Google previews A3 GPU VMs with NVIDIA H100s + scalable data fabric

Google unveiled its next-generation A3 GPU VMs designed for training and serving the most demanding generative AI and large language model applications.

Google Compute Engine A3 VMs combine NVIDIA H100 Tensor Core GPUs and Google’s custom-designed 200 Gbps IPUs. The design leverages GPU-to-GPU data transfers bypassing the CPU host and flowing over separate interfaces from other VM networks and data traffic. Google says this innovation enables up to 10x more network bandwidth compared to our A2 VMs, with low tail latencies and high bandwidth stability. 

Google also notes that its Jupiter data center networking fabric scales to tens of thousands of highly interconnected GPUs and allows for full-bandwidth reconfigurable optical links that can adjust the topology on demand. 

Highlights of the A3:

• 8 H100 GPUs utilizing NVIDIA’s Hopper architecture, delivering 3x compute throughput
• 3.6 TB/s bisectional bandwidth between A3’s 8 GPUs via NVIDIA NVSwitch and NVLink 4.0 
• Next-generation 4th Gen Intel Xeon Scalable processors
• 2TB of host memory via 4800 MHz DDR5 DIMMs
• 10x greater networking bandwidth powered by our hardware-enabled IPUs, specialized inter-server GPU communication stack and NCCL optimizations 

“Google Cloud's A3 VMs, powered by next-generation NVIDIA H100 GPUs, will accelerate training and serving of generative AI applications,” said Ian Buck, vice president of hyperscale and high performance computing at NVIDIA. “On the heels of Google Cloud’s recently launched G2 instances, we're proud to continue our work with Google Cloud to help transform enterprises around the world with purpose-built AI infrastructure.”

https://cloud.google.com/blog/products/compute/introducing-a3-supercomputers-with-nvidia-h100-gpus

Read More

Arelion adds high-capacity East Coast route

Arelion has deployed a new fully diverse, high-capacity route from Reston, VA down the U.S. East Coast, connecting directly to the Telxius cable landing station in Virginia Beach, then continuing to Arelion's new Point-of-Presence (PoP) in Raleigh, NC before connecting into Charlotte, NC and continuing south into Jacksonville, FL and Miami, FL. The route will be live in Q3. 

Arelion's new route leverages open optical line systems to support multi-vendor transponders and the latest coherent pluggable components, enabling cost-effective, scalable Wavelength services, including 10G, 100G and 400G, for its customers in thriving data center and technology markets on the U.S. East Coast. In addition, the unique route supports and sustains the growth of Arelion's No.1 ranked Internet backbone, AS1299.

Arelion said the new route enables it to serve the Research Triangle's booming technology and data center markets with scalable capacity and end-to-end trans-Atlantic solutions for multi-terabit scale traffic flowing into North America and Europe. Arelion's new deployment also provides additional capacity and capability to support the terrestrial transport of multi-terabit subsea traffic from new subsea cables to Latin America and the Caribbean terminating in Jacksonville and South Florida.

"This network expansion on the East Coast of the U.S. is in response to our customers' growing requirements for diverse, high-capacity connectivity and end-to-end trans-Atlantic solutions," said Art Kazmierczak, Director of Strategic Sales and Network Development, Arelion. "Arelion's new unique route continues our long-term commitment to invest in our North American network and provide resilient connectivity solutions that facilitate the growth of critical digital industries in rapidly evolving markets."

Additional highlights on Arelion's new unique, fully diverse East Coast route:

• Routing south, Arelion's new diverse network connects from Reston (CoreSite – VA1) to Virginia Beach (Telxius – Virginia Beach CLS).
• The route connects to three key subsea cables in Virginia Beach (Telxius – Virginia Beach CLS), providing trans-Atlantic backhaul into Europe and Latin America, including Derio, Spain (Telxius Derio Communications - MAREA), Paris (Telxius - DUNANT) and Rio de Janeiro (Telxius – BRUSA).
• The route connects to Arelion's new PoP in Raleigh (Flexential), continuing to Charlotte (Tierpoint) & (Digital Realty).
• The network continues south, connecting to Jacksonville (Cologix – JAX1) and Miami (Equinix – NOTA).

Read More

Microsoft signs fusion power purchase agreement with Helion

Microsoft signed a first of its kind agreement with Helion Energy to purchase electricity from a fusion power plant. 

Helion, a start-up based near Seattle, has been developing fusion technology for more than a decade. The company has previously built six working prototypes and was the first private fusion company to reach 100-million-degree plasma temperatures with its sixth fusion prototype. The company is currently building its seventh prototype, which is expected to demonstrate the ability to produce electricity in 2024. The first commercial plant is expected to be online by 2028 and will target power generation of 50 MW or greater after a 1-year ramp up period. 

