Tag: trillion

The socket is the motherboard, Part 2 — Intel archrival (and Nvidia’s BFF) plans to build giant chips that could use kilowatts of power but they won’t be as big as Cerebras trillion transistor behemoth

Post author By lisa nichols
Post date May 11, 2024
No Comments on The socket is the motherboard, Part 2 — Intel archrival (and Nvidia’s BFF) plans to build giant chips that could use kilowatts of power but they won’t be as big as Cerebras trillion transistor behemoth

[ad_1]

A few weeks ago, we wrote how Eliyan’s NuLink PHY could do away with silicon interposers and integrate everything into an single, elegant package. How, essentially, the socket could become the motherboard.

At the recent 30th annual North America Technology Symposium, the Taiwan Semiconductor Manufacturing Company (TSMC) revealed plans to construct a version of its chip-on-wafer-on-substrate (CoWoS) packaging technology that could lead to system-in-packages (SiPs) over twice the size of the current largest ones.

“With System-on-Wafer, TSMC is providing a revolutionary new option to enable a large array of dies on a 300mm wafer, offering more compute power while occupying far less data center space and boosting performance per watt by orders of magnitude,” the company said.

An enormous amount of power

TSMC’s first SoW offering, a logic-only wafer based on Integrated Fan-Out (InFO) technology, is already in production.

A chip-on-wafer version using CoWoS technology is expected to arrive in 2027, and will enable the “integration of SoIC, HBM and other components to create a powerful wafer-level system with computing power comparable to a data center server rack, or even an entire server.“

Reporting on the move, Tom’s Hardware expands on this saying, “One of the designs that TSMC envisions relies on four stacked SoICs mated with 12 HBM4 memory stacks and additional I/O dies. Such a giant will certainly draw an enormous amount of power – we are talking about thousands of watts here and will need a very sophisticated cooling technology. TSMC also expects such solutions to use a 120x120mm substrate.”

TSMC’s ambitious pursuit to create gigantic chips, however, is dwarfed by Cerebras Systems’ newest Wafer Scale Engine 3 (WSE-3), termed the “fastest AI chip in the world”. The WSE-3 boasts four trillion transistors and is twice as powerful as its predecessor, the WSE-2, while maintaining the same energy consumption and price. This new chip created on a 5nm TSMC process, provides a staggering peak AI performance of 125 petaflops – which is equivalent to 62 Nvidia H100 GPUs.

This camera captures 156.3 trillion frames per second

Post author By lisa nichols
Post date March 26, 2024
No Comments on This camera captures 156.3 trillion frames per second

[ad_1]

Scientists have created a blazing-fast scientific camera that shoots images at an encoding rate of 156.3 terahertz (THz) to individual pixels — equivalent to 156.3 trillion frames per second. Dubbed SCARF (swept-coded aperture real-time femtophotography), the research-grade camera could lead to breakthroughs in fields studying micro-events that come and go too quickly for today’s most expensive scientific sensors.

SCARF has successfully captured ultrafast events like absorption in a semiconductor and the demagnetization of a metal alloy. The research could open new frontiers in areas as diverse as shock wave mechanics or developing more effective medicine.

Leading the research team was Professor Jinyang Liang of Canada’s Institut national de la recherche scientifique (INRS). He’s a globally recognized pioneer in ultrafast photography who built on his breakthroughs from a separate study six years ago. The current research was published in Nature, summarized in a press release from INRS and first reported on by Science Daily.

Professor Liang and company tailored their research as a fresh take on ultrafast cameras. Typically, these systems use a sequential approach: capture frames one at a time and piece them together to observe the objects in motion. But that approach has limitations. “For example, phenomena such as femtosecond laser ablation, shock-wave interaction with living cells, and optical chaos cannot be studied this way,” Liang said.

Components of a research-grade camera spread in a row on a scientific table. — *SCARF* (Institut national de la recherche scientifique)

The new camera builds on Liang’s previous research to upend traditional ultrafast camera logic. “SCARF overcomes these challenges,” INRS communication officer Julie Robert wrote in a statement. “Its imaging modality enables ultrafast sweeping of a static coded aperture while not shearing the ultrafast phenomenon. This provides full-sequence encoding rates of up to 156.3 THz to individual pixels on a camera with a charge-coupled device (CCD). These results can be obtained in a single shot at tunable frame rates and spatial scales in both reflection and transmission modes.”

