➀ Footage from an Indonesian factory mistakenly identified a GeForce RTX 4070 Ti SUPER as a GeForce RTX 5090; ➁ The RTX 4070 Ti SUPER SOLID is one of ZOTAC's largest GPUs; ➂ NVIDIA's GeForce RTX 5090 and GeForce RTX 5080 are expected to be unveiled at CES 2025.
Recent #GPU news in the semiconductor industry
➀ Intel's Arc Battlemage GPU leaks reveal the G31 and G21 SKUs with expected performance levels; ➁ AIBs show little interest in Arc, with AMD reportedly pushing Ryzen + RDNA 4 bundles; ➂ Intel faces challenges with Battlemage, including budget constraints and competition with RDNA 4.
➀ Nvidia's Ian Buck discusses the shift from focusing on single chips to integrated systems;
➁ The transformation of data centers into AI factories due to the rise of AI;
➂ The increasing power and cooling requirements of GPUs and Nvidia's involvement in the Coolerchips project.
➀ Samsung has developed next-gen 24Gbps GDDR7 memory modules; ➁ Offering higher VRAM capacities on future consumer GPUs; ➂ Up to 42.5Gbps speeds.
➀ Dell has launched the new PowerEdge XE9712 with NVIDIA GB200 NVL72 AI servers, offering 30x faster real-time LLM performance over the H100 AI GPU; ➁ The system features 72 x B200 AI GPUs connected with NVLink technology, providing lightning-fast connectivity; ➂ Dell highlights the liquid-cooled system for maximizing datacenter power utilization and rapid deployment of AI clusters.
➀ Intel and AMD are testing 12V-2x6 PCIe power connectors for their next-gen Radeon and Arc GPUs; ➁ The new standard is expected to be used in upcoming GPUs; ➂ Despite the new standard's benefits, Intel and AMD are hesitant to change due to the reliability of traditional connectors.
➀ The evolution of rendering technology with GPU hardware upgrades from GeForce 256 to RTX with ray tracing;
➁ The perspective of software developers in understanding rendering and the need for a deeper understanding from the GPU architecture angle;
➂ Overview of GPU architecture including global memory, stream multiprocessors (SM), and their components.
➀ Arm is exploring the feasibility of running LLMs on mobile devices; ➁ Arm's optimization techniques for LLMs on mobile; ➂ The importance of practical use cases for LLMs in mobile devices
➀ AMD has detailed the Instinct MI300X at Hot Chips 2024, with MI325X expected to be released soon.
➁ MI300X is a significant revenue source for AMD with over $4 billion in sales in the AI industry.
➂ AMD has acquired ZT Systems, the manufacturer of the Microsoft Azure MI300X platform.
➃ MI300X features a 192MB HBM3, a multi-chiplet chip for computing applications.
➄ AMD's CDNA 3 architecture has evolved with 8-stack HBM3 memory arrays, reaching 192GB in capacity.
➅ MI300X can operate as a single partition or across different memory and compute partitions.
➆ AMD's current major platform is the 8-way MI300X OAM platform.
➇ AMD discusses ROCm, which is improving.
➈ AMD's MI300X can compete with NVIDIA H100 in some cases.
➉ AMD is expected to release MI325X this year and Instinct MI350 288GB GPU in 2025.
➀ Computing power is an important indicator of a computer's information processing capability, with AI computing power focusing on AI applications, commonly measured in TOPS and TFLOPS, and provided by dedicated chips such as GPU, ASIC, and FPGA for algorithm model training and inference.
➁ AI chip accuracy is a way to measure computing power level, with FP16 and FP32 used in model training, and FP16 and INT8 used in model inference.
➂ AI chips typically use GPU and ASIC architectures. GPUs are the key components in AI computing due to their advantages in computation and parallel task processing.
➃ Tensor Core, an enhanced AI computing core compared to the parallel computation performance of Cuda Core, is more focused on the deep learning field and accelerates AI deep learning training and inference tasks through optimized matrix operations.
➄ TPUs, a type of ASIC designed for machine learning, stand out in high energy efficiency in machine learning tasks compared to CPUs and GPUs.
➀ NVIDIA's next-generation GeForce RTX 5070 is rumored to launch with 18GB of GDDR7 memory; ➁ Initially, it will be available in a 12GB model, breaking the VRAM mold; ➂ The 18GB version could be the RTX 5070 SUPER 18GB.
1. Nvidia's upcoming Blackwell GPU is expected to drive significant revenue growth due to high demand from Data Centers and hyperscalers; 2. CEO Jensen Huang highlighted 'insane demand' for Blackwell GPUs, indicating strong pricing power and revenue potential; 3. The author's conservative estimate suggests Nvidia could generate $80-120B in incremental revenue from Blackwell in FY 2025, significantly higher than consensus forecasts.
➀ AMD is rumored to be preparing the RX 7650 GRE for launch early next year; ➁ The GPU is unlikely to be revealed at CES 2025; ➂ It is expected to launch shortly after CES in China, aimed at clearing inventory of AMD's Navi 33-based GPUs.
➀ AMD's Navi 44 GPU package size is claimed to be 31% smaller than Navi 33; ➁ The package size of Navi 44 is 29mm x 29mm; ➂ AMD may be using a newer 5nm-class process node for Navi 44.
➀ The research investigates the effects of adding TiC nanoparticles to aluminum alloy 7075, aiming to enhance casting performance and improve fluidity and surface quality; ➁ TiC nanoparticles were introduced in two concentrations, and their impact on fluidity and microstructure was analyzed; ➂ The results showed a significant improvement in fluidity and surface quality, with finer grain sizes and smoother surfaces.
➀ NVIDIA's Blackwell AI GPUs are 'sold out' for the next 12 months due to high demand; ➁ The GPUs are in high demand among major tech companies like Google, Meta, and Microsoft; ➂ Despite the delay in production, Blackwell is expected to deliver significant performance improvements for AI training and inference.
➀ Leaker 'kopite7kimi' disputes the rumored $1999 to $2499 pricing for the RTX 5090; ➁ The RTX 5090 is expected to have significant performance upgrades over the RTX 4090; ➂ The RTX 5090 is rumored to feature 32GB GDDR7 memory, over 20K CUDA cores, and a 600W power requirement.
➀ NVIDIA's next-gen Blackwell Ultra 'B300' AI GPUs are rumored to use a socketed design on GB300 AI servers; ➁ This design change is expected to simplify maintenance and upgrades; ➂ The B300 series is set to become the mainstream product for NVIDIA in the second half of 2025.
➀ AMD launched the new AI GPU, AMD Instinct MI325X, an upgrade from the MI300X; ➁ AMD also announced its mid-2025 accelerator, the MI355X; ➂ The MI325X features 256GB of HBM3E memory and is designed for deployment in the Universal Baseboard (UBB) platform.
➀ Nvidia is considering a socketed design for its upcoming Blackwell B300 GPUs; ➁ This could allow users to remove the GPUs from the motherboard; ➂ The move aims to address high failure rates and cooling challenges in AI GPUs.