High Bandwidth Memory (HBM) is emerging as the preferred solution for overcoming memory transfer speed restrictions due to the bandwidth limitations of DDR SDRAM in high-speed computation. HBM is recognized for its revolutionary transmission efficiency and plays a pivotal role in allowing core computational components to operate at their maximum capacity. Top-tier AI server GPUs have set a new industry standard by primarily using HBM. TrendForce forecasts that global demand for HBM will experience almost 60% growth annually in 2023, reaching 290 million GB, with a further 30% growth in 2024.

TrendForce's forecast for 2025, taking into account five large-scale AIGC products equivalent to ChatGPT, 25 mid-size AIGC products from Midjourney, and 80 small AIGC products, the minimum computing resources required globally could range from 145,600 to 233,700 Nvidia A100 GPUs. Emerging technologies such as supercomputers, 8K video streaming, and AR/VR, among others, are expected to simultaneously increase the workload on cloud computing systems due to escalating demands for high-speed computing.

A group of researchers in China have used a new approach to AI to create a full RISC-V processor from scratch. The team set out to answer the question of whether an AI could design an entire processor on its own without human intervention. While AI design tools do already exist and are used for complex circuit design and validation today, they are generally limited in use and scope. The key improvements shown in this approach over traditional or AI assisted logic design are the automated capabilities, as well as its speed. The traditional assistive tools for designing circuits still require many hours of manual programming and validation to design a functional circuit. Even for such a simple processor as the one created by the AI, the team claims the design would have taken 1000x as long to be done by humans. The AI was trained by observing specific inputs and outputs of existing CPU designs, with the paper summarizing the approach as such:

(...) a new AI approach, which generates large-scale Boolean function with almost 100% validation accuracy (e.g., > 99.99999999999% as Intel) from only external input-output examples rather than formal programs written by the human. This approach generates the Boolean function represented by a graph structure called Binary Speculation Diagram (BSD), with a theoretical accuracy lower bound by using the Monte Carlo based expansion, and the distance of Boolean functions is used to tackle the intractability.

Qualcomm Technologies, Inc. has announced the new Snapdragon 4 Gen 2 Mobile Platform, which has been creatively engineered to make incredible mobile experiences accessible to more consumers globally. Snapdragon 4 Gen 2 provides effortless, all-day use with fast CPU speeds, sharp photography and videography, plus speedy 5G and Wi-Fi for reliable connections.

"Snapdragon - at its core - is driving innovation while meeting the demands of OEMs and the broader industry," said Matthew Lopatka, director of product management, Qualcomm Technologies, Inc. "With this generational advancement in the Snapdragon 4-series, consumers will have greater access to the most popular and relevant mobile features and capabilities. We optimized every aspect of the platform in order to maximize the experiences for users."

Oracle today announced the addition of generative AI-powered capabilities within Oracle Fusion Cloud Human Capital Management (HCM). Supported by the Oracle Cloud Infrastructure (OCI) generative AI service, the new capabilities are embedded in existing HR processes to drive faster business value, improve productivity, enhance the candidate and employee experience, and streamline HR processes.

"Generative AI is boosting productivity and unlocking a new world of skills, ideas, and creativity that can have an immediate impact in the workplace," said Chris Leone, executive vice president, applications development, Oracle Cloud HCM. "With the ability to summarize, author, and recommend content, generative AI helps to reduce friction as employees complete important HR functions. For example, with the new embedded generative AI capabilities in Oracle Cloud HCM, our customers will be able to take advantage of large language models to drastically reduce the time required to complete tasks, improve the employee experience, enhance the accuracy of workforce insights, and ultimately increase business value."

What's New: As part of a collaboration with Intel and Dell Technologies, the United Kingdom Atomic Energy Authority (UKAEA) and the Cambridge Open Zettascale Lab plan to build a "digital twin" of the Spherical Tokamak for Energy Production (STEP) prototype fusion power plant. The UKAEA will utilize the lab's supercomputer based on Intel technologies, including 4th Gen Intel Xeon Scalable processors, distributed asynchronous object storage (DAOS) and oneAPI tools to streamline the development and delivery of fusion energy to the grid in the 2040s.

