more information about HARTING T1 Industrial Single Pair Ethernet (SPE) Products