In extremely simplified terms, that means the camera uses a computational imaging modality to capture spatial information by letting light enter its sensor at slightly different times. Not having to process the spatial data at the moment is part of what frees the camera to capture those extremely quick “chirped” laser pulses at up to 156.3 trillion times per second. The images’ raw data can then be processed by a computer algorithm that decodes the time-staggered inputs, transforming each of the trillions of frames into a complete picture.

Remarkably, it did so “using off-the-shelf and passive optical components,” as the paper describes. The team describes SCARF as low-cost with low power consumption and high measurement quality compared to existing techniques.

Although SCARF is focused more on research than consumers, the team is already working with two companies, Axis Photonique and Few-Cycle, to develop commercial versions, presumably for peers at other higher learning or scientific institutions.

For a more technical explanation of the camera and its potential applications, you can view the full paper in Nature.

[ad_2]

Source Article Link

Tags camera, captures, frames, trillion

Featured

Want to break the geekiest world record of them all? Here’s how engineers spent 75 days to calculate Pi to 105 trillion digits — just in time for Pi day

[ad_1]

Engineers at StorageReview decided to do something incredibly geeky for this year’s Pi day (March 14 – 3/14) – beat their own record for calculating Pi. Considering that the previous record, achieved last year, was 100 trillion digits, the challenge was no mean feat. While – spoiler alert – they smashed the record, it did take them 75 days to accomplish it.

The task was achieved using a dual processor 128-core AMD EPYC 9754 Bergamo system, equipped with 1.5TB of DRAM and nearly a petabyte of Solidigm QLC SSDs.

The team started their computation on December 14, 2023, and finished on February 27, 2024, spanning 75 days. They used the Chudnovsky (1988) algorithm to calculate Pi, and the computation required a total memory of 1.36 TiB.

New challenges

The journey to 105 trillion digits of Pi – the new record – was not without fresh challenges. The team had to deal with performance-related issues, which led them to delve into the intricacies of parallel computing and hardware interactions. They discovered a CPU hazard specific to the Zen4 architecture involving super-alignment and its effects on memory access patterns.

The engineers also encountered a critical floating-point arithmetic error within the AVX512 code path of the N63 multiply algorithm. With remote assistance from the developer, Alexander Yee, they were able to diagnose and fix the problem, resulting in the successful computation.

Summing up, StorageReview’s Jordan Ranous noted, “The run to 105 trillion digits of Pi was much more complex than we expected. Upon reflection, we should have expected to encounter new issues; after all, we’re completing a computation that had never been done before. But with the 100 trillion computation completed with a much more “duct tape and chicken wire” configuration, we thought we had it made. Ultimately, it took a collaborative effort to get this rig across the finish line.”

Was it worth it? Ranous says, “While we rejoice with our partners in this record-breaking run, we must ask, “What does this even mean?” Five more trillion digits of Pi probably won’t make a huge difference to mathematics. Still, we can draw some lines between computational workloads and the need for modern underlying hardware to support them. Fundamentally, this exercise reflects that the proper hardware makes all the difference, whether an enterprise data center cluster or a large HPC installment. For the Pi computation, we were completely restricted by storage. Faster CPUs will help accelerate the math, but the limiting factor to many new world records is the amount of local storage in the box.”

You can watch the video of the record breaking attempt below.

‘The fastest AI chip in the world’: Gigantic AI CPU has almost one million cores — Cerebras has Nvidia firmily in its sights as it unveils the WSE-3, a chip that can train AI models with 24 trillion parameters

[ad_1]

Cerebras Systems has unveiled its Wafer Scale Engine 3 (WSE-3), dubbed the “fastest AI chip in the world.”

The WSE-3, which powers the Cerebras CS-3 AI supercomputer, reportedly offers twice the performance of its predecessor, the WSE-2, at the same power consumption and price.

The chip is capable of training AI models with up to 24 trillion parameters, a significant leap from previous models.

