Nova Ltd.’s Post

The future of AI is dependent on computing power, and the semiconductor industry is rising to meet the challenge. With the potential of reaching the 1T transistor mark by the end of the decade, we are witnessing a significant increase in transistor counts. Regarding GPU performance and efficiency, transistor counts are a big decider. Firms like TSMC are optimistic about the future of semiconductor integration and plan to reach the 1T transistor goal over the next decade, categorizing it as an essential step for progress. NVIDIA's recently revealed Blackwell GPU architecture features 208B transistors; the industry expects to be five times this amount over the next decade. The future of AI is looking bright with the semiconductor industry's advancements. #chips #transistors #semiconductors #semiconductorindustry #semiconductormanufacturing #innovation #technology #technologynews

AI workload is changing the requirements of GPU vs. CPUs, and how efficient (and lower cost) to do machine learning. Opportunities for architecture and new startups to innovate to drive cost down, thus making AI more affordable. “In the data center, everything is a 32-bit floating point number. Alternative representations can reduce the size of the operators and the amount of data that needs to be moved and stored,” he noted. “Most AI algorithms do not need the full range that floating point numbers support and work fine with fixed point numbers. Fixed point multipliers are usually about ½ the area and power of a corresponding floating point multiplier, and they run faster."

TSMC boss says one-trillion transistor GPU is possible by early 2030s Timing predictions aside, multi-chip designs with 3D stacking will be the path forward 3D chiplets will be the key to building the world's first one-trillion transistor GPU, says TSMC chairman Mark Liu and chief scientist H.-S. Philip Wong. The semiconductor industry is always going to want to cram more transistors into processors, but as Liu and Wong outline in a IEEE Spectrum report, AI has made this demand even more insatiable. Established companies and startups are snapping up as many GPUs as they can for the use of running AI workloads, and naturally chips with higher performance density are in high demand. Liu and Wong argue that today's 100-billion transistor GPUs just won't cut it: one trillion transistors in a single GPU will be needed. The two argue that this trillion-transistor chip can come about as early as 2034, a decade from today. While newer nodes with increased transistor density will be an important part in getting to one trillion transistors, it's not enough on its own. Instead, Liu and Wong say 3D chiplets, the cutting-edge technology of connecting several chips together beside and on top of each other, are going to be crucial for achieving one trillion transistors.

🚀 TSMC- NVIDIA-Broadcom to Develop Cutting-Edge Silicon Photonics: TSMC has teamed up with Broadcom & NVIDIA to develop cutting-edge silicon photonics, in order to provide massive transmission speeds for AI computing. Silicon photonics is a next-gen adaption of the traditional copper transmission cables, which combines laser and silicon technology to provide a solution that guarantees high data transfer speeds. TSMC has been rumored to allocate 200 professionals for R&D, primarily focusing on the tech in silicon photonic integration into high-speed computing chips. NVIDIA and partners are working on expanding the technology from the 45-nanometer process up to the 7-nanometer, which will ultimately bring in significant performance uplift onboard. The process is expected to come into implementation by 2024, with mass production expected by 2025. Hence we could see next-gen AI GPUs to boast significant transmission speeds in the upcoming years. #semiconductor #semiconductorindustry #tsmc #intel #samsung #imec #globalfoundries #smic #umc #innovation #ai #computerchips #machinelearning #broadcomm #transistor #cowos #skhynix #microntechnology #kioxia #nanya #toshiba #ymtc #yangtze #scaling #moore #manufacturing #production #fabrication #apple #nvidia #arm #amd #qualcomm #ibm #huawei #chip #chipdesign #chipmaker #memory #logic #cpu #processor #FEOL #BEOL #interconnects #dram #nand #3Dnand #nandflash #storage #asml #euv #lithography #pellicle #photonics #photon #light

Tenstorrent heats up AI, GPU and RISC-V architecture, chip, and IP wars by partnering with Japanese LSTC at 2nm processing nodes. Excerpts: "Tenstorrent, the firm led by legendary chip architect Jim Keller, the mastermind behind AMD's Zen architecture and Tesla's original self-driving chip, has launched its first hardware. Grayskull is a RISC-V alternative to GPUs that is designed to be easier to program and scale, and reportedly excels at handling run-time sparsity and conditional computation." ... "The Santa Clara-based tech firm's milestone launch comes hot on the heels of a partnership with Japan's Leading-edge Semiconductor Technology Center (LSTC). Tenstorrent's RISC-V and Chiplet IP will be used to build a state-of-the-art 2nm AI Accelerator, with the ultimate goal of revolutionizing AI performance in Japan." ... "Tenstorrent processors comprise a grid of cores known as Tensix Cores and come with network communication hardware so they can talk with one another directly over networks, instead of through DRAM." https://lnkd.in/guPMTNBT

