[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-672775-105":3,"detail-sidebar-cat-0-en-105":80,"doc-detail-672775-en":130},{"code":4,"msg":5,"data":6},0,"ok",{"site_id":7,"language":8,"slug":9,"title":10,"keywords":11,"description":12,"schema_data":13,"social_meta":73,"head_meta":75,"extra_data":77,"updated_unix":79},105,"en","nvidia-a100-tensor-core-gpu-datasheet-jun21","NVIDIA A100 Tensor Core GPU - Datasheet - Jun21","","NVIDIA A100 Tensor Core GPU datasheet covering architecture-driven performance for AI training, AI inference, and high-performance computing. Highlights include up to 20X performance over NVIDIA Volta, elastic scaling via Multi-Instance GPU (MIG) up to seven isolated GPU instances, and support for multiple math precisions using a single accelerator. Specifies configurations for SXM4 and PCIe form factors, including 40GB and 80GB variants, memory bandwidth up to 2 TB/s, thermal design power ranges, and detailed FP64/FP32/TF32/BF16/FP16/INT8 Tensor Core throughput figures.",{"@graph":14,"@context":72},[15,34,55],{"@type":16,"itemListElement":17},"BreadcrumbList",[18,23,27,31],{"item":19,"name":20,"@type":21,"position":22},"https://docshare.wps.com","Home","ListItem",1,{"item":24,"name":25,"@type":21,"position":26},"https://docshare.wps.com/document/","Document",2,{"item":28,"name":29,"@type":21,"position":30},"https://docshare.wps.com/document/technology/","Technology",3,{"item":32,"name":10,"@type":21,"position":33},"https://docshare.wps.com/document/nvidia-a100-tensor-core-gpu-datasheet-jun21/672775/",4,{"url":32,"name":10,"@type":35,"image":36,"author":41,"headline":10,"publisher":44,"fileFormat":47,"inLanguage":8,"description":12,"dateModified":48,"datePublished":49,"encodingFormat":47,"isAccessibleForFree":50,"interactionStatistic":51},"DigitalDocument",{"url":37,"@type":38,"width":39,"height":40},"https://docshare.wps.com/thumbnails/nvidia-a100-tensor-core-gpu-datasheet-jun21/672775.png","ImageObject",300,407,{"name":42,"@type":43},"AditGamingID","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-11-04","2026-11-02",true,{"@type":52,"interactionType":53,"userInteractionCount":30},"InteractionCounter",{"@type":54},"ViewAction",{"@type":56,"mainEntity":57},"FAQPage",[58,64,68],{"name":59,"@type":60,"acceptedAnswer":61},"What is the main performance positioning of NVIDIA A100 for data centers?","Question",{"text":62,"@type":63},"NVIDIA A100 is positioned as an accelerator platform delivering unprecedented acceleration at every scale, powering elastic data centers for AI, data analytics, and HPC. It is also presented as up to 20X higher performance than the prior NVIDIA Volta generation.","Answer",{"name":65,"@type":60,"acceptedAnswer":66},"How does Multi-Instance GPU (MIG) work on A100?",{"text":67,"@type":63},"A100 can be efficiently partitioned into up to seven isolated GPU instances using Multi-Instance GPU (MIG). This enables dynamic adjustment to changing workload demands in an elastic data center.",{"name":69,"@type":60,"acceptedAnswer":70},"What memory and bandwidth capabilities are highlighted for the A100 80GB model?",{"text":71,"@type":63},"The datasheet states that the latest A100 80GB doubles GPU memory and debuts the world’s fastest memory bandwidth at 2 TB/s. This is intended to speed time to solution for large models and massive datasets.","https://schema.org",{"og:url":32,"og:type":74,"og:title":10,"og:site_name":45,"og:description":12},"article",{"robots":76,"canonical":32},"index,follow",{"doc_id":78,"site_id":7},672775,1793607139,{"code":4,"msg":81,"data":82},"success",[83,87,91,95,100,104,109,114,119,122,126],{"id":22,"doc_module":4,"doc_module_name":25,"category_name":84,"show_sort_weight":85,"slug":86},"Story & Novel",90,"story-novel",{"id":26,"doc_module":4,"doc_module_name":25,"category_name":88,"show_sort_weight":89,"slug":90},"Literature",80,"literature",{"id":33,"doc_module":4,"doc_module_name":25,"category_name":92,"show_sort_weight":93,"slug":94},"Exam",70,"exam",{"id":96,"doc_module":4,"doc_module_name":25,"category_name":97,"show_sort_weight":98,"slug":99},5,"Comic",60,"comic",{"id":101,"doc_module":4,"doc_module_name":25,"category_name":29,"show_sort_weight":102,"slug":103},6,50,"technology",{"id":105,"doc_module":4,"doc_module_name":25,"category_name":106,"show_sort_weight":107,"slug":108},7,"Healthcare",40,"healthcare",{"id":110,"doc_module":4,"doc_module_name":25,"category_name":111,"show_sort_weight":112,"slug":113},8,"Research & Report",30,"research-report",{"id":115,"doc_module":4,"doc_module_name":25,"category_name":116,"show_sort_weight":117,"slug":118},9,"Religion & Spirituality",20,"religion-spirituality",{"id":117,"doc_module":4,"doc_module_name":25,"category_name":120,"show_sort_weight":117,"slug":121},"World Cup","world-cup",{"id":123,"doc_module":4,"doc_module_name":25,"category_name":124,"show_sort_weight":123,"slug":125},10,"Lifestyle","lifestyle",{"id":127,"doc_module":4,"doc_module_name":25,"category_name":128,"show_sort_weight":96,"slug":129},19,"General","general",{"code":4,"msg":81,"data":131},{"doc_id":78,"user_id":132,"nickname":42,"user_avatar":133,"doc_module":4,"category_id":101,"category_name":29,"doc_title":10,"doc_description":12,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":30,"is_deleted":4,"is_public":22,"is_downloadable":22,"audit_status":22,"page_count":30,"language":139,"language_code":8,"site_id":7,"html_lang":8,"table_of_contents":140,"faqs":141,"seo_title":142,"seo_description":12,"update_tm":79,"read_time":110},962094249103,"https://ap-avatar.wpscdn.com/davatar_276721f389ce27ea32af1340a28f341c","NVIDIA A100 TENSOR CORE GPU  \nUnprecedented Acceleration at Every Scale  \nThe Most Powerful Compute Platform for Every Workload  \nThe NVIDIA A100 Tensor Core GPU delivers unprecedented acceleration—at every scale—to power the world’s highestperforming elastic data centers for AI, data analytics, and highperformance computing (HPC) applications. As the engine of the NVIDIA data center platform, A100 provides up to 20X higher performance over the prior NVIDIA Volta™ generation. A100 can efficiently scale up or be partitioned into seven isolated GPU instances with Multi-Instance GPU (MIG), providing a unified platform that enables elastic data centers to dynamically adjust to shifting workload demands.  \nNVIDIA A100 Tensor Core technology supports a broad range of math precisions, providing a single accelerator for every workload. The latest generation A100 80GB doubles GPU memory and debuts the world’s fastest memory bandwidth at 2 terabytes per second (TB/s), speeding time to solution for the largest models and most massive datasets.  \nA100 is part of the complete NVIDIA data center solution that incorporates building blocks across hardware, networking, software, libraries, and optimized AI models and applications from the NVIDIA NGC™ catalog. Representing the most powerful end-to-end AI and HPC platform for data centers, it allows researchers to deliver real-world results and deploy solutions into production at scale.  \nNVIDIA A100 TENSOR CORE GPU SPECIFICATIONS (SXM4 AND PCIE FORM FACTORS)  \n\n| A100 | A100 | A100 | A100 |\n| --- | --- | --- | --- |\n| 40GB PCIe | 80GB PCIe | 40GB SXM | 80GB SXM |\n\nFP64  \n9.7 TFLOPS  \n\n| FP64 Tensor Core | 19.5 TFLOPS |\n| --- | --- |\n| FP32 | 19.5 TFLOPS |\n| Tensor Float 32 (TF32) | 156 TFLOPS | 312 TFLOPS* |\n| BFLOAT16 Tensor Core | 312 TFLOPS | 624 TFLOPS* |\n| FP16 Tensor Core | 312 TFLOPS | 624 TFLOPS* |\n\nINT8 Tensor Core  \n624 TOPS | 1248 TOPS*  \n\n| GPU Memory 40GB\u003Cbr>HBM2 |  | 80GB\u003Cbr>HBM2e | 40GB\u003Cbr>HBM2 | 80GB\u003Cbr>HBM2e |\n| --- | --- | --- | --- | --- |\n| GPU Memory 1,555GB/s Bandwidth |  | 1,935GB/s | 1,555GB/s | 2,039GB/s |\n| Max Thermal\u003Cbr>Design Power (TDP) | 250W | 300W | 400W | 400W |\n| Multi-Instance Up to 7 |  | Up to 7 | Up to 7 | Up to 7 |\n| GPU MIGs @ |  | MIGs @ | MIGs @ | MIGs @ |\n| 5GB |  | 10GB | 5GB | 10GB |\n| Form Factor PCIe |  |  | SXM |  |\n| Interconnect NVIDIA® NVLink® Bridge for 2 GPUs: 600GB/s **\u003Cbr>PCIe Gen4: 64GB/s |  |  | NVLink: 600GB/s\u003Cbr>PCIe Gen4: 64GB/s |  |\n| Server Options Partner and NVIDIACertified Systems™ with\u003Cbr>1-8 GPUs |  |  | NVIDIA HGX™ A100-\u003Cbr>Partner and NVIDIACertified Systems with\u003Cbr>4,8, or 16 GPUs NVIDIA DGX™ A100 with\u003Cbr>8 GPUs |  |\n\n* With sparsity  \n** SXM4 GPUs via HGX A100 server boards; PCIe GPUs via NVLink Bridge for up to two GPUs  \nNVIDIA A100 TENSOR CORE GPU | DATA SHEET | JUN21 | 1  \nIncredible Performance Across Workloads  \n\n| Up to 3X Higher AI Training on Largest Models\u003Cbr>DLRM Training |  |\n| --- | --- |\n| 3X\u003Cbr>2X\u003Cbr>1X\u003Cbr>0 | |\n| DLRM on HugeCTR framework, precision = FP16 | NVIDIA A100 80GB batch size = 48 | NVIDIA A100 40GB batch size = 32 | NVIDIA V100 32GB batch size = 32. |  |\n\n\n| Up to 249X Higher AI Inference Performance over CPUs\u003Cbr>BERT-LARGE Inference |  |\n| --- | --- |\n| 250X\u003Cbr>200X\u003Cbr>150X\u003Cbr>100X\u003Cbr>50X\u003Cbr>0 | 245X 249X 1X\u003Cbr>\u003Cbr>|\n| CPU Only A100 40GB A100 80GB\u003Cbr>Sequences Per Second-Relative Performance\u003Cbr>BERT-Large Inference | CPU only: Dual Xeon Gold 6240 @2.60 GHz, precision = FP32, batch size = 128 | V100: NVIDIA TensorRT™ (TRT) 7.2, precision = INT8, batch size = 256 | A100 40GBand 80GB, batch size = 256, precision = INT8 with sparsity. |  |\n\n\n| 11X More HPC Performance in Four Years Throughput for Top HPC Apps |  |  |  |  |  |\n| --- | --- | --- | --- | --- | --- |\n| 11X\u003Cbr>10X\u003Cbr>9X\u003Cbr>8X\u003Cbr>7X\u003Cbr>6X\u003Cbr>5X\u003Cbr>4X\u003Cbr>3X\u003Cbr>2X\u003Cbr>1X\u003Cbr>0 | 11X 4X 3X 2X \u003Cbr>1X\u003Cbr>\u003Cbr>\u003Cbr>\u003Cbr>|  |  |  |  |\n|  | P100\u003Cbr>2016 | V100\u003Cbr>2017 | V100\u003Cbr>2018 | V100\u003Cbr>2019 | A100\u003Cbr>2020 |\n| Throughput-Relative Performance\u003Cbr>G","cbCaifs9dBC2LnVD","https://ap.wps.com/l/cbCaifs9dBC2LnVD","pdf",627682,"English","# NVIDIA A100 Tensor Core GPU\n## Key capabilities and performance highlights\n## Specifications by form factor (SXM4 and PCIe)\n## Tensor Core performance (FP64/FP32/TF32/BF16/FP16/INT8)\n## GPU memory, bandwidth, power, and MIG options\n## Performance across workloads\n## Groundbreaking innovations (Ampere architecture)","[{\"question\":\"What is the main performance positioning of NVIDIA A100 for data centers?\",\"answer\":\"NVIDIA A100 is positioned as an accelerator platform delivering unprecedented acceleration at every scale, powering elastic data centers for AI, data analytics, and HPC. It is also presented as up to 20X higher performance than the prior NVIDIA Volta generation.\"},{\"question\":\"How does Multi-Instance GPU (MIG) work on A100?\",\"answer\":\"A100 can be efficiently partitioned into up to seven isolated GPU instances using Multi-Instance GPU (MIG). This enables dynamic adjustment to changing workload demands in an elastic data center.\"},{\"question\":\"What memory and bandwidth capabilities are highlighted for the A100 80GB model?\",\"answer\":\"The datasheet states that the latest A100 80GB doubles GPU memory and debuts the world’s fastest memory bandwidth at 2 TB/s. This is intended to speed time to solution for large models and massive datasets.\"}]","NVIDIA A100 Tensor Core GPU - Datasheet - Jun21 | PDF"]