CS-3 supercomputer

The WSE-3 is built on a 5nm TSMC process and features 44GB on-chip SRAM. It boasts four trillion transistors and 900,000 AI-optimized compute cores, delivering a peak AI performance of 125 petaflops – that’s the theoretical equivalent to about 62 Nvidia H100 GPUs.

The CS-3 supercomputer, powered by the WSE-3, is designed to train next-generation AI models that are 10 times larger than GPT-4 and Gemini. With a memory system of up to 1.2 petabytes, it can reportedly store 24 trillion parameter models in a single logical memory space, simplifying training workflow and boosting developer productivity.

Cerebras says its CS-3 supercomputer is optimized for both enterprise and hyperscale needs, and it offers superior power efficiency and software simplicity, requiring 97% less code than GPUs for large language models (LLMs).

Cerebras CEO and co-founder, Andrew Feldman, said, “WSE-3 is the fastest AI chip in the world, purpose-built for the latest cutting-edge AI work, from mixture of experts to 24 trillion parameter models. We are thrilled for bring WSE-3 and CS-3 to market to help solve today’s biggest AI challenges.”

The company says it already has a backlog of orders for the CS-3 across enterprise, government, and international clouds. The CS-3 will also play a significant role in the strategic partnership between Cerebras and G42, which has already delivered 8 exaFLOPs of AI supercomputer performance via Condor Galaxy 1 and 2. A third installation, Condor Galaxy 3, is currently under construction and will be built with 64 CS-3 systems, producing 8 exaFLOPs of AI compute.

Inside NVIDIA HQ the offices of a $2 Trillion company explored

Post author By miranda cosgrove
Post date March 2, 2024
No Comments on Inside NVIDIA HQ the offices of a $2 Trillion company explored

Inside NVIDIA HQ the offices of a $2 Trillion company explored

Imagine stepping into a workspace that’s not just an office but a hub of innovation, where the design itself inspires creativity and collaboration. That’s what you’ll find at NVIDIA’s headquarters in the bustling tech epicenter of Silicon Valley. The Santa Clara, California campus, with its two main buildings, Endeavor and Voyager inspired by Star Trek, covers a vast area of 1.25 million square feet and is a reflection of NVIDIA’s pioneering spirit in the tech industry.

Inside NVIDIA HQ Offices outside

As you walk through the doors of this impressive campus, you’ll notice how the company has harnessed the power of natural light. NVIDIA has used its own cutting-edge visualization tools to create an environment that maximizes sunlight, reducing the need for artificial lighting and enhancing the well-being of its employees. The office space is bathed in light filtered through 511 triangular skylights, a clever design choice that not only illuminates but also showcases NVIDIA’s flair for innovation.

NVIDIA Offices

Architectural designs inspired by NVIDIA’s chips and the company’s focus on 3D graphics, with a recurring triangle motif.
The campus includes two buildings named Endeavor and Voyager, with Star Trek-inspired names and an outdoor park connecting them.
Endeavor and Voyager buildings, offer 500,000 and 750,000 square feet respectively.
The “Heart” of Endeavor and the “Mountain” in Voyager, central areas designed to encourage interaction and provide active spaces like reception, conference rooms, and coffee breaks.
Open office layouts in both buildings to enhance collaboration among the approximately 5,000 employees, with acoustical design features to manage noise.
An emphasis on visual connections to facilitate quick iteration and deeper relationships, supported by findings from an MIT study.
Numerous staircases and strategically placed elevators to promote movement and encounters among staff.
Efficient pathways and shortcuts, including a two-minute connection between the two buildings.
511 triangular skylights to control lighting, with certain areas like the center of the mountain intentionally shielded from daylight to accommodate lab spaces.
Integration of greenery, including an 80-foot living wall in Endeavor and over 14,000 plants in Voyager, to balance the technological environment.
NVIDIA hints at the potential for future expansion and the use of AI in the design and construction of additional buildings.

Guided tour of NVIDIA’s HQ workspaces

The architecture of the headquarters is a nod to NVIDIA’s heritage in 3D graphics. Triangles, the fundamental element of 3D rendering, are a recurring motif throughout the campus. This design choice is not just about aesthetics; it’s a constant visual stimulus for the staff, reminding them of the core products that have shaped the company’s success.

