Your browser is not supported. Please update it.

17 April 2020

Xilinx upgrades 5G-focused Versal ACAP to ride the storm; wins Samsung

Xilinx has announced that its Versal ACAP platform for intensive compute tasks will be used by Samsung in 5G base stations. This comes hard on the heels of the announcement of Versal Premium, a high end member of the family targeted directly at telecoms networks.

Both pieces of news are welcome for Xilinx supporters. The company has been open about several challenges recently:

  • It has been hit by the US/China trade war, with the prospect of being restricted in supplying Huawei and other key customers; and this uncertainty has been worsened by the impact of Covid-19 on the semiconductor supply chain and the delays to some customers’ deployment plans.
  • Field programmable gate arrays (FPGAs) – the product category which Xilinx introduced in the 1980s, and with which it remains most associated – suffered a dose of negativity when Nokia was forced to change direction on its homegrown 5G base station chipset, blaming the expense of its heavily FPGA-centric strategy. There will still be FPGAs in its platform, but it will rely more heavily on readymade system-on-chip products.
  • Intel acquired Xilinx’s main challenger in FPGAs, Altera, in 2015. That benefited Xilinx in some ways, causing uncertainty around Altera’s roadmap and how well Intel would support it. However, as Intel progresses solidly in 5G infrastructure, it is natural for it to take its Altera product line – as well as other programmable products acquired with eASIC and Omnitek – into those contracts. This could potentially leave Xilinx exposed in accounts where its FPGAs were used in 3G and 4G base stations, as the big RAN equipment makers move towards virtualized RAN architectures where they may look to buy CPUs and FPGAs from the same supplier.

So it is doubly important for Xilinx to make a 5G splash now, and prove its platform has superior capabilities in these new networks, even for OEMs which are buying processors from Intel. The company has a strong balance sheet and, according to most Wall Street watchers, solid fundamentals, making it well-placed to ride the difficult years ahead. And in recent years it has been expanding its portfolio considerably, to target a wider range of applications – from data centers and 5G networks to AI and connected devices – while adding full platform products and services to its FPGA core.

Hence the branding of Versal ACAP – which stands for adaptive compute acceleration platform, a far broader remit than FPGAs alone. The key roles of Xilinx’s offerings will still be to offload and accelerate compute-intensive tasks from the CPU, but they will rely on a variety of components and software to do that. Versal ACAP combines ‘adaptable hardware’ (programmable logic), scalar engines (CPUs), intelligent ‘engines’ (AI or DSP accelerators) and hard IP blocks for connectivity and security.

At the launch of Versal ACAP, CEO Victor Peng went as far as to say Xilinx was “no longer an FPGA company”.

Among the key tasks which Versal is designed to accelerate, according to Xilinx, are beamforming for 5G wireless communications, network packet communications, and smart controllers for large solid-state storage systems in data centers. Such demanding processes led Xilinx to create a platform with multiple different engines, each addressing a different type of requirement, from raw performance to power efficiency (see inset).

Some early 5G successes will be important to prove that was the right strategy, both for revenue growth and investor and customer confidence. In the wake of the Versal ACAP launch, indicators were good. For its third fiscal quarter of 2019, its wireless communications business was its strongest, with 41% year-on-year growth in revenue (compared to 34% for the whole company, reaching $800m). Part of this was already down to 5G, Peng said, with Xilinx being present in large numbers of pre-commercial proof of concept initiatives, and in the beginnings of commercial roll-out in China and South Korea.

But the environment is more challenging now, hence the importance of the timing of the Samsung announcement. The Korean company has staked a great deal on achieving a strong position in the RAN in the 5G era, something that has eluded it, outside Korea, in 4G. It has developed base station and RF chips inhouse, and is championing open RAN platforms to increase its impact. Now it will use Versal ACAP to offload demanding tasks from the base station processors.

“Samsung has been working closely with Xilinx, paving the way for enhancing our 5G technical leadership and opening up a new era in 5G,” said Jaeho Jeon, head of R&D at Samsung’s Networks business unit. “Taking a step further by applying Xilinx’s new advanced platform to our solutions, we expect to increase 5G performance and accelerate our leadership position in the global market.”

Versal ACAP will be used to perform complex, real time signal processing for 5G base stations, including sophisticated beamforming, especially in millimeter wave spectrum, an aspect of 5G in which Samsung has established some technology leadership. The adaptive element of Versal ACAP, and of FPGAs in general, are important in these challenging environments because 5G standards, as well as implementations to meet operator requirements, are very fluid in this early stage of the market, so it is essential to be able to modify and re-program systems in the field.

Xilinx explained: “Beamforming technology requires significant compute density and advanced high speed connectivity – on-chip and off-chip – to meet 5G’s low latency requirements. Adding to this, different system-functional partition requirements and algorithm implementations lead to a wide range of processing performance and compute precision. It is extremely challenging for traditional FPGAs to optimally address this requirement while meeting thermal and system footprint constraints.”