The new favorite of the AI era: HBM, High Bandwidth Memory. AI artificial intelligence is waking up the growth momentum of the technology industry in waves. After a year and a half of blowing, from the production of GPU graphics chipsets by NVIDIA to server assembly factories known as the "AI Five Kings," including Wiwynn, Quanta, Inventec, Foxconn, and Gigabyte, to high-end power supplies such as Delta and heatsink modules such as Cooler Master. Now, this spring breeze has finally blown onto the face of the memory industry. Starting in the second half of 2023, a new term in the semiconductor industry emerged: HBM (High Bandwidth Memory), which is a type of DRAM (Dynamic Random-Access Memory). It adopts 3D stacking to become an efficient memory for storing running programs and data. HBM is essentially DRAM stacked together. Compared with traditional DRAM, HBM has advantages such as high bandwidth, low power consumption, and small size, but it is more expensive due to the complexity of the manufacturing process. Therefore, since the technological breakthrough in 2013, only high-end products have adopted it, and it is not the mainstream in the market. However, the new generation of AI chips (GPU/CPU) has a high demand for computing power and bandwidth to handle a large amount of parallel data. The stronger the computing power, the faster the data processing speed per second. The larger the bandwidth, the more data can be accessed per second.

Sam Altman, CEO of OpenAI, plans to gather a staggering $5 to $7 trillion to build a network of fabs to produce AI-specific chips, an ambitious initiative sparking industry attention. NVIDIA's Jensen Huang and renowned CPU developer Jim Keller, currently at Tenstorrent, emphasize the importance of architectural innovation in AI processors over sheer quantity. Keller expressed confidence in achieving Altman's plan for less than $1 trillion and called for simplifying the supply chain and accelerating chip speed. Tenstorrent's roadmap focuses on developing processors for AI and HPC applications, aiming to increase processing units and enhance performance efficiency. Altman's fundraising target far exceeds the current valuation of the global semiconductor industry, with chip manufacturers investing heavily in fabrication equipment. #AIChip #SemiconductorIndustry #Chip #Innovation #AIHardware #Processor #ChipFabrication #AIProcessor #fab #fabrication #semiconductormanufacturing #CPU #OpenAI #Nvidia #Tenstorrent

Interviewing Lien Wei-han, the Chief CPU Architect - A Discussion #TechTalksWithLien 🤝 Follow us on Discord 🔜: https://lnkd.in/gt823Zd3 _ ❇️ Summary: Tenstorrent, a Canadian AI startup, is developing AI chips and servers to challenge Nvidia, AMD, and Intel. Lien Wei-han, the chief CPU architect at Tenstorrent and former lead architect of Apple's M1 chip, discussed the AI revolution and the role of hardware in an interview. He described the current era as an unprecedented time in human history, compared to the first industrial revolution. Tenstorrent focuses on "heterogeneous computation," combining CPUs and NPUs to create a computing environment that excels in AI tasks and general computing. Their scalable and modular architecture, empowered by RISC-V, ensures power efficiency and enables collaborations with electronic supply chains in Taiwan and Korea. Tenstorrent is also collaborating on automotive-grade ICs for AI-powered autonomous driving systems. Functional safety is a crucial consideration in their chip design process. Hashtags: #chatGPT #ChiefCPUArchitectConversations #TechTalksWithLienWeiHan

👋 Hello all! I have some industry news to share with you today. 📰🔥 The talk of the tech town is the XC6SLX25T-2CSG324C chip. It's not just any chip; it's the superstar of AI! 🖥️🚀🧠 This chip is really making waves in the market currently. We've delved into the details in our latest blog post! Learn why this "little brain" is so important and creating such a buzz. 📚💡🔍 I'm curious, how do you see this chip shaping the future of AI? Comment below and let's discuss! #Rowsum #PCB #PCBA #Tech #AI #Innovation #Electronics #Microcontrollers #Silicon #Circuits #Technology #Business #ChinaManufacturing #MarketInsights #Quality #Efficiency #CustomerService

Nvidia And Synopsys Punctuate AI Chip Design And Acceleration Leadership 1. Synopsys and NVIDIA, two major players in the semiconductor industry, held conferences in Silicon Valley last week focused on AI-driven chip design. 2. Synopsys announced a new AI tool, 3DSO.ai, which optimizes chip designs in three dimensions and provides thermal analysis. 3. NVIDIA unveiled its new Blackwell GPU architecture for AI and showcased its Project Gr00T, which aims to create humanoid robots with advanced AI capabilities. 4. Blackwell GPUs offer 2.5 times more transistors than NVIDIA's previous architecture and can be configured into powerful supercomputers for AI processing. 5. The advancements made by both companies in AI and chip design highlight the rapid pace of innovation in the semiconductor industry. Source: https://lnkd.in/gRA2Xm4H