CC Blog Newsletter Sponsored

Q & A: The Future of Embedded Computing

Written by Dan Demers

Q&A with Dan Demers, congatec

Q: The embedded sector is demanding more and more server power today. In the past, the emphasis was on low power. Where does this sudden demand for high performance come from?

A: The keywords here are edge servers and fog computing. Digitalization means that the volumes of data to be processed and the real-time requirements are constantly increasing. And since these data need processing at the edge, this is where edge data centers are deployed. They must not only function in particularly harsh operating conditions—they also have to keep running for many years, just like industrial machines. This applies to server technology for the process, energy, railway and transportation sectors. Autonomous vehicles in logistics or agriculture need massive mobile computing power, for example. In automation, collaborative robotics and entire networked manufacturing cells need higher performance. All vision-based inspection, diagnostics imaging and automation systems also require more and more performance at ever higher resolutions. And all of this is coupled with artificial intelligence (AI) and big data analytics at the edge. In addition, more functions are being consolidated onto one hardware platform by leveraging virtualization with real-time hypervisors like the one from Real-Time Systems. This reduces system costs and simultaneously increases reliability and resilience compared to multiple distributed systems. Today, systems are also expected to interact directly with enterprise applications. This can only be achieved with high-performance, high-bandwidth edge servers.

Q: COM-HPC is designed to provide sufficient computing power for these edge and server applications. What products will you be focusing on in the near future?

A: We have launched brand new COM-HPC Client modules with 11th generation Intel Core processors. They come in two variants, with Tiger Lake U  and Tiger Lake H . Tiger Lake H modules offer Intel Core processors with up to 4 cores and a TDP of 15W that can be scaled up to 28W. They are the first embedded processors with PCI Express Gen 4, i.e., the standard for ultra-fast peripheral connectivity, which was one of the reasons why a new form factor for Computer-on-Modules was needed. Its high performance is further enhanced by the new Intel Xeon, Core and Celeron processors of the Tiger Lake H generation, which offer almost twice as many interfaces as COM Express with the same processors. For me, these are the first processors to really make COM-HPC fun. Yet with up to 45W TDP and 20 PCIe Gen lanes, they mark only the beginning of the performance potential of COM-HPC designs. By the way, we offer both series in the extended temperature range of 0 to 60°C and the industrial range of -40 to +85°C. The first congatec COM-HPC Server modules will be ready for launch in Q1/2022, and we’ll be stepping on the gas even more with these. I can’t, or rather I’m not allowed to say, which processors they will support as the roadmaps are still secret. Since Intel is heavily involved in the COM-HPC working group, it’s a likely assumption that we will see relevant processor launches in the not-too-distant future. But AMD processors are also highly interesting for this environment.

The congatec portfolio already includes two COM-HPC Client Computer-on-Modules. They are based on the Intel Tiger Lake processors.

Q: The new modules also require new board designs. Which are the most important features for system and board design in your view today?

A: Key requirements for today’s high-performance, high-bandwidth industrial board and system designs are real-time performance; real-time connectivity with time sensitive networking support; and reliability. There’s also the need to integrate image and speech recognition into these systems and to make them more interactive using AI. This calls for complete embedded vision kits, which enable developers to start designs immediately without first having to connect boards and cameras to verify whether everything works or not. Especially with regards to AI, I expect to see significant performance improvements in the next few years. Remote management features are also important; examples include Intel AMT or the new COM-HPC Platform Management Interface (PMI) specification, which implements IPMI and Redfish.

Because of the many high-speed interfaces, the quality of the board layout also matters. Even minor layout errors can disrupt data transmission, which in turn can lead to dramatic errors in the final application. The right thermal design is also extremely important to ensure that systems deliver continuous performance, and that processors and other components don’t overheat. When the processor has to reduce the clock speed because of overheating, this has a massive impact on performance as well as reliability. As a rule of thumb, a temperature increase of 5-10°C can halve the chip lifetime.

Implementing new high-speed interfaces such as PCIe Gen 4 or 10 GbE is far from trivial, requiring great board layout expertise. Application-ready starter sets simplify the implementation immensely.

Q: What are the three main challenges facing developers in this context today?