Here are some other articles you may find of interest on the subject of NVIDIA technology :

The layout of the Endeavor and Voyager buildings is open and airy, designed to bring together the 5,000 employees who work there. This openness is complemented by acoustical design elements that strike the perfect balance between the lively energy of teamwork and the need for individual focus and quiet.

One of the most striking features of the design is the emphasis on visual connections. Central areas like the “heart” of Endeavor and the “mountain” in Voyager serve as communal hubs. These spaces are more than just meeting points; they’re strategic areas that encourage quick exchanges of ideas and foster strong relationships among colleagues, echoing findings from MIT on the importance of such interactions.

The headquarters is built to encourage movement and chance meetings. With staircases and elevators strategically placed to connect different areas, the campus is designed for dynamic interaction. A quick two-minute walk is all it takes to move between the two main buildings, underscoring the company’s focus on a vibrant, collaborative work environment.

But it’s not all about technology and design. Greenery is an essential part of the NVIDIA campus. The Endeavor building features an 80-foot living wall, and Voyager is adorned with over 14,000 plants. These natural elements provide a serene counterbalance to the high-tech surroundings, offering employees a place to recharge and find inspiration away from their screens.

NVIDIA isn’t stopping here. The company is looking to the future, planning to incorporate AI into the design and construction of its next buildings. This forward-thinking approach shows NVIDIA’s dedication to using its own technologies to push the boundaries of workplace design.

The NVIDIA headquarters stands as a space that not only optimizes work conditions but also nurtures a culture of collaboration and well-being. By integrating its advanced technology into every aspect of its office design, NVIDIA has crafted a work environment that goes beyond the conventional, creating a place where innovation is not just developed but lived and breathed by its workforce.

Image Credit : NVIDIA and Gensler Jason O’Rear Photography

Filed Under: Technology News, Top News

Latest timeswonderful Deals

Disclosure: Some of our articles include affiliate links. If you buy something through one of these links, timeswonderful may earn an affiliate commission. Learn about our Disclosure Policy.

Tags company, Explored, NVIDIA, offices, trillion

News

OpenAI CEO Sam Altman raising $7 trillion to build a network of fabs for AI processors

Post author By miranda cosgrove
Post date February 14, 2024
No Comments on OpenAI CEO Sam Altman raising $7 trillion to build a network of fabs for AI processors

OpenAI network of fabs for AI processors

Sam Altman, the visionary head of OpenAI,, is spearheading an ambitious initiative poised to revolutionize the technology landscape. His goal is to establish a global network of factories dedicated to producing the specialized silicon processors—the brains behind artificial intelligence. These processors are essential for powering advanced technologies from smarter phones to self-driving cars, marking a significant leap forward in AI capabilities.

This project, with an estimated cost between $5 and $7 trillion, aims not only to advance OpenAI’s technological base but also to catalyze a global shift towards more sophisticated AI applications. Altman’s strategy involves partnering with industry giants, including the Taiwan Semiconductor Manufacturing Company (TSMC), leveraging their expertise to operationalize these advanced manufacturing hubs. These factories are envisioned to be at the forefront of semiconductor technology, producing chips for a wide array of companies eager to harness AI’s full potential.

The initiative emerges against the backdrop of a global chip shortage that has underscored the fragility of the current semiconductor supply chain, heavily reliant on a few key providers like TSMC. This shortage, exacerbated by geopolitical tensions, particularly between Taiwan and China, has prompted significant moves to diversify production and mitigate risks. The U.S. government, recognizing the strategic importance of semiconductor independence, has blocked certain tech transactions and invested in domestic semiconductor R&D, complementing TSMC’s commitment to build a $40 billion chip plant in the United States.

OpenAI AI Silicon Processors

Altman’s vision transcends merely addressing the current chip shortage. It represents a foundational shift towards diversifying chip production globally, reducing the geopolitical risk currently concentrated in the semiconductor industry. This strategic expansion is crucial as AI becomes increasingly integrated into every aspect of our lives, from healthcare to transportation, necessitating reliable and powerful semiconductors.

