Wednesday, September 15, 2021

GlobalFoundries announces 45nm Silicon Photonics platform

GlobalFoundries announced a  new Silicon Photonics 45nm platform for combining RF CMOS and optical components on the same chip.

The monolithic platform includes an innovative new feature, the first micro ring resonator (MRR) optical component in 300 mm wafer technology.

GF said the new platform has passed critical technology milestones and is on track for full technology qualification by Q1 2022. The company also confirms that it is engaged with leading customers and partners on this new platform.


https://gf.com/blog/gf-innovation-moving-data-light-speed

Ayar Labs demos terabit link for Co-Packaged Optics and chip-to-chip

Ayar Labs demonstrated the industry’s first Terabit per second Wavelength Division Multiplexing (WDM) optical link with its TeraPHY optical I/O chiplet and SuperNova multi-wavelength optical source. The demonstration shows a fully functional TeraPHY chiplet with 8 optical ports running error free without Forward Error Correction (FEC) for a total bandwidth of 1.024 Tbps and at less than 5 pJ/bit energy efficiency. “This is yet another industry...

GlobalFoundries adapts silicon photonics for quantum computing

GlobalFoundries is now manufacturing the silicon photonic and electronic chips that form the foundation of PsiQuantum's Q1 quantum computer.

PsiQuantum, a start-up based in Palo Alto, California, aims to be the first to deliver a commercially viable quantum computer with one million qubits.

PsiQuantum and GF have now demonstrated the ability to manufacture core quantum components, such as single-photon sources and single-photon detectors, with precision and in volume, using the standard manufacturing processes of GF’s semiconductor fab. The companies have also installed proprietary production and manufacturing equipment in two of GF’s 300mm fabs to produce thousands of Q1 silicon photonic chips at its facility in upstate New York, and state-of-the-art electronic control chips at its Fab 1 facility in Dresden, Germany.

PsiQuantum’s Q1 system represents breakthroughs in silicon photonics, which the company believes is the only way to scale to 1 million-plus qubits and beyond and to deliver an error-corrected, fault-tolerant, general-purpose quantum computer.

“In the past year, we have experienced a decade of technological change. Now, due to the digital transformation and the explosion of data we are faced with problems that require quantum computing to further accelerate the Renaissance of Compute,” said Amir Faintuch, senior vice president and general manager of Compute and Wired Infrastructure at GF. “PsiQuantum and GF’s partnership is a powerful combination of PsiQuantum’s photonic quantum computing expertise and GF’s silicon photonics manufacturing capability that will transform industries and technology applications across climate, energy, healthcare, materials science, and government.”

“This is a major achievement for both the quantum and semiconductor industries, demonstrating that it’s possible to build the critical components of a quantum computer on a silicon chip, using the standard manufacturing processes of a world-leading semiconductor fab,” said Pete Shadbolt, chief strategy officer and co-founder of PsiQuantum. “When we first envisioned PsiQuantum, we knew that scaling the system would be the existential question. Together with GLOBALFOUNDRIES, we have validated the manufacturing path for silicon photonics and are confident that by the middle of this decade, PsiQuantum will have completely stood up all the manufacturing lines and processes necessary to begin assembling a final machine.”

Arrcus resets with new CEO, vision, and strategic partners

Arrcus, a hyperscale networking software start-up based in San Jose, California, named Shekar Ayyar as its CEO and chairman of the board. The company also announced a strategic shift toward edge-native, large-scale distributed and disaggregated networking opportunities. In addition, Arrcus announced the infusion of new capital from its strategic partners - Liberty Global, SoftBank Corp and Samsung Next. The new investors join the existing investors, Clear Ventures, General Catalyst Partners, and Lightspeed Venture Partners. 

Ayyar joins Arrcus from VMware where he was the executive vice president and general manager of the Telco and Edge Cloud business. He also led VMware’s strategy and corporate development efforts in enterprise software and communications while overseeing over 60 M&A transactions and investments including the acquisitions of Nicira, AirWatch, and VeloCloud.

