{"id":18682,"date":"2026-09-16T20:10:06","date_gmt":"2026-09-16T13:10:06","guid":{"rendered":"https:\/\/droneth.or.th\/?post_type=product&#038;p=18682"},"modified":"2026-09-16T20:10:06","modified_gmt":"2026-09-16T13:10:06","slug":"nvidia-h100","status":"publish","type":"product","link":"https:\/\/droneth.or.th\/ja\/product\/nvidia-h100\/","title":{"rendered":"NVIDIA H100"},"content":{"rendered":"<p><strong>NVIDIA H100<\/strong> \u0e40\u0e1b\u0e47\u0e19 Data Center GPU \/ Accelerator \u0e08\u0e32\u0e01 <strong>NVIDIA<\/strong> \u0e40\u0e2b\u0e21\u0e32\u0e30\u0e2a\u0e33\u0e2b\u0e23\u0e31\u0e1a\u0e07\u0e32\u0e19 AI training, AI inference, HPC, virtual workstation \u0e23\u0e2d\u0e07\u0e23\u0e31\u0e1a\u0e01\u0e32\u0e23\u0e43\u0e0a\u0e49\u0e07\u0e32\u0e19\u0e23\u0e30\u0e14\u0e31\u0e1a\u0e2d\u0e07\u0e04\u0e4c\u0e01\u0e23 \u0e28\u0e39\u0e19\u0e22\u0e4c\u0e02\u0e49\u0e2d\u0e21\u0e39\u0e25 \u0e41\u0e25\u0e30\u0e07\u0e32\u0e19\u0e27\u0e34\u0e08\u0e31\u0e22\u0e14\u0e49\u0e32\u0e19 AI\/HPC<\/p>\n<p><strong>\u0e04\u0e33\u0e2d\u0e18\u0e34\u0e1a\u0e32\u0e22\u0e08\u0e32\u0e01\u0e1c\u0e39\u0e49\u0e1c\u0e25\u0e34\u0e15:<\/strong> A Massive Leap in Accelerated Compute.<\/p>\n<p><strong>\u0e01\u0e32\u0e23\u0e43\u0e0a\u0e49\u0e07\u0e32\u0e19\u0e17\u0e35\u0e48\u0e40\u0e2b\u0e21\u0e32\u0e30\u0e2a\u0e21<\/strong><\/p>\n<ul>\n<li>AI training<\/li>\n<li>AI inference<\/li>\n<li>HPC<\/li>\n<li>virtual workstation<\/li>\n<\/ul>\n<p><strong>\u0e02\u0e49\u0e2d\u0e21\u0e39\u0e25\u0e17\u0e32\u0e07\u0e40\u0e17\u0e04\u0e19\u0e34\u0e04 (Product Specifications)<\/strong><\/p>\n<table style=\"width:100%\">\n<thead>\n<tr>\n<td><strong>\u30bb\u30af\u30b7\u30e7\u30f3<\/strong><\/td>\n<td><strong>\u0e23\u0e32\u0e22\u0e25\u0e30\u0e40\u0e2d\u0e35\u0e22\u0e14\u0e2a\u0e40\u0e1b\u0e04<\/strong><\/td>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td><strong>\u0e2a\u0e16\u0e32\u0e1b\u0e31\u0e15\u0e22\u0e01\u0e23\u0e23\u0e21 GPU<\/strong><\/td>\n<td>NVIDIA Hopper architecture<\/td>\n<\/tr>\n<tr>\n<td><strong>Variant<\/strong><\/td>\n<td>H100 SXM \/ H100 NVL (PCIe)<\/td>\n<\/tr>\n<tr>\n<td><strong>FP64 (Precise)<\/strong><\/td>\n<td>SXM: 34 TFLOPS \/ NVL: 30 TFLOPS<\/td>\n<\/tr>\n<tr>\n<td><strong>FP64 