All articles tagged with #ai workloads
ASUS has launched pre-orders for its ProArt GR1X mini-PC, powered by the NVIDIA RTX Spark SoC. The company markets the device as an efficient gaming powerhouse, claiming it can achieve 100 FPS at 1440p resolution with a 140W power limit. While the system supports high-end creative tasks and local AI models, questions remain regarding the specific upscaling technologies used to reach these frame rates.
Intel’s Q2 2026 results show server CPU ASP up 48% year over year due to a higher mix of premium products and demand-based pricing, while data center revenue rose 59% to $6.3 billion as AI workloads drive demand; server volume grew 9% year over year but supply constraints kept product availability tight, with Granite Rapids demand described as extremely tight.
Apple’s A20 Pro will not feature LPDDR6 RAM; instead it supports six-channel LPDDR5X on a 96-bit bus, boosting memory bandwidth to meet AI workloads, with DRAM separated from the die via WMCM packaging to improve heat dissipation and sustained performance.
IBM announced a sub-1‑nanometer NanoStack transistor architecture that stacks nanosheets in a 3D structure to fit almost 100 billion transistors on a fingernail‑size die at a 0.7 nm class, aiming for up to 50% higher performance or 70% lower power versus its 2 nm node and a 40% SRAM density boost. The platform is designed for AI and data-center workloads and relies on advanced lithography and materials (High-NA EUV, new dielectrics), with IBM targeting a production path at sub‑1 nm within about five years through partnerships like Rapidus. While presented as a generational leap, it remains a research milestone rather than immediate mass production.
At CES 2026, Acemagic showcased mini PCs featuring AMD's Gorgon Point and Intel's Panther Lake chips, including a retro-inspired AMD Ryzen AI 9 HX 470-based system and a modern Intel Core Ultra X9 388H-based model, with high memory capacities and AI-focused configurations, highlighting powerful compact computing solutions.
ZLUDA has released version 6-preview.48, now supporting CUDA 13.1, enabling unmodified CUDA applications to run on non-NVIDIA GPUs, with a focus on AI workloads and broader hardware compatibility.
Applied Digital and EKSO announced a non-binding agreement to combine their cloud and exoskeleton businesses to form ChronoScale, a platform focused on high-performance GPU infrastructure for AI workloads, aiming to meet growing demand for specialized AI compute solutions. The transaction is expected to close in the first half of 2026, with Applied Digital owning about 97% of the new company.
The article discusses how the energy crisis is prompting a reimagining of data centers, with innovative concepts like underground, suspended, and space-based data centers being explored to improve sustainability and efficiency, though widespread adoption faces regulatory, technical, and cost challenges.
Bit-Brick Cluster K1 is a versatile 4-slot RISC-V cluster board designed for edge computing, AI, and embedded applications, supporting up to four SSOM-K1 modules with high-performance features, extensive I/O options, and software support for various operating systems, available for around $125 without modules.
Cisco has launched the Cisco Unified Edge platform, an integrated, modular system that extends data center capabilities to the edge, supporting real-time AI inferencing and agentic workloads across various industries with enhanced performance, security, and operational simplicity.
The M5 chip offers significant performance improvements over the M4, especially in AI, graphics, and memory bandwidth, making it ideal for demanding tasks, though for everyday use the difference may be less noticeable. It is currently available in select Apple devices like the 14-inch MacBook Pro, iPad Pro, and Vision Pro, with future models expected to feature higher-end variants.
Microsoft has announced the deployment of its first large-scale AI 'factory' using over 4,600 Nvidia GPUs, with plans to expand globally, highlighting its existing extensive data center infrastructure and readiness to support advanced AI models, amidst competition with OpenAI and Nvidia.
AMD is using stock warrants as a novel way to finance OpenAI's multi-billion dollar GPU purchases, with the potential for OpenAI to hold a stake worth around $100 billion if certain stock price milestones are met. This arrangement serves as a validation of AMD's AI GPU capabilities and could boost its market share, although OpenAI may sell the stock along the way to fund the purchases. The deal highlights innovative financing in the AI hardware industry and AMD's strategic push into AI data centers.
The GMKtec EVO-X2 mini PC, powered by the powerful AMD Ryzen AI Max+ 395 with 16 cores and Radeon 8060S graphics, offers desktop-class performance in a compact form factor, making it an excellent choice for AI, gaming, and as a desktop upgrade, especially at its current Prime Day price of $1,499.89.
Google's cloud computing business reported a 32% revenue increase in Q2, driven by AI workloads, leading to positive Wall Street reactions and a stock rise. The company also raised its 2025 capital spending guidance to $85 billion, emphasizing AI infrastructure. Analysts see Google Cloud as a strong competitor to Amazon and Microsoft, with growth prospects supported by high demand and expanding backlog.