All articles tagged with #cuda
NVIDIA announced that they will bring their CUDA software to RISC-V processors, expanding support beyond x86_64, AArch64, and previously IBM POWER, reflecting the growing interest in RISC-V for data centers. The Linux driver stack is portable, and similar support has been seen with AMD's ROCm on RISC-V and LoongArch processors.
British startup Spectral Compute has introduced "SCALE," a GPGPU toolchain that enables NVIDIA's CUDA to run directly on AMD GPUs without code porting. This development challenges NVIDIA's software exclusivity by allowing a single codebase to function across multiple hardware platforms. Despite NVIDIA's resistance, SCALE aims to bridge the compatibility gap and has been tested on various applications using AMD's RDNA architectures.
Nvidia's market cap could surge to $10 trillion by 2030, driven by its dominant position in AI chips and software, according to analyst Beth Kindig. The company's hardware, including upcoming Blackwell chips, and its CUDA platform provide a significant competitive edge. Nvidia's role in AI infrastructure and the automotive sector further bolster its growth prospects.
Nvidia's stock is projected to surge 258% and reach a $10 trillion valuation by 2030, driven by its next-generation Blackwell GPU and CUDA software platform, according to I/O Fund tech analyst Beth Kindig. Nvidia's strong market position and "impenetrable moat" around its GPU business are expected to capture a significant share of the AI data center market, outpacing competitors like AMD and Intel.
Nvidia's stock is predicted to surge 258% by 2030, reaching a $10 trillion valuation, driven by its next-generation Blackwell GPU and CUDA software platform. Analyst Beth Kindig highlights Nvidia's strong market position and its significant role in the AI data center market, projecting substantial revenue growth and an "impenetrable moat" around its business.
Amazon is struggling to compete with Nvidia's dominant AI chips, facing low usage and compatibility issues with its Trainium and Inferentia chips, which threatens AWS's profitability. Despite efforts to innovate and feedback from customers, Nvidia's established CUDA platform remains a significant barrier. Amazon's AI chips have seen limited adoption, and even internal projects often rely on Nvidia GPUs. The company aims to improve its AI chip offerings and work more closely with the open-source community to gain a larger market share.
NVIDIA has launched enterprise-grade generative AI microservices that enable businesses to create and deploy custom applications on their own platforms while retaining ownership of their intellectual property. These microservices, built on the NVIDIA CUDA platform, include NIM microservices for optimized inference on popular AI models and CUDA-X microservices for data processing, retrieval-augmented generation, and more. The microservices are adopted by leading application platform providers and can be accessed through NVIDIA AI Enterprise 5.0, offering a standardized path to run custom AI models optimized for NVIDIA's CUDA installed base.
Nvidia has updated its licensing terms to ban running CUDA-based software on non-Nvidia hardware using translation layers, potentially targeting initiatives like ZLUDA and Chinese GPU makers. This move aims to protect Nvidia's dominance in the accelerated computing space, particularly with AI applications, but may also impact competition from AMD, Intel, and other developers of AI compute hardware. While recompiling existing CUDA programs remains legal, the ban on translation layers could affect the competitive positioning of other hardware platforms in the GPGPU arena.
ZLUDA, a project funded by AMD, enables unmodified CUDA code to run on AMD Radeon GPUs, allowing native CUDA applications to run on AMD hardware with near-native performance. Despite AMD discontinuing its support for the project, ZLUDA has been released as open-source software. While ZLUDA cannot run Nvidia Optix code, it has shown improved performance over AMD's Radeon HIP code in various applications. The developer plans to continue working on ZLUDA, focusing on personal interests without financial backing from AMD.
AMD's ROCm now indirectly supports NVIDIA CUDA libraries through the open-source ZLUDA project, allowing AMD GPU owners to run CUDA apps within ROCm without code adaptation. ZLUDA, originally designed for NVIDIA CUDA support on Intel GPUs, has been successfully tested with AMD's Radeon GPUs on ROCm, enabling performance on par with NVIDIA for rendering tasks. While native CUDA support on Radeon GPUs remains unavailable, the potential of ZLUDA in bridging ROCm and CUDA stacks for AI applications is promising, with the developer exploring NVIDIA's upscaling capabilities on RDNA GPUs.
AMD quietly funded a project to create a drop-in replacement for CUDA on AMD GPUs, allowing many CUDA applications to run on the ROCm stack without modification. The project, based on the ZLUDA implementation, was developed by Andrzej Janik and is now open-source, dual-licensed under Apache 2.0 or MIT. While it's a promising solution, it's not without limitations, such as incomplete NVIDIA OptiX support. The implementation, leveraging the Rust programming language, has been tested successfully, and the exposed device name for Radeon GPUs via CUDA will be updated as part of the open-sourcing.
Intel's CEO, Pat Gelsinger, stated that the entire industry is motivated to eliminate NVIDIA's CUDA dominance in the AI market. Intel believes that the future of AI lies in inference rather than training models and aims to prioritize inference developments. Gelsinger sees NVIDIA's success as a temporary "bubble" and believes that the industry will adopt new training methods to bring a broader set of technologies. Intel praised its OpenVINO model and aims to transition towards next-gen markets. However, Intel needs to do more work to challenge CUDA's dominance, and for now, NVIDIA remains the leader in the AI segment.