A: Ever faster I/Os such as PCIe Gen 4 or 10 Gigabit Ethernet present developers with entirely new challenges. This starts with the routing on the carrier boards. We support our customers with comprehensive workshops, best-practice design solutions that they can use directly in their own developments, and detailed design guides. Key issue number two is real-time networking, for example in smart factories. To this end, we have launched a Time Sensitive Networking (TSN) initiative to enable real-time Ethernet communication based on open standards. Since TSN even supports single pair Ethernet, it can replace both proprietary industrial Ethernet protocols and fieldbuses. The third challenge pertains to system consolidation. Here, OEMs need a simple way to easily partition their multicore-based designs. This means assigning the individual resources to the relevant virtual machines without increasing latency, i.e., without losing real-time capability. To achieve this, we not only offer validated packages, but also support customers with the extensive real-time hypervisor expertise we acquired with Real-Time Systems.

Q: A major advantage of PCIe is that it’s backward compatible with previous generations. This means there’s no problem using PCIe Gen 4 based processors with standards such as COM Express. So, why should developers switch to the new standard?

A: There is a very clear and straightforward answer: COM-HPC targets the performance class above COM Express. So, whenever the maximum CPU performance or the PCIe interface bandwidth of COM Express is insufficient, or if developers know that they will need more than COM Express can deliver in the future, they should switch to COM-HPC sooner rather than later. All those whose applications are and will be powered entirely adequately with COM Express in the future should of course stay with this most successful open Computer-on-Module standard. In the majority of cases, support of the new PCIe Gen 4 and the upcoming PCIe Gen 5 specification is as of yet an unlikely reason compelling embedded OEMs to switch, since the range of suitable peripherals is still quite small today. As both standards are hosted by the PICMG and the largest COM Express players are also involved in the COM-HPC working group, the offered support can be regarded as identical. This ensures highest continuity, which means it will always be easy for OEMs to migrate their COM Express designs to COM-HPC. Newcomers to the Computer-on-Module world will receive the best possible support for both standards. So, whether they choose COM Express or COM-HPC, they have an equally low learning curve to master.

To date, only the COM-HPC footprints and pinouts are known. The first COM-HPC Server Computer-on-Modules are expected to be launched in Q1/2022. At Embedded World, perhaps?

Q: What challenges do you see for congatec in 2022 and how do you plan to meet them?

A: A major task is the market launch of new COM-HPC Server modules. After all, this new form factor is only just beginning to open up the field of modular edge and fog servers with performance requirements that go beyond COM Express Type 7 Server-on-Modules. Industrial micro servers are one of the easier new targets for us here. Manufacturers at the telco-grade edge data center level, on the other hand, are still fairly focused on in-house development, which means we have a lot more convincing to do. However, the as-a-service trend, as well as the need to be able to increase performance per rack very quickly and more and more sustainably, play well into our hands as a module supplier. We see a new – very demanding but also very promising – market opening up for us besides the classic embedded market. Supplying into 5G networks is associated with attractive quantities, of course. In principle, the same also applies to the automotive and mobile vehicle market.

congatec | www.congatec.com

Keep up-to-date with our FREE Weekly Newsletter!

Don't miss out on upcoming issues of Circuit Cellar.


Note: We’ve made the May 2020 issue of Circuit Cellar available as a free sample issue. In it, you’ll find a rich variety of the kinds of articles and information that exemplify a typical issue of the current magazine.

Would you like to write for Circuit Cellar? We are always accepting articles/posts from the technical community. Get in touch with us and let's discuss your ideas.
| Website | + posts

Dan Demers is Director of Sales & Marketing, congatec Americas. Dan Demers holds a B.B.S degree in International Business from Grand Valley State University, Grand Rapids, Michigan and an M.B.A. from Ashford University, Clinton, Iowa. Mr. Demers has over 22 years of experience in embedded computing having worked with Fortune 500 companies in the Industrial, Medical, and Communications markets.

About the author

Dan Demers

Dan Demers is Director of Sales & Marketing, congatec Americas. Dan Demers holds a B.B.S degree in International Business from Grand Valley State University, Grand Rapids, Michigan and an M.B.A. from Ashford University, Clinton, Iowa. Mr. Demers has over 22 years of experience in embedded computing having worked with Fortune 500 companies in the Industrial, Medical, and Communications markets.

Supporting Companies

Upcoming Events


Copyright © KCK Media Corp.
All Rights Reserved

Copyright © 2021 KCK Media Corp.

Q & A: The Future of Embedded Computing

by Dan Demers time to read: 6 min