Regarding its new focus, Arrcus said 5G and edge computing are accelerating the network infrastructure transformation requiring programmable and massively scalable, multi-cloud optimized, cost-effective connectivity for the next generation of distributed applications including remote work environments, industrial automation, and content distribution. 

The Arrcus Connected Edge (ACE) platform targets the convergence of communications and compute infrastructure at the edge, delivering massive network scale and performance, and enabling service providers and enterprises to create new services and deliver superior end-user experiences at a low total cost of ownership (TCO).


“Shekar is a passionate technology leader with a proven track record of translating strategy into execution and corralling high-performing teams to deliver results. On behalf of Arrcus employees and the board, I am excited to have him join us as our CEO at this important time,” said Keyur Patel, Founder and CTO, Arrcus. “We are confident in Shekar’s leadership to take advantage of the opportunities ahead of us and drive Arrcus forward as we enter the next phase of product innovation and growth.”

“I am delighted to be joining Arrcus as CEO at this inflection point in the industry when customers are seeking edge-native, large-scale distributed and disaggregated networking to support their 5G and edge deployments. The Arrcus Connected Edge (ACE) is uniquely positioned to provide network performance at scale, while delivering efficiently low TCO for enterprises and service providers,” said Shekar Ayyar, CEO, Arrcus.

“Service providers are under relentless pressure to modernize their networks to deliver new, innovative services at the edge while meeting stringent requirements for high bandwidth and low latency,” said Ram Velaga, senior vice president and general manager, Core Switching Group, Broadcom. “At the time when carrier networks are going through this transformation, we are excited to see the Arrcus Connected Edge infrastructure leverage our StrataDNX family of switch SOCs, including Jericho2, Qumran2C, Jericho2C, and Qumran2A merchant silicon platforms to offer a flexible, high-performance, cost-optimized, low-power solution and address the challenging needs of the most demanding telco workloads.”



Arrcus: Opportunities at the Edge for Network Operating Systems?

In this video, Shekar Ayyar shares his perspective on the need for disaggregated networking solutions at the 5G edge.

Ayyarr recently joined Arrcus as CEO and Chairman of the Board. Previously, he was the executive vice president and general manager of VMware's Telco and Edge Cloud business. He also led VMware’s strategy and corporate development efforts in enterprise software and communications while overseeing over 60 M&A transactions and investments including the acquisitions of Nicira, AirWatch, and VeloCloud.

https://youtu.be/UoDRvAAIFJY


Credo's Active Electrical Cables carry 400G over 7m

Credo announced its second generation HiWire LP SPAN AEC (Low Power SPAN Active Electrical Cable) family, promising a reduction in power of nearly 50% and while increasing reach 40% compared to first generation.

The hot-swappable LP SPAN AEC is positioned as a reliable, plug and play replacement of Active Optical Cables (AOC) for high-speed interconnects up to 400Gbps at approximately half the power and lower cost. 

Credo says its LP SPAN AECs are easier to deploy than DACs due to their longer reach, lighter weight, and tighter bend radius.  It’s designed for rack-to-rack connectivity, especially in Distributed Disaggregated Chassis (DDC) implementations.


400G LP SPAN AECs are terminated with QSFP-DD modules and support 8x56G lanes while 200G LP SPAN AECs are terminated with QSFP65 modules and support 4x56G lanes. The cables use PAM4 modulation at both ends and have integrated retimers.

https://www.credosemi.com/releases/2021/9/7/credo-expands-aec-active-electrical-cable-family-with-second-generation-hiwire-low-power-span-aec

Spirent and MultiLane collaborate on QSFP-DD 800G testing

Spirent Communications has partnered with MultiLane, a global leader in data center interconnects, to offer vendor-neutral testing solutions supporting the 800G ecosystem. 

Specifically, Spirent and MultiLane are working together to resolve QSFP-DD800 testing challenges. 