Financing this colossal venture poses a significant challenge, with Altman and his team in discussions with potential investors capable of supporting such a groundbreaking initiative. Beyond financial investment, the project requires global cooperation and alignment with the vision of a future where AI chips are ubiquitously manufactured worldwide, fostering a new era of technological innovation and independence.

The implications of Altman’s project extend far beyond the current technological landscape. By laying the groundwork for the next generation of AI technologies, this initiative not only aims to solve the immediate issues of chip shortages and geopolitical dependencies but also sets the stage for a future where AI’s potential can be fully realized, ensuring the tech industry can meet the growing demands of AI applications.

However, navigating the intricate web of political and financial challenges remains a formidable task. The endeavor is set against a complex backdrop of international politics and economic considerations, where the strategic importance of semiconductor manufacturing has become a focal point of global attention. As Altman’s project moves forward, it will require careful negotiation and strategic partnerships to realize this bold vision for a technologically empowered future.

Sam Altman’s project is not just an ambitious venture; it’s a visionary leap towards redefining the global semiconductor industry, promising to usher in a new era of technology that could fundamentally change how we interact with the world around us. With high stakes and the world watching, the success of this initiative could signal the dawn of an unprecedented technological age.

Source : Toms Hardware

Filed Under: Technology News, Top News

Latest timeswonderful Deals

Disclosure: Some of our articles include affiliate links. If you buy something through one of these links, timeswonderful may earn an affiliate commission. Learn about our Disclosure Policy.

Tags Altman, Build, CEO, fabs, Network, OpenAI, processors, raising, Sam, trillion

News

TinyLlama 1.1B powerful small AI model trained on 3 trillion tokens

Post author By miranda cosgrove
Post date January 5, 2024
No Comments on TinyLlama 1.1B powerful small AI model trained on 3 trillion tokens

TinyLlama powerful AI model trained on 3 trillion tokens

If you are interested in using and installing TinyLlama 1.1B, a new language model that packs a punch despite its small size. This quick guide will take you through the process. TinyLlama is an innovative compact AI model making waves by offering high-level language processing capabilities that can be used on a variety of devices, from desktops to smartphones. It’s a big deal for developers and researchers who need advanced language understanding but don’t have the luxury of unlimited computing power.

TinyLlama 1.1B is built on the efficient Llama 2 architecture, which means it’s not only powerful but also designed to work smoothly with many different open-source projects. This is great news for users who want to add TinyLlama’s advanced features to their existing systems without any hassle. The model comes with a specialized tokenizer that ensures it can communicate effectively with other software, which is a key requirement for anyone looking to upgrade their tech with the latest AI capabilities.

The development of TinyLlama was no small feat. It underwent a rigorous 90-day training period that started on September 1st, 2023, using 16 high-performance GPUs. The goal was to make the model as efficient as possible, teaching it to understand complex language and concepts, including logic and common sense. The training process was closely watched to avoid overfitting, which can reduce a model’s effectiveness. The result is a language model that performs exceptionally well, even when compared to other models that have many more parameters.

How to install TinyLlama 1.1B

Here are some other articles you may find of interest on the subject of compact AI models :

What sets TinyLlama 1.1B apart is its ability to handle complex tasks using far fewer resources than you might expect. This efficiency is a testament to the developers’ focus on optimizing training and making sure the model learns as much as possible without wasting energy or computing power.

For those eager to try out TinyLlama, the model is readily available for download on Hugging Face, a popular platform for sharing machine learning models. This move makes cutting-edge AI technology accessible to a wide audience, from experienced developers to those just starting to dip their toes into the world of artificial intelligence.

TinyLlama 1.1B is a noteworthy development in the field of language modeling and more information is available over on the Huggingface website. It manages to balance a compact size with strong computational abilities, making it an excellent choice for anyone interested in exploring AI. Its compatibility with standard devices and ease of integration make it a valuable resource for those who want to push the boundaries of what’s possible with AI, without needing a supercomputer to do so.

Filed Under: Guides, Top News

Latest timeswonderful Deals

Disclosure: Some of our articles include affiliate links. If you buy something through one of these links, timeswonderful may earn an affiliate commission. Learn about our Disclosure Policy.

Tags 1.1B, model, Powerful, Small, TinyLlama, tokens, trained, trillion