Marvell has a long track record in supporting cellular infrastructure dating back to the 3G networking era. My principal Patrick Moorhead captured his insights in this regard at Marvell’s inaugural analyst event in late January. If interested, you can find that article here. This article contains contributions from him as well.
The company made two strategic announcement this week that should further their leadership. I believe they are significant given the demands that will be placed on future 5G network deployments from a compute and an edge-enabled use case perspective. I would like to summarize the highlights and provide insight into what I believe is compelling.
The first announcement is centered on a refresh of Marvell’s programmable infrastructure processor platform that is targeted for control and data plane workloads. There’s a lot of attention recently on the power of combining 5G and multi-access edge computing, and Marvell claims that this new generation of processors provides a 2x improvement in performance and scalability over its last generation.
Based on Arm v8-A and with support from 4 to 36 Arm cores, I believe the TX2 is well positioned to support demanding tasks such as packet processing and load balancing designed to deliver an improved and consistent quality of service (QoS). With the ability for operators to virtualize and network slice 5G deployments, I believe Marvell’s enhancements are significant and have the potential to support new service monetization opportunities.
The second announcement is centered on Marvell’s next generation of processors that are targeted for 5G base station designs that include baseband and smart radio unit applications. These are built upon the TX2 platform that incorporates a combination of Arm v8 cores and hardware accelerators.
What I find compelling is the improvement in computing power for low-latency intensive 5G applications as well as support for massive MIMO antenna arrays that are demanding given beamforming algorithms. Programmable DSP cores should also provide exceptional investment protection given an ability to flash upgrades without the need for costly truck rolls. The processor platform also supports a wide spectrum footprint including sub 7 GHz and higher band mmWave for deployment flexibility.
Nokia and Samsung Wins
Nokia and Marvell announced a collaboration for 5G radio access SoCs that will be incorporated into the Nokia AirScale RAN product family through the Nokia ReefShark portfolio. This should help shore up some of Nokia’s technical gaps and provide solid support for future 5G standalone and non-standalone deployments. The benefits appear compelling- a reduction in footprint for deployment flexibility, less power for improved OpEx control, and enhanced performance and capacity. The second is a collaboration between Samsung and Marvell also in radio access with an emphasis on delivering the necessary compute power required to support massive MIMO beamforming. Samsung Networks has been growing by leaps and bounds with both fixed and mobile 5G infrastructure deployments outside of its traditional Asian market, so I see great upside for Marvell in this alliance. 5G is inherently more software-defined than the previous “G”s, thus I see great value in their collaboration that should deliver new levels of programmability and performance for demanding use cases that are cloud and edge enabled.
I’m impressed with Marvell’s latest silicon enhancements for 5G. Proof lies in acceptance by cellular infrastructure providers, and this week Marvell made the two aforementioned announcements. These recent Nokia and Samsung collaborations as well as the company’s long track record dating back to the introduction of the first iPod all point to a proven platform supported by a mature ecosystem incorporating both open source and commercial software applications. It’s certainly a competitive space with recent announcements from Intel that they are looking to extend the Intel architecture from the core through access to the edge. Intel and x86 certainly have a strong position in the core, which is essentially a data center, but the access network is a different story and not one Intel has traditionally played in. The decision to go with an x86-based SoC from Intel or an Arm-based SoC from Marvell is a big one and likely to last for multiple generations as the software would have to be written for one instruction set architecture or the other. In the coming weeks, I’ll be consequently diving deeper into RAN silicon and providing further insights. However, with these recent wins, Marvell and Arm seem to be gaining a foothold and these new processor families should position Marvell well into the future of 5G and beyond.