 “We are optimistic that fusion energy can be an important technology to help the world transition to clean energy,” said Brad Smith, Vice Chair and President at Microsoft. “Helion’s announcement supports our own long term clean energy goals and will advance the market to establish a new, efficient method for bringing more clean energy to the grid, faster.”

Constellation will serve as the power marketer and will manage transmission for the project.

“This collaboration represents a significant milestone for Helion and the fusion industry as a whole,” said David Kirtley, CEO at Helion. “We are grateful for the support of a visionary company like Microsoft. We still have a lot of work to do, but we are confident in our ability to deliver the world’s first fusion power facility.”

https://www.helionenergy.com/articles/announcing-helion-fusion-ppa-with-microsoft-constellation/

Read More

SES now serves 369 million TV homes worldwide

SES now delivers more than 8,000 channels to a total of 369 million TV homes worldwide, an increase of three million over the previous year. 

The overall growth was driven primarily by increases in Europe (+6.3 million) and Africa (+3 million), with stabilisation in Latin America and Asia-Pacific. The reach in North America (-5.4 million) continues to decline due to the ‘cord cutting’ trend and the continued expansion of fibre connectivity in the United States.

TV Homes:

• Africa - 49.5M
• APAC - 33.1M
• LATAM - 44M
• Middle East - 13.4M
• North America - 53.2M
• Europe - 175.8M

Further insights from the Satellite Monitor include:

• While Video on Demand continues to attract viewers, Free to Air TV proves to be growing globally.
• As fibre connectivity expands globally, IPTV continues to grow as a TV reception method, taking share from traditional cable TV.
• In Africa, satellite continues to drive an increase in the number of direct-to-home (DTH) TV homes as an easily scalable and cost-effective method of TV reception. In Ethiopia, SES has gained 90% penetration of TV homes through its partner Ethiosat.

“For years, we have been collecting data and insight from the annual Satellite Monitor market research on our TV neighbourhoods and the shifting market dynamics around video delivery and consumption trends. We do this because we believe we can add value to our customers with this data and empower them to make informed business decisions on the best way to deliver content to a large audience in the most cost-effective way,” said Norbert Hölzle, Global Head of Media at SES.

Read More

Orange tests Google’s generative AI in contact centers

 Orange is testing generative AI from Google Cloud in a next-gen contact center.

Generative AI could be used to transcribe the call, summarize the exchange between the customer and service representatives, and suggest possible follow up actions to the agent based on the discussion.

Google Cloud said the experiment has the potential to dramatically improve both the efficiency and quality of customer interactions.

Read More

MOX deploys Ciena for 400G wavelength service

MOX Networks will be deploying Ciena’s 6500 flexible grid Reconfigurable Optical Add/Drop Multiplexer (ROADM) photonic layer with WaveLogic 5 Extreme (WL5e) coherent optics to enable multi-terabit capacity on its dark fiber network in the US. Both 100G and 400G connectivity will be offered on MOX network segments in the Pacific Northwest region and on Midwest fiber routes connecting the Mid-Atlantic and Southeast regions. 

Ciena confirmed that its equipment will be installed specifically on the Hillsboro-Portland-Seattle and Columbus-Ashburn Atlanta routes in the coming months. 

MOX also recently deployed Telescent’s Gen 4 Network Topology Manager (G4 NTM), an automated interconnect system, to its Hillsboro-Portland-Seattle fiber routes to provide the company with a remote controlled, reconfigurable fiber-optic patch panel and cross-connect platform to implement physical layer automation and diagnostics.

"This is a very exciting time for MOX as we launch our next-generation fiber routes. We continue to invest in our network and the technologies associated with those assets so that our customers have an exceptional experience with our networks. It’s a pleasure to work with Ciena on our latest network upgrades,” said Allen Meeks, President and COO, MOX. “This collaboration with Ciena will allow us to reconfigure backhaul on the fly while providing infinite and intelligent capacity to our customers."

"With Ciena’s WaveLogic 5 Extreme, MOX can deliver its wavelength services with increased capacity, incredible speed, and improved efficiency,” said Kevin Sheehan, CTO of the Americas, Ciena. “This will allow MOX to address the growing bandwidth needs of its content provider and carrier customers and provide an enhanced customer experience—both today and in the future.”

 

Read More

IonQ appoints VP of R&D

IonQ, a quantum computing start-up, appointed Pat Tang as Vice President of Research and Development. 

Tang most recently served as an Engineering Vice President at Amazon Lab126, where he led engineering teams behind the Kindle, Echo, FireTV, and emerging devices group. Previously, as a team lead at Apple, Tang designed and managed the testing and development of several hardware products. He earned his PhD in the solid-state physics group at Imperial College London and most recently founded Project Mustard Seed, a charity dedicated to providing shelter and education to those in need.

IonQ also recently hired Rima Alameddine, formerly of NVIDIA, as Chief Revenue Officer and appointed Wendy Thomas, CEO of Secureworks, to the IonQ Board of Directors. 

Read More