Tensor Core<\/strong><\/td>\n<td>SXM: 67 TFLOPS \/ NVL: 60 TFLOPS<\/td>\n<\/tr>\n<tr>\n<td><strong>FP32<\/strong><\/td>\n<td>SXM: 67 TFLOPS \/ NVL: 60 TFLOPS<\/td>\n<\/tr>\n<tr>\n<td><strong>TF32 Tensor Core<\/strong><\/td>\n<td>SXM: 989 TFLOPS \/ NVL: 835 TFLOPS<\/td>\n<\/tr>\n<tr>\n<td><strong>BF16 Tensor Core<\/strong><\/td>\n<td>SXM: 1,979 TFLOPS \/ NVL: 1,671 TFLOPS<\/td>\n<\/tr>\n<tr>\n<td><strong>FP16 Tensor Core<\/strong><\/td>\n<td>SXM: 1,979 TFLOPS \/ NVL: 1,671 TFLOPS<\/td>\n<\/tr>\n<tr>\n<td><strong>FP8 Tensor Core<\/strong><\/td>\n<td>SXM: 3,958 TFLOPS \/ NVL: 3,341 TFLOPS<\/td>\n<\/tr>\n<tr>\n<td><strong>INT8 TOPS<\/strong><\/td>\n<td>SXM: 3,958 \/ NVL: 3,341<\/td>\n<\/tr>\n<tr>\n<td><strong>GPU Memory<\/strong><\/td>\n<td>SXM: 80 GB \/ NVL: 94 GB<\/td>\n<\/tr>\n<tr>\n<td><strong>\u30e1\u30e2\u30ea\u5e2f\u57df\u5e45<\/strong><\/td>\n<td>SXM: 3.35 TB\/s \/ NVL: 3.9 TB\/s<\/td>\n<\/tr>\n<tr>\n<td><strong>TDP (\u0e04\u0e27\u0e32\u0e21\u0e23\u0e49\u0e2d\u0e19\u0e2a\u0e39\u0e07\u0e2a\u0e38\u0e14)<\/strong><\/td>\n<td>SXM: up to 700 W \/ NVL: 350\u2013400 W<\/td>\n<\/tr>\n<tr>\n<td><strong>Multi\u2010Instance GPU (MIG)<\/strong><\/td>\n<td>Up to 7 MIGs (10 GB each SXM, 12 GB each NVL)<\/td>\n<\/tr>\n<tr>\n<td><strong>Form Factor<\/strong><\/td>\n<td>SXM \/ PCIe dual-slot air-cooled<\/td>\n<\/tr>\n<tr>\n<td><strong>Interconnect<\/strong><\/td>\n<td>NVLink: 900 GB\/s (SXM) or 600 GB\/s (NVL); PCIe Gen5: 128 GB\/s<\/td>\n<\/tr>\n<tr>\n<td><strong>Transformer Engine &amp; DPX<\/strong><\/td>\n<td>\u0e40\u0e23\u0e48\u0e07 LLM training\/inference \u0e41\u0e25\u0e30 dynamic programming \u0e44\u0e14\u0e49\u0e40\u0e23\u0e47\u0e27 30\u00d7, 7\u00d7<\/td>\n<\/tr>\n<tr>\n<td><strong>\u0e08\u0e4d\u0e32\u0e19\u0e27\u0e19 CUDA cores (PCIe)<\/strong><\/td>\n<td>14,592 CUDA cores<\/td>\n<\/tr>\n<tr>\n<td><strong>Transistors &amp; Process<\/strong><\/td>\n<td>80 billion transistors on TSMC 4N process<\/td>\n<\/tr>\n<tr>\n<td><strong>L2 Cache<\/strong><\/td>\n<td>50 MB<\/td>\n<\/tr>\n<tr>\n<td><strong>Asynchronous architecture<\/strong><\/td>\n<td>\u0e23\u0e2d\u0e07\u0e23\u0e31\u0e1a tensor memory accelerator, shared memory communications<\/td>\n<\/tr>\n<tr>\n<td><strong>Technical