# NVIDIA A800 40GB Active Graphics Card The ultimate workstation development platform for AI, data science, and HPC. ## Where to Buy Find an NVIDIA design and visualization partner. [Find a Partner](https://www.nvidia.com/en-us/products/workstations/where-to-buy.md) [Datasheet](https://resources.nvidia.com/en-us-briefcase-for-datasheets/proviz-a800-40gb-dat?ncid=no-ncid) | [Specs](#specs) ## The Supercomputing Platform for Workstations The NVIDIA® A800 40GB Active GPU, powered by the NVIDIA Ampere architecture, is the ultimate workstation development platform, delivering powerful performance to accelerate next-generation data science, AI, HPC, and engineering simulation/CAE workloads. It comes with a five-year NVIDIA Enterprise subscription, which includes [NVIDIA AI Enterprise](https://www.nvidia.com/en-us/data-center/products/ai-enterprise.md) to help users tackle demanding AI, data science, and analytics workflows on workstations. ### Get Started With NVIDIA A800 40GB Active Learn how to set up NVIDIA A800 40GB Active with a companion GPU for display and activate the NVIDIA Enterprise license. [Watch Video](#watch-video) ## Highlights ## Industry Leading Performance ### Double-Precision (FP64) Performance 9.7 TFLOPS¹ ### Tensor Performance 1,247 AI TOPS² ### Memory Bandwidth 1.5 TB/s 1 Peak rates based on GPU Boost Clock. 2 Theoretical INT8 TOPS using sparsity. ## Features ## Powered by the NVIDIA Ampere Architecture ### Third-Generation Tensor Cores Performance and versatility for a wide range of AI and HPC applications with support for double-precision (FP64) and Tensor Float 32 (TF32) precision provides up to 2X the performance and efficiency over the previous generation, enabling rapid model training and inferencing directly on [RTX-powered AI workstations](https://www.nvidia.com/en-us/ai-data-science/workstations.md). Hardware support for structural sparsity doubles the throughput for inferencing. ### Multi-Instance GPU Fully isolated and secure multi-tenancy at the hardware level with dedicated high-bandwidth memory, cache, and compute cores. [Multi-Instance GPU](https://www.nvidia.com/en-us/technologies/multi-instance-gpu.md) (MIG) maximizes the utilization of GPU-accelerated infrastructure, allowing an A800 40GB Active GPU to be partitioned into as many as seven independent instances, giving multiple users access to GPU acceleration. ### Third-Generation NVIDIA NVLink Increased GPU-to-GPU interconnect bandwidth provides a single scalable memory to accelerate compute workloads and tackle larger datasets. Connect a pair of NVIDIA A800 40GB Active GPUs with [NVIDIA NVLink®](https://www.nvidia.com/en-us/products/workstations/nvlink-bridges.md) to increase the effective memory footprint to 80GB and scale application performance by enabling GPU-to-GPU data transfers at rates up to 400 GB/s (bidirectional). ### Ultra-Fast HBM2 Memory Deliver massive computational throughput with 40GB of high-speed HBM2 memory with a class-leading 1.5 TB/s of memory bandwidth—an over 70% increase compared to previous generation—and significantly more on-chip memory, including a 40MB level 2 cache to accelerate the most computationally intense AI and HPC workloads. [Learn more about the NVIDIA Ampere GPU architecture](https://www.nvidia.com/en-us/data-center/ampere-architecture.md) ## Workloads ## Supercharge AI and HPC Workflows Across Industries #### Generative AI Using neural networks to identify patterns and structures within existing data, generative AI applications enable users to generate new and original content from a wide variety of inputs and outputs, including images, sounds, animation, and 3D models. Leverage NVIDIA’s generative AI solution—NeMo™ Framework, included in NVIDIA AI Enterprise—along with the