Nvidia on July 1 announced software optimizations that boost token throughput by up to 5x for DeepSeek V4 on Blackwell systems, with throughput improvements reaching 20x compared to baseline configura…
Discarded NVIDIA enterprise Tesla GPUs, including K80 (24 GB, ~$60), P100 (16 GB, ~$75), and V100 (16 GB, under $200), remain usable for local AI inference and training workloads when paired with olde…
TornadoVM now natively integrates NVIDIA CUDA libraries cuBLAS, cuFFT, and cuDNN, allowing Java applications to call GPU-accelerated linear algebra and deep learning primitives directly from JIT-compi…
Morgan Stanley and Goldman Sachs warn that AI chip revenue forecasts for 2026 and 2027 have outpaced realistic capacity and demand trajectories, with sustained annual growth exceeding 30% seen as unsu…
A new GPU inference method for large language models (LLMs) with moderate unstructured sparsity (around 50%) outperforms dense matrix multiplication on modern GPUs with high-bandwidth memory, achievin…
A hyper-optimized, zero-dependency C/CUDA inference engine for the Qwen 3.6 35B model on RTX 5090 Blackwell GPUs achieves 13.4k tokens/sec prefill throughput at 2,048 context depth and 270+ tokens/sec…
A benchmark comparing Ollama and llama-server on a Tesla V100 found that both backends achieve nearly identical token generation speeds (about 111 tok/s), but llama-server processed prompts 22% faster…
A developer released an experimental Godot 4.7 project that runs Gemma 4 inference entirely in GDScript and Vulkan compute shaders, without external dependencies. The project is about an order of magn…
Mariana Souza published a practical guide for fine-tuning Qwen2.5-7B-Instruct using QLoRA on custom instruction datasets, including cost estimates and a loss-masking sanity check. The tutorial covers …
A software engineer and technical program manager documents the first week of a 32-week journey from AI-API consumer to local LLM systems architect using an NVIDIA DGX Spark. The engineer inventories …
Hugging Face announced that custom GPU kernels on its Hub are now signed artifacts governed by a trusted publisher model, requiring a dedicated repository type that replaces the old model-type format.…
PolymathicAI released The Well, a 15TB collection of 16 physics simulation datasets for machine learning, covering domains like fluid dynamics and supernova explosions. The open-source repository prov…
Researchers have developed a new approach to 1-WL stable coloring for Graph Neural Networks that breaks traditional scalability bottlenecks by using a probabilistically backed refinement algorithm. A …
NVIDIA published a technical blog post explaining how kernel fusion in CUDA can optimize GPU code by combining multiple operations into a single kernel, reducing memory traffic and launch overhead. Th…
A developer released a real-time n-body simulation using the Barnes-Hut algorithm on GPU, implemented in CUDA C++ and OpenGL, scaling to millions of particles on an NVIDIA RTX 500 Ada Laptop GPU. The …
Researchers introduced DominoTree, a training-free best-first draft tree method for speculative decoding that uses Domino's conditional correction to achieve up to 6.6x speedup over autoregressive dec…
NVIDIA announced the BioNeMo Agent Toolkit to accelerate biomolecular structure prediction and co-folding with OpenFold3, using GPU-accelerated MSA generation and multi-GPU scaling to enable virtual s…
NVIDIA expanded partnerships with cloud providers in Q1 2026 to offer scalable AI computing solutions, inviting a new class of partners to build on its accelerated compute infrastructure. The company'…
NVIDIA, AMD, and Intel compete in the 2026 AI GPU market, with NVIDIA's Blackwell RTX 50-series, AMD's Radeon AI Pro R9700, and Intel's Arc Pro B70 targeting local LLM inference. VRAM capacity and mem…
France's antitrust regulator is finalizing an investigation into Nvidia over alleged anti-competitive practices tied to its GPU dominance and CUDA software ecosystem, with potential fines of up to 10%…