QSFP-DD800 is one of the first port standards that uses individual electrical 100G lanes. The design, optimization, and fabrication of these 100G lanes presents a significant challenge for host and pluggable developers alike. The partnership facilitates the test and verification of QSFP-DD hardware, including transceivers, switches that use these transceivers, and hyperscaler use cases.


“Successful 100G electrical lane design requires novel techniques and robust test procedures,” said Kees Propstra, Vice President of Marketing at MultiLane. “Our joint effort with Spirent to tackle these challenges facilitates the transition to the 800G node, the next frontier in data center interconnects.”

Spirent offers 800G validation solutions for the entire ecosystem, with support for various speed modes including 1x800G, 2x400G, 4x200G and 8x100G. The company helps its customers qualify equipment so they can determine which supplier will best meet their requirements. For partners, Spirent provides a neutral test solution to validate their vendors.

“Spirent has developed a comprehensive, end-to-end 800G testing suite that leverages our decades of experience in Ethernet testing,” said Aniket Khosla, VP of Product Management, Cloud and IP at Spirent Communications. “Our partnership with MultiLane will assure successful deployments of this complex new technology while helping customers improve product quality and ensure reliable, high-quality user experiences.”

Multilane's 800G test solutions include QSFPDD-800 thermal load loopbacks, module/host compliance boards and noise injection Bit Error Rate Testers (BERTs).

http://www.spirent.com/HSE

MACOM demos TIA interoperability with Broadcom 7nm PAM-4 DSP

 MACOM Technology Solutions has demonstrated interoperability of Broadcom’s PAM-4 DSP and its transimpedance amplifiers (TIAs) for 53 Gbps per lane VCSEL based short reach and 100 Gbps per lane EML based long reach data center applications.

The pairing for single-mode long reach solutions includes Broadcom’s 7nm Centenario DSP (BCM874xx) with MACOM’s four-channel 4x106 Gbps PAM-4 linear transimpedance amplifiers (TIAs) optimized for use in 400G and 800G DR4 and FR4 optical modules. The MATA-03819, MATA-03820 and MATA-03821 are available in flip chip and wire bonding options, respectively, for fast, flexible deployment of QSFP-DD and OSFP optical modules.


The pairing for multi-mode short reach solutions includes Broadcom’s 7nm Estoque DSP (BCM875xx) with MACOM’s two-chip analog solution for 200 Gbps and 400 Gbps short reach SR4 and SR8 modules (QSFP, OSFP, QSFP-DD) and Active Optical Cable (AOC) applications in the data center. 

The MACOM chipset consists of two devices, the MATA-38434 4x53 Gbps PAM-4 TIA and MATA-38435 4x53 Gbps PAM-4 VCSEL driver, that deliver IEEE standard specified performance and are compliant with Open Eye MSA eye mask specifications.

“Our customers require high-performance solutions to enable their data center applications,” said Marek Tlalka, Senior Director, High-Performance Analog, MACOM. “This collaboration offers the industry another proof point towards the interoperability of DSPs to MACOM’s TIA portfolio, and a path to shorter design cycle for 400G and 800G module manufacturers.”

“Broadcom's 7nm DSPs along with MACOM’s TIAs provides industry-leading performance at the lowest power for 200G/400G/800G optical modules,” said Khushrow Machhi, Senior Director of Marketing, Physical Layer Products Division, Broadcom. “Additionally these various modules show the interoperability of Broadcom's DSP with various components needed to develop industry-leading optical modules.”

https://www.macom.com/about/news-and-events/press-release-archive/row-col1/news--event-archive/macom-demonstrates-tia-interoper

Google's Grace Hopper subsea cable lands in Cornwall

The Grace Hopper subsea cable has landed in Bude, Cornwall, UK. This follows a successful Bilbao landing earlier in September.

