|"Chips" for 5G Infrastructure ...|
Intel, Marvell, Arm, Qualcomm, AMD and others
[and a Hedy Lamarr reference]
>> Major Chip Vendors Driving Revolutionary Changes In 5G Infrastructure
March 10, 2020
Though it’s absolutely essential to the operation of any wireless network, the truth is that the network infrastructure—everything from cell towers to all the bits of networking and computing equipment that works behind the scenes—is arguably the least understood element of technology that we all use on a regular basis.
Part of the reason for this is that, until very recently, most of the work to create the equipment—including the designing of the chips that run inside them—has been done by only a few vendors. Companies like Ericsson, Nokia, Huawei, and a few others are the biggest players here, and frankly, most of them (the notable exception being Huawei as of late) aren’t as closely followed as most other tech companies. Additionally, as with many aspects of the telecom business, the details of network infrastructure equipment and how it works can be extraordinarily complicated, making it difficult, even for those who want to understand more, to make sense of network infrastructure and its effects on all of our devices.
With the roll-out of 5G, however, we’re starting to see some dramatic changes, not only in the technologies being used to power network infrastructure, but also in the companies that are starting to do some of the work. As with a number of 5G-related topics, some of this is directly because of the additional technical requirements that 5G demands, but another significant aspect is that the widespread move to software-defined networking (which is helping drive many of these changes) just happened to coincide with the development and deployment of 5G.
Over the past two weeks, virtually all of the major semiconductor companies made announcements of either new products, new partnerships, or new technologies focused on enabling more 5G infrastructure. (Obviously, much of it was originally intended for Mobile World Congress, which was cancelled, hence the similar timing).
Intel, Marvell, Arm, Qualcomm, AMD and others all had news related to 5G infrastructure. In addition, we’ve started to see Samsung play a bigger role as a network infrastructure provider for 5G deployments, in part because of its more aggressive use of new approaches to building the equipment that these other tech companies are enabling.
Intel kicked things off with the debut of a range of chips designed for various parts of the 5G infrastructure chain. Its new Atom P5900 is a low-power x86-based CPU that’s optimized for wireless base stations, thanks in part to the addition of dedicated network acceleration circuits to the traditional core components. The company also announced the debut of a new type of chip design called a structured ASIC (application-specific integrated circuit)—the first iteration is codenamed Diamond Mesa—that’s also specifically intended for network infrastructure. Much of the traditional infrastructure equipment has used FPGAs (Field Programmable Gate Arrays), in order to process network-critical data. FPGAs are a flexible type of chip that can be programmed (and even reprogrammed) to accelerate specific functions. The Diamond Mesa part provides some of that same level of flexibility, but without the power consumption and pricing concerns that some FPGAs demand.
Marvell, for its part, launched several significant additions to its 5G infrastructure-focused product line, notably the Octeon Fusion and Octeon Tx2 processors, both of which are based around Arm core architectures. The Octeon Fusion combines multiple Arm V8 cores with several programmable DSP (digital signal processing) cores and hardware-accelerated baseband processors all designed to handle the increase in network data flow and processing that 5G networks will require. The Octeon Tx2 is designed to work further up the networking stack, with a focus on higher-level transport and control plane functions. Together, these two chips allow network infrastructure providers to build equipment optimized for the increasingly complex radio signal processing that 5G networks demand, with a particular focus on meeting the increased latency and bandwidth requirements. Notably, Marvell also announced that they were working with Samsung’s network division on enabling a new range of RAN (Radio Access Network)-focused products, powered by the Octeon Fusion and TX2.
At Qualcomm's post-MWC press event, the company discussed its progress with the Snapdragon 865 processor in a range of forthcoming 5G phones and also unveiled a 3rd generation X60 5G modem coming in 2021. On the infrastructure side, however, the company caught many off guard with a lengthy discussion on its growing presence in 5G infrastructure, thanks in part to a partnership with Japan-based Rakuten. Though Rakuten participates in many businesses (some have called it, for example, the Amazon of Japan for its strong online retail business), the company’s discussion at the Qualcomm event was all about a radical new approach to building a completely virtualized 5G network. Specifically, Rakuten talked about leveraging Qualcomm’s FSM100xx line of infrastructure-focused chips, which are designed to enable mmWave-based 5G RAN infrastructure. Qualcomm first unveiled the FMS100 line back in 2018, but only now is starting to see more widespread usage of the platform, with Rakuten, Samsung Networks, and a few other companies talking about their ongoing efforts with the Qualcomm chips. The Rakuten example—which they plan to start deploying in Japan next month—is particularly intriguing because they claim it allows macro-based cell sites to be set up nearly as easily and as quickly as managed WiFi hotspots. That’s an enormous improvement versus traditional macro cell sites, and if the company does bring the technology to other carriers outside of Japan, as they said they were planning to do, that could represent a big shift in how 5G infrastructure gets deployed in certain environments.
Of course, given all these large shifts, it’s easy to presume that the traditional infrastructure players could be challenged, but a few other announcements highlight how even the existing players are taking new approaches to 5G infrastructure. At AMD’s financial analyst day, for example, among the many announcements that the company made was that it was working with Nokia to power its 5G core network with 2nd generation Epyc server processors. Interestingly, Marvell also announced a partnership with Nokia on 5G infrastructure-focused products, with a specific emphasis on combining some customized Marvell components along with Nokia’s ReefShark chipsets to enable 5G multi-RAT (Radio Access Technology). Shortly thereafter, Nokia also announced a partnership with Intel that uses its ReefShark chipsets in Nokia’s 5G Airscale radio access devices.
All told, it’s clear that there are a number of major developments happening in 5G infrastructure by large semiconductor vendors that haven’t traditionally been associated with this typically closed-box world. The software-based demands for applications such as network function virtualization (NFV), as well as the greatly increased throughput demands that 5G networks will eventually enable, are clearly forcing some significant changes for existing providers and potentially opening up opportunities for new providers.
To be clear, shifts in infrastructure typically don’t happen very quickly. However, there’s no doubt that the world of 5G network infrastructure is going to be something that receives a lot more attention from a lot more people in the months and years to come.
On a separate but related note, at the urging of my wife and as the father of a STEM daughter, in honor of International Women’s Day yesterday and Women’s History Month this month, I would like to use this as a reminder of the many woman who have made contributions to the tech industry. Related to network infrastructure, communications, and the progression to 5G, one such woman in history is Hedy Lamarr. The world remembers her as a Hollywood starlet from the 1940s, but she was also a passionate inventor. In fact, she worked to create a “Secret Communications System” for switching radio frequencies during World War ll to prevent enemies from decoding messages that turned out to be the basis for spread spectrum technology. Fast forward to today and Lamar’s spread spectrum invention has ended up as a building block for many of today’s wireless communications technologies, including WiFi, GPS and cellular networks.
The Author: Bob O'Donnell is the president, founder and chief analyst at TECHnalysis Research, a technology market research and consulting firm that counts many of tech industry’s largest vendors among its clients. The firm's research and O’Donnell’s opinions are also regularly used by major media outlets, including Bloomberg TV, CNBC, CNN, Investor's Business Daily, the Wall Street Journal, Yahoo Finance, and more. O'Donnell writes regular columns for USAToday and Forbes, as well as a weekly blog for Tech.pinions.com that's also published on TechSpot, SeekingAlpha and LinkedIn. Prior to founding TECHnalysis Research, O'Donnell served as Program Vice President, Clients and Displays for industry research firm IDC. O'Donnell is a graduate of the University of Notre Dame. <<
# # #
- Eric L. -