Intel Makes 5G Play With Wide Range Of New Chips

While the latest versions of the second-generation Intel Xeon Scalable processors offer improved performance for cloud and other workloads, they also are the base of a an entire new silicon-based infrastructure targeting the nascent 5G industry, Intel says.

Intel Monday enhanced its second-generation Xeon Scalable processor line with new levels of performance, and introduced several other new chips that the company said work with the new Xeon CPUs to build a foundation on which to build future-facing 5G infrastructures.

Lisa Spelman, corporate vice president and general manager of Intel's Xeon products and Data Platforms Group, said Intel's new offerings stem from the company's view that 5G is ground-breaking new technology that will change how the industry looks at data creation and service delivery, and will make it easier to process data on the edge rather than at the core.

"We're readying all of the infrastructure from the data center through the core network out to the edge in order to unlock that full 5G potential," she said. "And we've built our product portfolio with the goal of moving data faster, storing more data, and processing all of it. And we think that the rest of that portfolio that we've built enables us to address the unique requirements of 5G infrastructure."

[Related: Cisco Intros Cisco Silicon One: New ASIC Play Beyond The Network Into The Cloud]

Spelman, addressing a press conference that was held in lieu of the Mobile World Conference that this week was canceled due to concerns over the coronavirus crisis, said Intel sees compute and its surrounding technology generating vast amounts of data to be moved, stored and processed, with it being centered on the cloud.

5G technology builds on that foundation, making other technology trends like artificial intelligence, IoT, and edge computing available for building next-generation services across all applications and use cases and eventually permeating throughout the application layer, Spelman said.

"It is changing the face of entire industries and turns data from [a] burden into opportunity. ... We see more and more compute being pushed out to the edge because fundamentally it is more economical and a better experience to bring the compute closer to the data generation and consumption than to bring all the data back to the computer," she said.

For Intel, this means investing well beyond the CPU into a wide range of technologies to accelerate data movement including Ethernet silicon, photonics and network switching, as well as storage and memory technologies including Intel Optane persistent memory and PLC 3D NAND SSDs, along with the software to stitch it all together, Spelman said.

"I have to say I still laugh every once in a while when I see comments about Intel being a CPU company," she said.

Intel Monday unveiled several new 2nd-Gen Intel Xeon Scalable processors.

For high-performance workloads, Intel introduced the Intel Xeon Gold 6256, which features 12 cores with a base clock speed of 3.6GHz, and the Intel Xeon Gold 6250, which features eight cores with a base clock speed of 3.9GHz. Both also offer a turbo speed of up to 4.5GHz. This gives them a 36 percent increase in performance versus the first-generation Xeon Scalable processors, Spelman said.

The new processors include Intel Deep Learning Boost, which Spelman said gives them up to six times the performance versus competitive processors when used with artificial intelligence workloads.

For mainstream workloads, Intel introduced several Xeon Gold 6200R and Intel Gold 5200R series processors with increased performance and larger cache memory than previous models.

The company also introduced several processors it said are targeted at single-socket entry-level servers and for edge, networking and IoT workloads.

Intel also enhanced its "N" SKUs, which are versions of the second-generation Intel Xeon Scalable processors designed specifically for networking and NFV, or network functions virtualization.

Spelman said the "N" SKUs are part of Intel's goal to move more of the networking processing to NFV.

"There's tweaking and tuning inside of the processor capabilities that goes beyond just the frequency or the cores that actually delivers increased acceleration and increased quality of service, which is incredibly important in this space," she said.

Dan Rodriguez, corporate vice president and general manager of Intel's Network Platforms Group, said at the press conference that Intel has become the world's No. 1 network silicon manufacturer, and helped drive the IT industry to transition to server-based networks.

"We innovated and integrated network optimizations in your Intel Architecture-based CPUs and other Intel-based platform ingredients," Rodriguez said. "This enabled service providers to converge multiple workloads on a server, a rack of servers, or racks of servers."

Intel sees network and edge computing to be a $65 billion opportunity by 2024, and Intel in 2019 already saw $10 billion in revenue from that opportunity, Rodriguez said.

"With our history of helping to build the cloud and driving network transformation, no one is better suited to address the evolving high-performance computing needs of the 5G network and the edge," he said.

To help build that business, Intel Monday unveiled the Intel Atom P5900 processor, which uses the company's 10-nanometer manufacturing process. Rodriguez said the Atom P5900 was designed from the ground up to meet the needs of the radio access network while helping bring the Intel Architecture, or IA, platform into the base station.

The Intel Atom P5900 was designed for 5G's high-bandwidth and low-latency requirements, Rodriguez said.

It features Intel Dynamic Load Balancer, which Rodriguez said offers 3.7 times the packet processing throughput of software-based technologies. It also features up to 1.8 times the integer throughput of the Atom C3000 processors, and up to 5.6 times the cryptography throughput of software-based technologies, he said.

In addition to its standard systems-on-chips, or SoCs, designed specifically for the base station market, Intel offers custom silicon technologies that allow its cloud and communications service provider customers a way to build their own intellectual property into custom Intel silicon, Rodriguez said.

These include FPGAs targeting flexible designs, ASICs for lowest cost and power use and higher performance, and structured ASICs with improved power and cost versus FPGAs with lower complexity versus ASICs, he said.

Intel's eASIC-branded structured ASICs provide customers with custom low-power capabilities for a broad range of data and IO-intensive applications in markets such as 5G wireless, data center, embedded, video, industrial, as well as military applications, Rodriguez said.

"And today I'm happy to introduce the next generation of structured ASICs from Intel code-named Diamond Mesa, which is now open for early access customers," he said. "Diamond Mesa is footprint-compatible with Intel FPGAs and delivers 2X higher performance for up to 50 percent lower power than previous structured ASIC families from Intel."

Intel also introduced the Intel Ethernet 700 network adapters with hardware-enhanced precision time protocol, designed specifically for 5G and other use cases that have very tight latency and timing requirements, Rodriguez said. That lets them access external GPS clocks for cross-network 5G service synchronization targeting 100 nanoseconds phase accuracy, he said.

The new adapters are targeted at latency-sensitive applications such as radio access networks and industrial and financial service markets, Rodriguez said.

"As 5G evolves, we expect new use cases to emerge that can encompass lengthier time-sensitive solutions from across the spectrum as part of a cost-effective solution," he said.

The Intel Ethernet 700 series network adapters are sampling now and will enter in production in the second quarter, he said.