"Planning for the commercialization of fusion power requires organizations like UKAEA to utilize extreme amounts of computational resources and artificial intelligence for simulations. These HPC workloads may be performed using a variety of different architectures, which is why open software solutions that optimize performance needs can lend portability to code that isn't available in closed, proprietary systems. Overall, advanced hardware and software can make the journey to commercial fusion power lower risk and accelerated - a key benefit on the path to sustainable energy."—Adam Roe, Intel EMEA HPC technical director

New research from Omdia reveals that the semiconductor market declined in revenue for a fifth straight quarter in the first quarter of 2023. This is the longest recorded period of decline since Omdia began tracking the market in 2002. Revenue in 1Q23 settled at $120.5B, down 9% from 4Q22. The semiconductor market is cyclical, and this prolonged decline follows the upsurge as the market grew to record revenues in each quarter between 4Q20 through 4Q21 following increased demand from the global pandemic.

The memory and MPU market are major areas of the semiconductor market that are contributing to the decline. The MPU market in 1Q23 was $13.1B, just 65% of its size in 1Q22 when it was $20B. The memory market fared worse, with 1Q23 coming in at $19.3B, just 44% of the market in 1Q22 when it was $43.6B. The combined MPU and memory markets declined 19% in 1Q23, dragging the market down to the 9% quarter-over-quarter (QoQ) decline.

According to Taiwanese media, TSMC will start production of its first 1.4 nm fab in 2026, with chip production in the fab said to start sometime in 2027 or 2028. The new fab will be located in Longtan Science Park outside of Hsinchu in Taiwan, where many of TSMC's current fabs are located. TSMC is currently constructing a 2 nm and below node R&D facility at a nearby plot of land to where the new fab is expected to be built. This facility is expected to be finished in 2025 and TSMC has been allocated a total area of just over 158 hectares of land for future expansion in the area.

In related news, TSMC is expected to be charging US$25,000 per 2 nm GAA wafer, which is an increase of about a fifth compared to its 3 nm wafers which are going for around US$20,000. This is largely due to the nodes being fully booked and TSMC being able to charge a premium for its cutting edge nodes. TSMC is also expanding in CoWoS packaging facilities due to increased demand from both AMD and NVIDIA for AI related products. Currently TSMC is said to be able to output 12,000 CoWoS wafers per month and this is twice as much as last year, yet TSMC is unable to meet demand from its customers.

Samsung Electronics, a world leader in advanced semiconductor technology, today announced its latest foundry technology innovations and business strategy at the 7th annual Samsung Foundry Forum (SFF) 2023. Under the theme "Innovation Beyond Boundaries," this year's forum delved into Samsung Foundry's mission to address customer needs in the artificial intelligence (AI) era through advanced semiconductor technology.

Over 700 guests, from customers and partners of Samsung Foundry, attended this year's event, of which 38 companies hosted their own booths to share the latest technology trends in the foundry industry.

In a new industry-standard benchmark, a cluster of 3,584 H100 GPUs at cloud service provider CoreWeave trained a massive GPT-3-based model in just 11 minutes. Leading users and industry-standard benchmarks agree: NVIDIA H100 Tensor Core GPUs deliver the best AI performance, especially on the large language models (LLMs) powering generative AI.

H100 GPUs set new records on all eight tests in the latest MLPerf training benchmarks released today, excelling on a new MLPerf test for generative AI. That excellence is delivered both per-accelerator and at-scale in massive servers. For example, on a commercially available cluster of 3,584 H100 GPUs co-developed by startup Inflection AI and operated by CoreWeave, a cloud service provider specializing in GPU-accelerated workloads, the system completed the massive GPT-3-based training benchmark in less than eleven minutes.

Today, MLCommons published results of its industry AI performance benchmark, MLPerf Training 3.0, in which both the Habana Gaudi2 deep learning accelerator and the 4th Gen Intel Xeon Scalable processor delivered impressive training results.

"The latest MLPerf results published by MLCommons validates the TCO value Intel Xeon processors and Intel Gaudi deep learning accelerators provide to customers in the area of AI. Xeon's built-in accelerators make it an ideal solution to run volume AI workloads on general-purpose processors, while Gaudi delivers competitive performance for large language models and generative AI. Intel's scalable systems with optimized, easy-to-program open software lowers the barrier for customers and partners to deploy a broad array of AI-based solutions in the data center from the cloud to the intelligent edge." - Sandra Rivera, Intel executive vice president and general manager of the Data Center and AI Group

AMD today announced the AMD Versal Premium VP1902 adaptive system-on-chip (SoC), the world's largest adaptive SoC. The VP1902 adaptive SoC is an emulation-class, chiplet-based device designed to streamline the verification of increasingly complex semiconductor designs. Offering 2X the capacity over the prior generation, designers can confidently innovate and validate application-specific integrated circuits (ASICs) and SoC designs to help bring next generation technologies to market faster.