Google says Grace Hopper will use a new switching architecture to provide optimum levels of network flexibility and resilience to adjust to unforeseen failures or traffic patterns. 

https://blog.google/around-the-globe/google-europe/united-kingdom/our-grace-hopper-subsea-cable-has-landed-uk/?linkId=131609522

Google's Grace Hopper subsea cable to link US-UK-Spain

Google unveiled plans for a new subsea cable — Grace Hopper — which will run between the United States, the United Kingdom and Spain. The cable is named for computer science pioneer Grace Brewster Murray Hopper (1906–1992), an admiral in the U.S. Navy best known for her work in developing the COBOL programming language.

The Grace Hopper cable will be equipped with 16 fiber pairs (32 fibers). Subcom has been selected as the lead contractor. The project is expected to be completed in 2022.


Google said the Grace Hopper cable system will be the first to use a novel optical fiber switching that allows for increased reliability by moving traffic around outages.

Grace Hopper joins Google's other private subsea cables, Curie, Dunant and Equiano.

https://cloud.google.com/blog/products/infrastructure/announcing-googles-grace-hopper-subsea-cable-system

Leading vendors agree to tackle vulnerabilities in global routing

The Mutually Agreed Norms for Routing Security (MANRS) initiative, backed by Arista, Cisco, Huawei, Juniper, and Nokia and supported by the Internet Society, today announced an Equipment Vendor Program aimed at reducing the most common threats to the Internet’s routing system.

The MANRS Equipment Vendor Program provides best practices for all network equipment vendors. By joining MANRS, equipment vendors agree to the baseline of routing security defined by a set of two actions and a commitment:

  • Equipment vendors agree to provide security features network operators can use to implement MANRS recommendations such as preventing incorrect routing information and IP address spoofing.
  • Participants agree to promote MANRS through training and technical content.
  • They also agree to actively participate in the MANRS community to advise members, develop solutions, and contribute to routing security resources.

“Having five of the leading network equipment vendors in the world working together with the MANRS community to launch the new Equipment Vendor Program demonstrates the importance of routing security in making the Internet safe for business and consumers alike,” said Andrei Robachevsky, Senior Director, Technology Programs for the Internet Society. “These founding participants are leading by example to improve routing security and hopefully will motivate the entire Internet infrastructure community to make the Internet more secure for everyone.”

“With a greater reliance on the Internet now more than ever, it’s crucial that the Internet is stable and offers secure and trusted methods to connect users. It is the responsibility of the global community to work together to ensure that the ecosystem forming the Internet works well and is protected against threat actors. Juniper Networks is a proud supporter of the MANRS program and will continue to strive for a better internet in today’s connected world,” says Melchior Aelmans, Consulting Engineer, Juniper Networks.

VIAVI and Capgemini collaborate on 5G and O-RAN validation

Viavi Solutions is working with Capgemini Engineering in Portugal to deliver a 5G and O-RAN lab test capability, powered by VIAVI's O-RAN Lab as a Service (LaaS).


The companies are collaborating to ensure successful network integration in 5G open environments. The VIAVI portfolio of test tools enables measurements for the entire network lifecycle, providing functional, system integration and performance testing of gNBs, core networks, core network components and O-RAN subsystems, delivering true end-user QoE visibility.

"Capgemini is at the forefront of 5G and O-RAN innovation, where testing and maintaining the resilience and performance of an entire network chain is crucial. With an ecosystem of both technology and telecom partners and focus on end-to-end solutions for industries, Capgemini is committed to support our clients to take advantage of the 5G and Edge revolution, driving their data-driven transformation towards Intelligent Industry," said Shamik Mishra, Chief Technology Officer Connectivity, Capgemini Engineering. "We chose to team up with VIAVI to ensure we continue to deliver the best service to our clients, augmenting our 5G Lab As A Service capability in Portugal to reduce time to market and mitigate technical and business risks significantly."


GlobalFoundries and Qualcomm team on 5G RF front-end

GlobalFoundries (GF) and Qualcomm Global Trading PTE., are extending their RF collaboration on 5G multi-Gigabit speed RF front-end products for 5G-enabled products.

