Almost exactly one year ago at CES 2023, Qualcomm announced its Snapdragon Ride Flex system-on-chip (SoC) product family. As an expansion of the company’s Snapdragon Digital Chassis product portfolio, the new SoC family is meant to support advanced driver assistance systems (ADAS) as well as digital cockpit and infotainment applications spanning from entry level to premium vehicles. At the time, Qualcomm announced that the Ride Flex SoC was sampling with an expected start of production in early 2024. It’s now early 2024 and CES is about to kick off again. Tirias Research is expecting to hear an update on the product family next week. We anticipate the update will include, at the very least, some of the partners who will be bringing the Ride Flex SoCs to market in production volumes this year and into 2025. Given Qualcomm’s track record for hitting their estimated timelines, we felt that a re-cap of the product family is warranted leading up to next week’s anticipated update.
“Flex-ing” Resources to Support Mixed Criticality and Multiple Tiers
The Snapdragon Ride Flex is actually two monolithically integrated 4nm SoCs – a primary SoC and what Qualcomm are calling a Safety Island SoC. The primary SoC consists of a Kryo Gen 6 Arm v8.2 central processing unit (CPU) with integrated L3 cache, an Adreno 663 graphics processing unit (GPU), a Hexagon neural processing unit (NPU), a Spectra 690 image signal processing (ISP), two Adreno display processing units (DPUs) for multiple high-resolution display support and associated memory and I/O interconnects. This part of the SoC is Automotive Safety Integrity Level (ASIL) B certified. The Safety Island SoC, which is ASIL-D certified, consists of a multi-core real-time CPU with enhanced error managements support and isolated memory and peripherals. ASIL is a risk classification methodology established under ISO 26262 from the International Organization for Standardization which defines functional safety for road vehicles. There are four levels defined from A-D with D denoting the highest safety integrity level. The Safety Island SoC can work independently or in a lockstep mode with the primary SoC. This flexible approach allows the Ride Flex SoCs to address its primary mission of supporting mixed-criticality workloads with one chip – workloads with varying levels of safety critical classifications such as ADAS, digital cockpit and infotainment. The ability to do this on one chip helps Tier 1 and automotive OEM partners to save on cost, board space and lower cabling requirements.
Being an extension of the company’s Digital Chassis portfolio, the Ride Flex SoCs have been optimized with the Snapdragon Ride Vision software stack. This optimization between the hardware and the software is essential to support various levels of safety-critical applications in one chip. With the use of two SoCs for ASIL B and ASIL D support, the company was able to partition compute, memory and I/O resources. In addition, Qualcomm also implemented spatial and temporal freedom from interference processing for those resources. ISO 26262 defines freedom from interference as the “absence of cascading failures between two or more elements that could lead to the violation of a safety requirement.” More specifically, spatial freedom from interference is the protection of memory resources for safety and non-safety critical workloads accomplished via hardware as well as low-level software. Temporal freedom from interference, on the other hand, ensures that compute resources are available when needed by safety critical applications. This can be accomplished by the partitioning of application resources as well as the prioritization of common resources like the memory controller for safety critical applications. Last is the implementation of system monitoring where safety-critical applications are capable of self-measurement of key performance indicators (KPIs) and the monitoring software is notified by the application if the required performance levels can’t be met with the current resources.
Another factor that the Snapdragon Ride Flex family is addressing is scalability from entry-level to premium vehicles. In conjunction with leveraging the Ride Vision software stack, the product family also leverages common software development environments (SDE) and tool sets. On the hardware front, Ride Flex SoCs can scale from 50 tera operations per second (TOPS) to support basic safety workloads such as autonomous emergency braking to 2000 TOPS in the future to support fully automated driving like robotaxis. According to Qualcomm, all compute resources from the CPU, GPU and NPU scale accordingly based on the workloads that need to be supported. The common stack, SDE, toolsets and scalable hardware allows Tier 1 partners as well as automotive OEMs to deliver a cohesive user experience across all vehicle models. Additionally, saving on non-recurring engineering and development costs, they also allow those partners to leverage their hardware and software investments while easily enabling capability customization as dictated by their product positioning strategies.
On The Road to Software Defined Vehicles
The expected ramp to production volumes of Qualcomm’s Snapdragon Ride Flex SoC comes at an auspicious time in the evolution of in-vehicle automotive infrastructures. Over the last 10-15 years, there has been a proliferation of sensors in vehicles, an explosion in the amount of data generated by those sensors and an exponential increase in compute resources required to process that data. As a result, the automotive industry is currently undergoing a paradigm shift from legacy de-centralized network topologies to centrally controlled zonal architectures. This shift is facilitating a need for powerful central compute systems like the Ride Flex SoC which are in turn enabling a move towards software defined vehicles or SDVs. Perhaps the strongest indicator of the momentum behind this shift is the robust ecosystem of competitive solutions that are available. Along with Qualcomm, Nvidia has their Drive Thor platform and Mobileye, their EyeQ platform.
The Snapdragon Ride Flex is promising the ability to support multiple levels of safety and non-safety critical applications in one chip as well as scalability from entry level to premium tiers. If it delivers on those promises, it will have the potential to accelerate the industry’s evolution to zonal architecture and ultimately to the SDV. Supporting multiple levels of safety critical applications as well as multi-tier scalability both help with cost per performance as well as scalable reuse of investment which are required for any possibility of mass market adoption. Looking ahead to CES kicking off next week, automotive technologies are looking like they will be prominently featured. Along with announcements from other companies like Nvidia and Mobileye around this topic, Tirias Research is anticipating major innovation and adoption news signaling the beginning of mass market central compute platform adoption and SDVs.
