Blackwell
Blackwell is a GPU microarchitecture introduced by NVIDIA, targeting high-performance computing, artificial intelligence training, and inference applications. The architecture improves computational throughput and energy efficiency relative to its predecessor, supports large-scale parallel computing tasks, and incorporates optimizations specific to AI workloads such as large language models. The Blackwell architecture is deployed across multiple NVIDIA GPU product lines serving data center and professional computing domains.
Core Facts
NVFP4 is a 4-bit floating-point (FP4) quantization format introduced by NVIDIA, designed specifically for the hardware acceleration capabilities of Blackwell architecture GPUs
NVIDIA's H100/H200 series GPUs hold more than 80% of the AI training and inference market.
The RTX 5090 has 32 GB of VRAM.
Timeline (last 90 days)
The B200 belongs to the Blackwell architecture and uses a dual-die package.
RTX Spark is based on Blackwell-architecture RTX capabilities and targets thin-and-light laptops and compact desktop form factors.
The RTX 6000 Pro is a workstation-class GPU NVIDIA launched based on the Blackwell architecture (a member of the RTX PRO 6000 Blackwell series), equipped with 96GB of GDDR7 memory and aimed at professional rendering, scientific computing and local AI inference.
B200 is an NVIDIA GPU product for large-scale AI workloads; it is a data center GPU of the Blackwell architecture, uses a dual-die package, and is equipped with HBM3e high-bandwidth memory.
SM100 corresponds to GPUs with compute capability 10.0 in NVIDIA's Blackwell architecture, such as B200 and GB200; it is an architecture identifier rather than a specific graphics card model.
DeepInfra's kernels are written for NVIDIA data center Blackwell chips (such as the B200) and cannot be used out of the box on devices using different Blackwell chips, such as the DGX Spark.
DeepInfer is released under the Apache 2.0 license, supports NVIDIA Blackwell and Huawei Ascend 950, and includes components such as expert load balancing, speculative decoding, and asynchronous scheduling.
Compute Capability is a version number NVIDIA defines for its GPU architectures, expressed as "major.minor" (e.g., 8.x for Ampere, 8.9 for Ada Lovelace, 12.0 for Blackwell), encoded directly in the GPU hardware identity and read by the driver via PCI configuration space and GPU registers; it does not depend on upper-layer software or platform labels and cannot be forged by editing product descriptions or the virtualization layer (MIG partitioning only affects the visibility of memory and compute units, not the architecture version number).
NVFP4 is a Blackwell tensor core format, and the FP4-Local model card explicitly states that Blackwell is the only supported microarchitecture, with Hopper and Ampere not supported.
Running this engine requires a Blackwell GPU (the kernels use the NVFP4 hardware path), about 70GB of space on fast NVMe, and 32GB of RAM.
40 more timeline events
All Facts (20)
NVFP4 is a 4-bit floating-point (FP4) quantization format introduced by NVIDIA, designed specifically for the hardware acceleration capabilities of Blackwell architecture GPUs
90%VerifiedNVIDIA's H100/H200 series GPUs hold more than 80% of the AI training and inference market.
90%VerifiedThe RTX 5090 has 32 GB of VRAM.
80%VerifiedCore innovations of the Blackwell architecture include the second-generation Transformer Engine (supporting FP4 precision), the NVLink Switch system (supporting interconnection of 576 GPUs), 20 petaFLOPS of FP4 compute, and 192 GB HBM3e memory (with bandwidth reaching 8 TB/s).
80%VerifiedBlackwell's second-generation Transformer Engine adds FP4 precision support, which in theory can improve Transformer model training throughput by 2–4x compared to the H100.
75%VerifiedThe GB10 Grace Blackwell Superchip integrates an ARM-architecture Grace CPU and a Blackwell GPU within the same package via NVLink-C2C chip-to-chip interconnect technology.
75%VerifiedNVIDIA designed a dedicated Tensor Core instruction mma.kind::mxf4 for NVFP4 in the Blackwell architecture (SM 100/103, such as the B100 and B200 series), and adopted an FP8 scaling factor mechanism with one scale per group of 16 elements.
75%VerifiedNVIDIA Blackwell 架构于2024年发布，是继 Hopper（H100）之后的新一代数据中心 GPU 架构
75%VerifiedThe NVIDIA Blackwell architecture (GB200/B200 series) introduces native FP4 tensor cores at the hardware level
75%VerifiedThe 'G' in GB200 stands for Grace CPU, and the 'B' stands for Blackwell GPU.
75%VerifiedBlackwell (B100/B200 series) was released in 2024, manufactured using TSMC's 4NP process, with a single GPU integrating 208 billion transistors
75%VerifiedThe NVIDIA RTX 5090 is based on the Blackwell architecture, is equipped with 32 GB of GDDR7 memory, and has a theoretical FP16 computing power exceeding 100 TFLOPS.
70%VerifiedNVIDIA's upcoming RTX Spark laptop offers up to 128 GB of unified memory in its highest configuration.
65%VerifiedThe RTX 5070 belongs to NVIDIA's Blackwell architecture consumer graphics card product line and features approximately 12 GB of VRAM.
65%VerifiedThe RTX 5090 is based on NVIDIA's Blackwell architecture.
65%VerifiedThe RTX 5090 is based on NVIDIA's Blackwell architecture.
65%VerifiedThe GB300 is NVIDIA's flagship data center chip in the Blackwell Ultra series, integrating HBM3e memory with a single-card capacity of 288 GB.
65%VerifiedThe Blackwell architecture is named after American mathematician and statistician David Blackwell.
65%Unverified2024年GTC上发布了Blackwell架构GPU和NIM（NVIDIA Inference Microservices）等产品
90%Unverified每颗Blackwell GPU拥有高达192GB HBM3e显存
90%