“GlobalFoundries continues to lead in RF with feature-rich technology solutions for 5G,” said Dr. Bami Bastani, senior vice president and general manager, mobile and wireless infrastructure strategic business unit at GF. “Our strong collaboration with Qualcomm Technologies includes sub-6 GHz to unlock everyday access to 5G, and cutting-edge mmWave technology to take 5G to the next level by delivering unmatched data speeds while continuing to provide the longest possible battery life for smartphones, computers, automobiles, network access points and many other 5G connected products.”

“With accelerating demand for RF front-end products in a 5G world, robust low-power semiconductor solutions are critical,” said Christian Block, senior vice president and general manager, RFFE, QUALCOMM Germany RFFE GmbH. “Our collaboration with

GlobalFoundries, and their leadership in RF-specific, feature-rich foundry solutions, helps to ensure that we’re able to meet the high-performance requirements of our cutting-edge 5G products.”


SambaNova announces a key hire for its AI team

SambaNova Systems, a Silicon Valley start-up developing AI-driven Dataflow-as-a-Service, announced the appointment of Poonacha Kongetira as Vice President of Hardware.

Kongetira hails from Google where he led a team of engineers developing custom silicon accelerators for machine learning, video transcode, and datacenter infrastructure. He also led NVIDIA’s Bangalore design center for mobile and GPU chips with a team of more than 800 engineers. Kongetira was an early member of the leadership team developing the first Niagara processor at Afara which was acquired by Sun Microsystems and eventually Oracle.

“I am thrilled to join SambaNova working with the best in the industry and to rejoin the familiar faces of a talented team which led the transition from single to multicore processors with Niagara,” said Poonacha Kongetira, VP of Hardware at SambaNova. “Armed with SambaNova’s Reconfigurable Dataflow Architecture and Dataflow-as-a-Service, we have the unique opportunity to establish this unrivaled technology as the industry standard for machine learning and scientific applications just like we did with Niagara and apply the vision of machine learning and AI for enterprises.”

SambaNova raises $676 million for its AI platform

SambaNova Systems, a start-up based in Palo Alto, California, announced $676 million in Series D funding for its software, hardware and services to run AI applications.

SambaNova’s flagship offering is Dataflow-as-a-Service (DaaS), a subscription-based, extensible AI services platform designed to jump-start enterprise-level AI initiatives, augmenting organizations’ AI capabilities and accelerating the work of existing data centers, allowing the organization to focus on its business objectives instead of infrastructure.

At the core of DaaS is SambaNova’s DataScale, an integrated software and hardware systems platform with optimized algorithms and next-generation processors delivering unmatched capabilities and efficiency across applications for training, inference, data analytics, and high-performance computing. SambaNova’s software-defined-hardware approach has set world records in AI performance, accuracy, scale, and ease of use.

The funding round was led by SoftBank Vision Fund 2, and included additional new investors Temasek and GIC, plus existing backers including funds and accounts managed by BlackRock, Intel Capital, GV (formerly Google Ventures), Walden International and WRVI. This Series D brings SambaNova’s total funding to more than $1 billion and rockets its valuation to more than $5 billion.

SambaNova says it is working "to shatter the computational limits of AI hardware and software currently on the market — all while making AI solutions for private and public sectors more accessible."

“We’re here to revolutionize the AI market, and this round greatly accelerates that mission,” said Rodrigo Liang, SambaNova co-founder and CEO. “Traditional CPU and GPU architectures have reached their computational limits. To truly unleash AI’s potential to solve humanity’s greatest technology challenges, a new approach is needed. We’ve figured out that approach, and it’s exciting to see a wealth of prudent investors validate that.”

Stanford Professors Kunle Olukotun and Chris Ré, along with Liang, founded SambaNova in 2017 and came out of stealth in December 2020. Olukotun is known as the “father of the multi-core processor” and the leader of the Stanford Hydra Chip Multiprocessor (CMP) research project. Ré is an associate professor in the Department of Computer Science at Stanford University. He is a MacArthur Genius Award recipient, and is affiliated with the Statistical Machine Learning Group, Pervasive Parallelism Lab, and Stanford AI Lab.

http://www.sambanova.ai