A800 40GB Active GPU for easy, fast, and customizable generative AI model development. [Read the AI Workstation Solution Brief](https://resources.nvidia.com/en-us-design-viz-stories-ep/proviz-rtx-ai-powered-workstations-solution-overview?lx=CCKW39) #### Engineering Simulation/CAE The A800 40GB Active GPU delivers remarkable performance for GPU-accelerated computer-aided engineering (CAE) applications. Engineering Analysts and CAE Specialists can run large-scale simulations and engineering analysis codes in full FP64 precision with incredible speed, shortening development timelines and accelerating time to value. With the addition of RTX-accelerated GPUs, providing display capabilities for pre- and post-processing, designers and engineers can visualize large-scale simulations and models in full-design fidelity. Image credit: Dassault Systèmes #### Data Science and Data Analytics Accelerate end-to-end data science and analytics workflows with powerful performance to extract meaningful insights from large-scale datasets quickly. By combining the high-performance computing capabilities of the A800 40GB Active with NVIDIA AI Enterprise, data practitioners can leverage a large collection of libraries, tools, and technologies to accelerate data science workflows—from data prep and analysis to modeling. #### AI Training and Inference Offload data center and cloud-based computing resources and bring supercomputing performance to the desktop for local AI training and inference workloads. Powerful workstations with four A800 40GB Active GPUs provide over 2.5 petaflops of AI computing performance and 160GB of combined HBM2 memory. [Read the AI Training Solution Brief](https://resources.nvidia.com/en-us-design-viz-stories-ep/proviz-ai-workstation-training-solution?lx=CCKW39&search=training) [Read the AI Inference Solution Brief](https://resources.nvidia.com/en-us-design-viz-stories-ep/proviz-ai-workstation-inference-solution?lx=CCKW39&search=inference) ## AI Platform ## Production-Ready AI With NVIDIA AI Enterprise ### Out-of-the-Box AI Development [NVIDIA AI Enterprise](https://www.nvidia.com/en-us/data-center/products/ai-enterprise.md) is an end-to-end enterprise software platform for rapid development and deployment of production-ready generative AI, computer vision, speech AI, and more. ### Accelerated Data Pipelines NVIDIA AI Enterprise includes data science libraries and tools to speed time to insights. Organizations can use [NVIDIA RAPIDS™](https://www.nvidia.com/en-us/deep-learning-ai/software/rapids.md) for up to 50X faster end-to-end data science pipelines. ### AI Training and Inference NVIDIA AI Enterprise accelerates every stage of the AI journey, from data prep and model training through inference and deployment at scale: * Access pretrained models with [NVIDIA TAO](https://developer.nvidia.com/tao-toolkit) * Optimize for inference with [NVIDIA® TensorRT](https://developer.nvidia.com/tensorrt) * Deploy at scale with [NVIDIA Triton™ Inference Server](https://www.nvidia.com/en-us/ai/dynamo-triton.md) [Learn more about NVIDIA AI Enterprise](https://www.nvidia.com/en-us/data-center/products/ai-enterprise.md) ## Performance ## Tackle Demanding AI and HPC Workloads The NVIDIA A800 40GB Active GPU delivers incredible performance to conquer the most demanding workflows on workstation platforms—from AI training and inference, to complex engineering simulations, modeling, and data analysis. With more than 2X the performance of the previous generation, the A800 40GB Active supports a wide range of compute-intensive workloads flawlessly. ### AI Training - ResNet-50 V1.5 ResNet-50 V1.5 Training. Batch Size=256; Precision=Mixed. ### AI Training - BERT - Large BERT Large Pre-Training Phase 2 Batch Size=8; Precision=Mixed. ### HPC - GTC GTC Version 4.5, TAE, Precision=FP32. ### HPC - LAMMPS LAMMPS patch\_8Feb2023, Atomic Fluid Lennard-Jones 2.5 (cutoff); Precision=FP64. ### AI Inference - ResNet-50 V1.5 ResNet-50 V1.5 Inference. Batch Size=128; Precision=Mixed. ### AI Inference - BERT - Large BERT Large Inference. Batch Size=128; Precision=INT8. Performance testing with A800 40GB Active and Quadro GV100 GPUs and Intel Xeon Gold 6126 processor. ## Supercomputing Performance on Desktop Workstations Offload demand for data center resources with NVIDIA RTX™-powered AI workstations that deliver the power of a supercomputer to the desktop. Workstation platforms equipped with the latest NVIDIA RTX GPUs and NVIDIA AI Enterprise software provide powerful AI performance to build, train, and deploy the next generation of AI-augmented applications and models. [Learn More About NVIDIA RTX-Powered AI Workstations](https://www.nvidia.com/en-us/ai-data-science/workstations.md) ## Specifications ## NVIDIA A800 40GB Active | | | | --- | --- | | GPU Memory | 40GB HBM2 | | Memory Interface | 5,120-bit | | Memory Bandwidth | 1,555.2 GB/s | | CUDA Cores | 6,912 | | Tensor Cores | 432 | | Double-Precision Performance | 9.7 TFLOPS | | Single-Precision Performance | 19.5 TFLOPS | | Peak Tensor Performance | 1,247 AI TOPS | 623.8 TFLOPS | | Multi-Instance GPU | Up to 7 MIG instances @ 5GB | | NVIDIA NVLink | Yes | | NVLink Bandwidth | 400GB/s | | Graphics Bus | PCIe 4.0 x 16 | | Max Power Consumption | 240W | | Thermal | Active | | Form Factor | 4.4” H x 10.5” L, dual slot | | Display Capability\* | – | | \*The A800 40GB Active does not come equipped with display ports. Either the NVIDIA RTX 4000 Ada Generation, NVIDIA RTX A4000, NVIDIA RTX A1000, or the NVIDIA T1000 GPU is required to support display out capabilities. | | [View NVIDIA A800 40GB Active Datasheet](https://resources.nvidia.com/en-us-briefcase-for-datasheets/proviz-a800-40gb-dat?ncid=no-ncid) ## Get Started ### Ready to Purchase? Talk with an NVIDIA design and visualization partner. [Find a Partner](https://www.nvidia.com/en-us/products/workstations/where-to-buy.md) ### Need Help Selecting the Right Product or Partner? Talk to an NVIDIA product specialist about your professional needs. [Contact Us](https://www.nvidia.com/en-us/contact/sales.md) ### Get the Latest on NVIDIA RTX Sign up for the latest news, updates, and more from NVIDIA. [Subscribe](https://www.nvidia.com/en-us/preferences/email-signup.md) ## Sign Up To Be Notified On Availability Welcome back. Not you? Log Out Welcome back. Not you? Clear form ## NVIDIA A800 40GB Active Quick Specs | | | | --- | --- | | FP64 | 9.7 teraFLOPS | | FP32 | 19.5 teraFLOPS | | Memory Bandwidth | 1,555 GB/s | | GPU Memory | 40GB HBM2 | | Max Power Consumption | 240 W | [View NVIDIA A800 40GB Active Datasheet](https://resources.nvidia.com/en-us-briefcase-for-datasheets/proviz-a800-40gb-dat?ncid=no-ncid) ![YouTube Video](https://img.youtube.com/vi_webp/YBPnYNci1co/maxresdefault.webp) Consent for Optional Cookies (googleCookiePolicyLink)YouTube sets performance, advertising, and other optional cookies(/googleCookiePolicyLink) when you watch embedded videos. To watch this video, you need to turn on optional cookies for the site. By clicking “Accept and Play Video,” you will automatically turn on advertising and other optional cookies for the site and accept our (nvidiaTermsOfServiceLink)Terms of Service(/nvidiaTermsOfServiceLink) (which contains important waivers). Please see our (nvidiaPrivacyPolicyLink)Privacy Policy(/nvidiaPrivacyPolicyLink) and (nvidiaCookiePolicyLink)Cookie Policy(/nvidiaCookiePolicyLink) for more information. Cancel Accept and Play Video Alternatively, you can (youtubeLink)watch this video on YouTube(/youtubeLink).