AI workloads are driving increased complexity in chipmaking, requiring next-generation solutions to develop the chips of tomorrow. FPGA-based emulation and prototyping provides the highest level of performance, allowing faster silicon verification and enabling developers to shift left in the design cycle and begin software development well before silicon tape-out. AMD, through Xilinx, brings over 17 years of leadership and six generations of the industry's highest capacity emulation devices, which have nearly doubled in capacity each generation.

Metalenz, the world leader in metasurface optics, today announced it has partnered with leading semiconductor foundry United Microelectronics Corporation ("UMC") to release its direct supply chain to mass production and bring the unrivaled scale and precision of semiconductor manufacturing to the optics industry. The announcement marks the launch of metasurface optics on the open market for the first time and follows multiple design wins for Metalenz with leading OEMs in Asia.

"After initially designing meta-optics in partnership with one of the leading suppliers of 3D sensing solutions, we are now engaged with OEMs directly to bring the benefits of metasurface optics to their 3D sensing applications. By partnering with a world-class foundry like UMC, we gain the manufacturing capabilities, expertise, and global reach to serve customers interested in adopting our meta-optics technology," said Rob Devlin, Co-founder and CEO of Metalenz. "This will further accelerate our growth as we are becoming the leading provider of precision optics for 3D sensing solutions."

A new global study by the IBM Institute for Business Value found that nearly half of CEOs surveyed identify productivity as their highest business priority—up from sixth place in 2022. They recognize technology modernization is key to achieving their productivity goals, ranking it as second highest priority. Yet, CEOs can face key barriers as they race to modernize and adopt new technologies like generative AI.

The annual CEO study, CEO decision-making in the age of AI, Act with intention, found three-quarters of CEO respondents believe that competitive advantage will depend on who has the most advanced generative AI. However, executives are also weighing potential risks or barriers of the technology such as bias, ethics and security. More than half (57%) of CEOs surveyed are concerned about data security and 48% worry about bias or data accuracy.

NVIDIA's compute and AI-oriented H100 GPU is supposedly getting an upgrade. The H100 GPU is NVIDIA's most powerful offering and comes in a few different flavors: H100 PCIe, H100 SXM, and H100 NVL (a duo of two GPUs). Currently, the H100 GPU comes with 80 GB of HBM2E, both in the PCIe and SXM5 version of the card. A notable exception if the H100 NVL, which comes with 188 GB of HBM3, but that is for two cards, making it 94 GB per each. However, we could see NVIDIA enable 94 and 64 GB options for the H100 accelerator soon, as the latest PCI ID Repository shows.

According to the PCI ID Repository listing, two messages are posted: "Kindly help to add H100 SXM5 64 GB into 2337." and "Kindly help to add H100 SXM5 94 GB into 2339." These two messages indicate that NVIDIA could prepare its H100 in more variations. In September 2022, we saw NVIDIA prepare an H100 variation with 120 GB of memory, but that still isn't official. These PCIe IDs could just come from engineering samples that NVIDIA is testing in the labs, and these cards could never appear on any market. So, we have to wait and see how it plays out.

NVIDIA DLSS 3 is a neural graphics technology that multiplies performance using AI image reconstruction and frame generation. It's a combination of three core innovations:

• Super Resolution uses deep learning algorithms to upscale a lower-resolution input into a higher-resolution output, creating a sharp image with a boosted frame rate.
• Frame Generation uses AI rendering to generate entirely new frames with best-in-class quality and responsiveness.
• NVIDIA Reflex is a low-latency technology that minimizes input lag by synchronizing the CPU and the GPU for optimal responsiveness.

Intel Labs, in collaboration with Blockade Labs, has introduced Latent Diffusion Model for 3D (LDM3D), a novel diffusion model that uses generative AI to create realistic 3D visual content. LDM3D is the industry's first model to generate a depth map using the diffusion process to create 3D images with 360-degree views that are vivid and immersive. LDM3D has the potential to revolutionize content creation, metaverse applications and digital experiences, transforming a wide range of industries, from entertainment and gaming to architecture and design.

"Generative AI technology aims to further augment and enhance human creativity and save time. However, most of today's generative AI models are limited to generating 2D images and only very few can generate 3D images from text prompts. Unlike existing latent stable diffusion models, LDM3D allows users to generate an image and a depth map from a given text prompt using almost the same number of parameters. It provides more accurate relative depth for each pixel in an image compared to standard post-processing methods for depth estimation and saves developers significant time to develop scenes," said Vasudev Lal, AI/ML research scientist, Intel Labs.