It claims that Versal ACAP combines leading levels of compute density and low power consumption in order to support beamforming efficiently. It takes advantage of multiple ‘engines’, including the AI Engines, which are made up of a tiled array of vector processors, capable of handling high compute density and advanced connectivity.

Liam Madden, general manager of Xilinx’s wired and wireless group, said in a statement: “Versal ACAPs will provide Samsung with the superior signal processing performance and adaptability needed to deliver an exceptional 5G experience to its customers now and into the future.”

Meanwhile, Xilinx recently announced more details of the Versal Premium family, targeted in particular at telco core infrastructure and cloud applications. It claims these devices provide three times the bandwidth and enable double the compute density for accelerated workloads, compared to Xilinx’s previous generation of FPGAs.

This is the third of six groups of products within the Versal portfolio to be announced in detail. Each one has a different combination of all the ACAP elements, according to its target applications. We have previously seen the general purpose Versal Prime and the Versal AI Core family for AI acceleration. Versal AI Edge and AI RF series – supporting RF and edge compute applications with AI engines – will follow later this year, while high bandwidth memory (HBM) series, an extension of Premium, will debut next year.

Kirk Saban, VP of product and platform marketing at Xilinx, said the increase in traffic and device connections driven by 5G would put the core network under huge stress, with operators needing to provision for up to 100 times more traffic capacity within the same space and power constraints. Traditional CPUs, GPUs and FPGAs would not address that challenge alone, hence the need for combinations of different elements.

In Premium, the focus is on adding hardened IP blocks to support networking processes. These include 400Gbps high speed cryptography engines for network security applications, multiple 600Gbps Ethernet cores and 112Gbps PAM4 high speed serial transceivers. The hardened IP saves a lot of space and power. In the previous generation of 16-nanometer Xilinx FPGAs, 22 chips would have been needed to program the capabilities Versal Prime offers from a single 7nm chip.

In addition, the chip’s programmable logic can be used to customize the hardware implementation for a particular application, and Saban told EETimes that, in core networks, that increasingly involved AI/ML.

“That trend is moving more and more to the core of the networks, using AI and machine learning algorithms for network optimization, anomaly detection, security threat analysis and prevention,” he said. “These networks are very complex, particularly on the metro and transport side of things. Even a small failure can be responsible for large maintenance costs. Operators can leverage AI algorithms to detect when something is going wrong in the network and then fix it. This is huge from a cost savings perspective.”

However, Versal Premium does not include the AI Engine which Xilinx includes in three of its Versal families – AI Core, AI Edge and AI RF. These are targeted at customers with greater AI acceleration requirements.

There will be seven devices in the Premium range and Versal Premium silicon will start sampling in the first half of 2020 and will begin shipping in volume in the first half of 2021.

The Versal ACAP architecture:

Versal is the brand name for the products based on Xilinx’s ACAP (adaptive compute acceleration platform) architecture, which it has been gradually unveiling through most of the past year. The initial offerings consist of six families of devices. While they all contain Xilinx’s FPGA array fabric, which it now calls the ‘adaptable engine’, they also include several other elements – hence the company’s defocusing on the FPGA term which has defined its history.

The other main elements are:

  • Scalar Engines based on ARM Cortex-72 and Cortex R5 processors.
  • Massively parallel Intelligent Engines for AI and digital signal processing (DSP), based on hundreds of purpose-built, networked and programmable processors.
  • DSP Engines (formerly known as DSP slices), now enhanced with hardened, floating-point extensions.
  • Various hardened protocol blocks for standard interface and memory protocols including Ethernet, PCIe, CCIX and SDRAM controllers.
  • High speed Serdes ports from 32Gbps to 112Gbps, and programmable I/O.
  • On-chip memory distributed all over the device as local RAM for the scalar and AI engines, and embedded in the FPGA fabric.
  • A pervasive network-on-chip (NoC) to tie all the engines to the large numbers of on-chip memories.
  • A software-controlled Platform Management Controller (PMC) to manage booting, configuration, dynamic reconfiguration, encryption, authentication, power management, and system monitoring for the entire device. The PMC claims to speed up dynamic device reconfiguration eightfold compared to

previous generation Xilinx devices.

The products are designed to be programmable in high level languages like C and Python, further moving away from the notion that FPGAs are extremely difficult to program.

Of the six Versal product families, three are targeted specifically at AI processing in many markets, including mobile operators, which are increasingly planning to use machine learning for advanced network optimization as well as extreme personalization of user experiences. The three AI products are the Versal AI Edge, AI Core and AI RF. Then there are three non-AI families, which include all the components listed above except the AI Engines. These are the Versal Prime, Premium and HBM.