RISC-V GPU will take an important step

Ventana has high-performance RISC-V CPUs that compete with Arm, x86, and others. There is a lot of interest in finding alternatives to Arm, especially in the mobile and automotive sectors. Arm is clearly not interested in supporting RISC-V via GPUs, so the partnership with Imagination is crucial for the emerging RISC-V client space. Partners in the mobility and automotive sectors are interested in seeing this work.

Founded in 2018 in San Jose, California, Ventana Micro Systems plans to build a chiplet processor based on RISC-V that can be adapted to different applications, from automotive and edge computing to data centers, AI/ML and ultra-large scaler) scales up to 192 cores. The company plans to provide chips and IP, with its first product being Veyron V1, launched at the RISC-V Summit in December 2022.

However, a CPU is still just a CPU - RISC, CISC or otherwise. To make it stand out and become a performance leader, you need a SIMD parallel processor accelerator - some call it a GPU. They can do graphics, but the G stands for everything.

To build such a heterogeneous SoC, Ventana turned to one of the most experienced GPU IP companies, Imagination Technologies, for help. Both companies are premier members of RISC-V International and the RISC-V Software Ecosystem (RISE) project, and are strong advocates of open architecture.

While there are more than 3,000 member companies involved in software ecosystem development, the partnership between Ventana and Imagination is an important step forward for the industry as it brings together two processor developers in their respective fields. With their combined expertise in CPUs and GPUs, the two companies are well-positioned to bring new configurations to the RISC-V open architecture.

The two companies will demonstrate simulation models of Imagination GPUs used with Ventana RISC-V CPUs at the RISC-V Summit next week. Ventana will launch the Veyron V2 next week, so this launch is centered around that goal. The key change in V2 is that the RISC-V ISA now has most of the necessary features to compete with x86/Arm, such as the standard vector extensions and IOMMU specifications. It will include:

• Hypervisor extension

• Type 1 and Type 2 hypervisors; nested virtualization

• Advanced Interrupt Architecture (AIA)

• Includes native MSI handling and interrupt virtualization

• External and autonomous debugging; trace memory

• Rich performance events and performance counters

The RVA23 feature set includes UCIe chiplet interface, +40% performance, 512b vector units + AI matrix extensions, server-grade IOMMU and RISE software compatibility platform, and Domain Specific Acceleration (DSA).

It's safe to say that if you think of Imagination, you can't not think of our GPUs, and in our eyes, if you want to use GPUs for RISC-V, we're a good choice because of our long history in graphics processors. history.

With over 30 years of experience with PowerVR GPUs, we know how to use graphics processors. During this time, we have continued to innovate to maintain technology leadership in the graphics market. PowerVR's unique Tile-Based Deferred Rendering (TBDR) architecture was revolutionary in the early 1990s, and just in 2020, we delivered multi-core GPU technology with the B-Series, opening up a whole new range of performance points. In 2021 we launched IMG CXT, the first and still the most efficient mobile ray tracing core on the market at Level 4 of the Ray Tracing Level System (RTLS).

All of our GPU innovations run with RISC-V CPUs. Our broad graphics performance range means we have the IP to suit any project. High-quality, error-free IP is our focus, and as one of the most established players in the industry, we always provide long-term support to our customers and can help you quickly integrate our IP into your RISC-V projects to ensure its successful.

When assembling complex chips, you want to know whether the technology inside is reliable. PowerVR GPUs were created by one of the world’s most dedicated graphics processor development teams and are tried and tested in many leading markets, for example, in smartphones powered by the MediaTek Helio P90 and P95 series, or in Renesas Electronics’ R-Car automotive SoC (system on chip).

No matter the size of the product you're building, our range of GPUs has a solution to match. Our GPUs have performance ranges from 1 to 72 pixels per clock and 16 to 2,304 FLOPs per clock - as well as multi-core performance ranges. Different GPUs have different optimizations, so depending on your target SoC, you can choose a GPU that targets area efficiency, power efficiency, or performance efficiency. For example, in our B-Series, the BXE-1-16 is highly area efficient and ideally suited for cost-optimized RISC-V-based set-top box products, while the desktop BXT-32-1024 MC4 is focused on top-tier performance efficiency .

Of course, hardware is only half the puzzle. We work hard to ensure that our GPU software is best-in-class and supports the latest APIs (application programming interfaces), enabling our customers to bring differentiated products to market faster and making it easier for developers to Create cutting-edge applications on IP. Our business drivers are fully optimized for market-specific functionality and are continuously updated throughout the year to support the latest software trends. Starting with version 1.13, it supports RISC-V as the target application processor for Imagination GPU-based SoCs.

In addition to this, we are committed to developing high-quality open source GPU drivers that comply with open source coding standards, enabling chip vendors, OEMs and graphics developers to control their complete graphics software stack. This means users can add, remove and customize functionality for their specific use cases without requiring licenses, approvals or any other costs or logistics. The open source driver initially supports Vulkan 1.0 on Linux, with plans to expand support for OpenGL ES.

At Imagination, we are committed to our long-term strategy of supporting open source, and our resources and community are a perfect match for the open source RISC-V architecture.

It's all well and good to talk about PowerVR GPUs being the best partners for RISC-V, but we can also put it into practice. There are already RISC-V products on the market using our GPU IP.

Last year, StarFive selected our IMG BXE GPU IP for its VisionFive 2 single-board computer (SBC), the world's first RISC-V SBC with 3D graphics capabilities. StarFive stated that our GPUs were selected because of our “proven track record of delivering high-quality semiconductor IP across industries,” stating that our GPUs provide “ and area (PPA) has considerable advantages.” Cloud, industrial, networking, consumer and automotive applications. "

In the following demo video of an automotive cluster application running at Embedded World 2023, you can see the StarFive VisionFive 2 board and its quad-core 64-bit U74 CPU running at 1.5GHz working seamlessly with the IMG BXE GPU.

Another example is that Alibaba licenses our IMG B-series GPUs for use in its T-Head Semiconductor AIoT applications. The video below shows a high-performance T-Head TH1520 running a preliminary version of Android 12 for RISC-V at Embedded World 2023, equipped with a quad-core Xuantie C910 CPU and supported by an IMG B-series GPU and NNA (Neural Network Accelerator) .

*Disclaimer: This article is original by the author. The content of the article is the personal opinion of the author. The reprinting by Semiconductor Industry Watch is only to convey a different point of view. It does not mean that Semiconductor Industry Watch agrees or supports the view. If you have any objections, please contact Semiconductor Industry Watch.

Today is the 3576th issue shared by "Semiconductor Industry Observation" with you. Welcome to pay attention.