NVIDIA's H100 Hopper GPU is a device designed for pure AI and other compute workloads, with the least amount of consideration for gaming workloads that involve graphics processing. However, it is still interesting to see how this 30,000 USD GPU fairs in comparison to other gaming GPUs and whether it is even possible to run games on it. It turns out that it is technically feasible but not making much sense, as the Chinese YouTube channel Geekerwan notes. Based on the GH100 GPU SKU with 14,592 CUDA, the H100 PCIe version tested here can achieve 204.9 TeraFLOPS at FP16, 51.22 TeraFLOPS at FP32, and 25.61 TeraFLOPS at FP64, with its natural power laying in accelerating AI workloads.

However, how does it fare in gaming benchmarks? Not very well, as the testing shows. It scored 2681 points in 3DMark Time Spy, which is lower than AMD's integrated Radeon 680M, which managed to score 2710 points. Interestingly, the GH100 has only 24 ROPs (render output units), while the gaming-oriented GA102 (highest-end gaming GPU SKU) has 112 ROPs. This is self-explanatory and provides a clear picture as to why the H100 GPU is used for computing only. Since it doesn't have any display outputs, the system needed another regular GPU to provide the picture, while the computation happened on the H100 GPU.

Intel Graphics today released the latest version of Arc GPU Graphics Drivers. Version 101.4499 beta adds optimization for F1 23, Aliens: Dark Descent, Forever Skies, and Counter Strike 2. There are major performance uplifts to be had. Counter Strike 2 sees up to 8% uplift at 1440p with high settings, and up to 10% at 1080p with Very High settings. F1 23 players can expect up to 33% uplift at 1080p with Ultra High settings and RT on, and an 18% uplift with RT off; and a 27% uplift at 1440p with high settings. Intel also updated its Destiny 2 optimization, with a neat 11% uplift to be had at 1080p with the highest settings. An error or black screen seen on applications embedding WebView2 frames, has been fixed.

DOWNLOAD: Intel GPU Graphics Drivers 101.4499 Beta

The European parliament has voted today on a proposed set of rules that aim to govern artificial intelligence development in the region. The main branch has approved the text of draft of this legislation—a final tally showed participant counts of 499 in favor, and 28 against, and 93 abstentions at the Strasbourg HQ-based meeting. The so called "AI Act" could be a world first as well as a global standard for regulation over AI technology—members of the European Parliament (MEPs) are expected to work on more detailed specifics with all involved countries before new legislation is set in stone.

Thierry Breton, the European commissioner for the internal market stated today: "AI raises a lot of questions socially, ethically, economically. But now is not the time to hit any 'pause button'. On the contrary, it is about acting fast and taking responsibility." The council is aiming to gain control of several fields of AI applications including drone operation, automated medical diagnostic equipment, "high risk" large language models and deepfake production methods. Critics of AI have reasoned that uncontrolled technological advancements could enable computers to perform tasks faster than humans—thus creating the potential for large portions of the working population to become redundant.

AMD today launched its Ryzen 7040HS line of mobile processors targeting consumer notebooks of conventional thickness and portability, which AMD considers thin-and-light. This class of devices is positioned between ultraportable notebooks, and gaming notebooks or portable workstations. AMD already powers several segments of gaming notebooks and portable workstations with its Ryzen 7045HX series "Dragon Range" mobile processors, with CPU core counts ranging between 6 and 16; as well as the ultraportable segment with the Ryzen 7040U series; but while the 7045 series have their TDP rated in the 45 W to 65 W range, and the Ryzen 7040U in the 15 W to 28 W range, the company was lacking a current-generation processor lineup in the 35 W to 54 W segment, which the company is filling up with today's Ryzen 7040HS series launch.

The Ryzen 7040HS series processors are based on the 4 nm "Phoenix" monolithic silicon, just like the 7040U series, and is based on a combination of "Zen 4" microarchitecture for its CPU, RDNA3 graphics architecture for its iGPU, and the new XDNA architecture for its Ryzen AI on-chip accelerator. Physically, the "Phoenix" silicon features an 8-core/16-thread "Zen 4" CPU. Each core has 1 MB of dedicated L2 cache, and a 16 MB L3 cache is shared among the 8 cores. The iGPU features 12 RDNA3 compute units, which amount to 768 stream processors, along with 24 AI Accelerators (specific to the RDNA3 architecture); 24 Ray Accelerators, 48 TMUs, and 32 ROPs. The iGPU meets the full DirectX 12 Ultimate logo requirements. The Ryzen AI XDNA accelerator features 20 AI acceleration tiles, each with local memory.

