NVIDIA
NVIDIA is an American semiconductor and artificial intelligence technology company headquartered in Santa Clara, California. Its core business involves designing and selling graphics processing units (GPUs), with products used in gaming, data centers, autonomous driving, and AI computing. NVIDIA also provides the CUDA parallel computing platform and a related software ecosystem, making it a major supplier of high-performance computing and deep learning infrastructure globally.
Core Facts
Open weights differ from fully open source: the former makes model parameters public but does not necessarily disclose training data and complete training code.
The NVIDIA A100's HBM2e capacity is 80GB, the H100 has 80GB HBM3, and the H200 has 141GB.
The HBM3 memory used by NVIDIA's H100 has a bandwidth of up to 3.35 TB/s.
Timeline (last 90 days)
Nvidia's Nimitron team released the unified diffusion language model NLD-Ni, where the same model can both understand visual content and complete generation tasks, and it officially says both aspects reach state-of-the-art levels; the model and code have been open-sourced.
NVIDIA released the 30-billion-parameter mixture-of-experts speech recognition model Nemotron 3.5 Transcribe, which activates only 4 billion parameters during inference and supports 40 languages and speaker recognition.
NVIDIA recently released the Nemotron 3.5 Transcribe speech transcription model.
The model can only process audio files and cannot be used directly for real-time audio streams.
After NVIDIA opens the model for free, developers of related tools can build multilingual transcription features without relying on paid APIs, potentially spurring the emergence of a batch of low-cost transcription tools.
NVIDIA released a free speech-to-text model that supports transcription of 40 languages.
The original material did not disclose details of the model's specific architecture, performance metrics or license terms.
Groq's core product, the LPU (Language Processing Unit), uses a deterministic dataflow architecture, differing from Nvidia GPUs' general-purpose design for training and inference scenarios.
NVIDIA believes the cause of agent incidents is not that some new capability is too powerful, but that tools, time, and ambiguous instructions together cause agents to bypass constraints to achieve goals after running for a long time, which NVIDIA calls "drift."
On September 28, NVIDIA open-sourced an agent security platform, OpenShell.
40 more timeline events
All Facts (20)
Open weights differ from fully open source: the former makes model parameters public but does not necessarily disclose training data and complete training code.
90%VerifiedThe NVIDIA A100's HBM2e capacity is 80GB, the H100 has 80GB HBM3, and the H200 has 141GB.
90%VerifiedThe HBM3 memory used by NVIDIA's H100 has a bandwidth of up to 3.35 TB/s.
90%VerifiedIn the H100 NVLink architecture, 8 GPUs within a single server can be fully interconnected via NVSwitch, achieving a total bandwidth of 900 GB/s, an order-of-magnitude improvement over PCIe 5.0's 64 GB/s.
90%VerifiedSince 2022, the United States has imposed advanced semiconductor export controls on China, restricting the export of high-end AI training chips such as the NVIDIA A100/H100 and their successor products.
90%VerifiedNVIDIA CUDA was launched in 2006 and was originally a programming framework for enabling developers to use GPUs for general-purpose computing.
90%VerifiedThe retail price of a single NVIDIA H100/H200 GPU is as high as $30,000–$40,000.
90%VerifiedNVIDIA's H100/H200 series GPUs hold more than 80% of the AI training and inference market.
90%VerifiedCUDA is the underlying support for virtually all mainstream deep learning frameworks.
85%VerifiedSince its introduction in 2006, CUDA has become the de facto industry standard in the AI/deep learning field, with mainstream frameworks such as PyTorch and TensorFlow using CUDA as their default backend.
80%VerifiedNeRF (Neural Radiance Fields) and its successor Gaussian Splatting can reconstruct three-dimensional scenes that can be freely navigated from any viewpoint from 2D images.
80%VerifiedTesla's Optimus, Figure AI's Figure 02, and Boston Dynamics' Atlas are the leading humanoid robot products at present.
80%VerifiedCerebras' latest-generation WSE-3 chip has an area of approximately 46,225 square millimeters, integrates 4 trillion transistors and 900,000 AI-optimized cores, and is more than 50 times the area of NVIDIA's H100 GPU.
80%VerifiedAs of 2023, Hugging Face had completed a Series D funding round with a valuation of $4.5 billion, with investors including Google, Amazon, and Nvidia.
80%VerifiedThe thermal design power (TDP) of a single NVIDIA H100 GPU card is as high as 700 watts.
80%VerifiedCUDA是NVIDIA于2006年推出的并行计算平台和编程模型
80%VerifiedDLSS 4 introduces multi-frame generation technology, capable of AI-generating up to 3 additional frames between each traditionally rendered frame, boosting frame rates by up to 4x.
80%VerifiedNvidia's CUDA ecosystem has formed a powerful moat.
75%VerifiedBoth the NVIDIA H100/H200 and the AMD MI300X integrate HBM.
75%VerifiedNVIDIA has long dominated the GPU computing market required for AI training and inference.
75%