Today, at the "Data Center and AI Technology Premiere," AMD announced the products, strategy and ecosystem partners that will shape the future of computing, highlighting the next phase of data center innovation. AMD was joined on stage with executives from Amazon Web Services (AWS), Citadel, Hugging Face, Meta, Microsoft Azure and PyTorch to showcase the technological partnerships with industry leaders to bring the next generation of high performance CPU and AI accelerator solutions to market.

"Today, we took another significant step forward in our data center strategy as we expanded our 4th Gen EPYC processor family with new leadership solutions for cloud and technical computing workloads and announced new public instances and internal deployments with the largest cloud providers," said AMD Chair and CEO Dr. Lisa Su. "AI is the defining technology shaping the next generation of computing and the largest strategic growth opportunity for AMD. We are laser focused on accelerating the deployment of AMD AI platforms at scale in the data center, led by the launch of our Instinct MI300 accelerators planned for later this year and the growing ecosystem of enterprise-ready AI software optimized for our hardware."

MetaX a new Chinese GPU developer had physical products on display at the recent International Social Public Safety Products Expo—ITHome noted that the company's first ever offering is the N100 GPGPU, specifically designed for artificial intelligence and video processing tasks. MetaX's Yang Jian, a company co-founder, said that his team's primary focus is on AI and data center applications—presumably due to recent demand presenting a lucrative prospect. China's access to Western-developed hardware has been heavily restricted, so native silicon developers have received heavy investment from the State. A MetaX MXN single-slot low-profile card sporting their N100 GPGPU and onboard HBM2E memory is said to offer compute performance of 160 INT8 TOPS and 80 FP16 TFLOPS.

MetaX is also targeting an entry into the gaming GPU market—ITHome reckons that non-enterprise cards are due in 2025. The company's website outlines an MXG-series for graphics rendering, but the information presented indicates that these products are targeting cloud gaming and data center sectors. The company likely requires more time and experience to develop complicated software that will be necessary for the operation of desktop graphics solutions. A similar Chinese chip startup - Biren Technology - has been working on BR100 and BR104 GPGPUs, in a bid to take on NVIDIA's downgraded Hopper H100 accelerators for the region. Biren's engineering team faced challenging circumstances when sanctions blocked their access to TSMC's foundry, plus the departure of their chief architect, who was alleged to be working on an unreleased range of mainstream/gaming graphics cards.

With the launch of 4th Gen Intel Xeon Scalable processors in January 2023, Intel delivered significant advancements in performance with industry-leading Intel accelerator engines and improved performance per watt across key workloads like AI, data analytics, high performance computing (HPC) and others. The industry has taken notice: 4th Gen Xeon has seen a rapid ramp, global customer adoption and leadership performance on a myriad of critical workloads for a broad range of business use cases.

Today, after weeks of rigorous and comprehensive head-to-head testing against the most comparable competitive processors, Intel is sharing compelling results that go far beyond simple industry benchmarks.

As many of you have seen or noticed, our credits in Firmament mention "AI Assisted Content". This has been present in the credits for Firmament since release day, and we have never hidden this information: "The contents listed as "AI Assisted" in the credits are "Journals, logs, checklists, newspapers, stories, songs, poems, letters, loosely scattered papers; all backer portraits; all founders portraits; the "sunset" paintings; the art-nouveau wallpaper in the Swan dormitory hallways; propaganda banners; coastal spill decal kit; all voiced mentor, announcer, founder, and other speeches; backer-exclusive content."

"AI Assisted" does not mean wholly AI-generated. Unfortunately, there have been articles published recently which have implied (especially in their headlines) that Cyan generated much of Firmament using AI tools. This is categorically false and misleading, and we are disappointed and frustrated to see this happening. Some folks may be concerned about our usage of AI, so in the interest of clarification: The voice performances in Firmament were voiced 100% of the time by a talented member of our development team who elected not to be credited by name. Their voice was simply modulated for the final product with one of these tools (and with their full permission and control). This same member of the development team has elected not to be credited in prior games of ours as well, for privacy concerns, and not anything to do with tools usage in our games.

According to a new research report from the analyst firm Jon Peddie Research, unit shipments in the add-in board (AIB) market decreased in Q1 2023 by -12.6% and decreased by -38.2% year to year. Intel increased its add-in board market share by 2% during the first quarter.

The percentage of AIBs in desktop PCs is referred to as the attach rate. The attach rate grew from last quarter by 8% but was down -21% year to year. Approximately 6.3 million add-in boards shipped in Q1 2023. The market shares for the desktop discrete GPU suppliers shifted in the quarter, as AMD's market share remained flat from last quarter. Intel, which entered the AIB market in Q3'22 with the Arc A770 and A750, gained 2% in market share, while Nvidia retains its dominant position in the add-in board space with an 84% market share.