detail<\/strong><\/td>\n<td>NVIDIA Hopper GPU architecture<\/td>\n<\/tr>\n<tr>\n<td><strong>Technical detail<\/strong><\/td>\n<td>80GB Memory<\/td>\n<\/tr>\n<tr>\n<td><strong>Max. Power Consumption<\/strong><\/td>\n<td>300W~350W (configurable)<\/td>\n<\/tr>\n<tr>\n<td><strong>Interconnect Bus<\/strong><\/td>\n<td>PCIe Gen. 5: 128GB\/s NVLink: 600GB\/s<\/td>\n<\/tr>\n<tr>\n<td><strong>Multi-instance GPU (MIG)<\/strong><\/td>\n<td>7 GPU instances @ 10GB each<\/td>\n<\/tr>\n<tr>\n<td><strong>Technical detail<\/strong><\/td>\n<td>NVIDIA H100 \u0e04\u0e37\u0e2d GPU \u0e28\u0e39\u0e19\u0e22\u0e4c\u0e02\u0e49\u0e2d\u0e21\u0e39\u0e25\u0e17\u0e35\u0e48\u0e43\u0e0a\u0e49\u0e2a\u0e16\u0e32\u0e1b\u0e31\u0e15\u0e22\u0e01\u0e23\u0e23\u0e21 Hopper \u0e2d\u0e31\u0e19\u0e17\u0e31\u0e19\u0e2a\u0e21\u0e31\u0e22 \u0e16\u0e39\u0e01\u0e2d\u0e2d\u0e01\u0e41\u0e1a\u0e1a\u0e40\u0e1e\u0e37\u0e48\u0e2d\u0e23\u0e2d\u0e07\u0e23\u0e31\u0e1a\u0e07\u0e32\u0e19 AI, HPC, \u0e41\u0e25\u0e30\u0e01\u0e32\u0e23\u0e27\u0e34\u0e40\u0e04\u0e23\u0e32\u0e30\u0e2b\u0e4c\u0e02\u0e49\u0e2d\u0e21\u0e39\u0e25\u0e2d\u0e22\u0e48\u0e32\u0e07\u0e40\u0e15\u0e47\u0e21\u0e23\u0e39\u0e1b\u0e41\u0e1a\u0e1a \u0e42\u0e14\u0e22\u0e21\u0e35 Tensor Cores \u0e23\u0e38\u0e48\u0e19\u0e17\u0e35\u0e48 4, \u0e2b\u0e19\u0e48\u0e27\u0e22\u0e04\u0e27\u0e32\u0e21\u0e08\u0e4d\u0e32 HBM3 \u0e41\u0e25\u0e30\u0e01\u0e32\u0e23\u0e40\u0e0a\u0e37\u0e48\u0e2d\u0e21\u0e15\u0e48\u0e2d\u0e17\u0e35\u0e48\u0e40\u0e2b\u0e19\u0e37\u0e2d\u0e0a\u0e31\u0e49\u0e19 \u0e0a\u0e48\u0e27\u0e22\u0e40\u0e23\u0e48\u0e07\u0e1d\u0e36\u0e01\u0e41\u0e25\u0e30\u0e1b\u0e23\u0e30\u0e21\u0e27\u0e25\u0e1c\u0e25\u0e42\u0e21\u0e40\u0e14\u0e25\u0e02\u0e19\u0e32\u0e14\u0e43\u0e2b\u0e0d\u0e48\u0e41\u0e25\u0e30\u0e07\u0e32\u0e19\u0e1b\u0e23\u0e30\u0e21\u0e27\u0e25\u0e1c\u0e25\u0e40\u0e0a\u0e34\u0e07\u0e27\u0e34\u0e17\u0e22\u0e32\u0e28\u0e32\u0e2a\u0e15\u0e23\u0e4c\u0e44\u0e14\u0e49\u0e2d\u0e22\u0e48\u0e32\u0e07\u0e21\u0e35\u0e1b\u0e23\u0e30\u0e2a\u0e34\u0e17\u0e18\u0e34\u0e20\u0e32\u0e1e\u0e2a\u0e39\u0e07\u0e2a\u0e38\u0e14<\/td>\n<\/tr>\n<tr>\n<td><strong>Technical detail<\/strong><\/td>\n<td>NVIDIA H100 \u0e21\u0e35\u0e43\u0e2b\u0e49\u0e40\u0e25\u0e37\u0e2d\u0e01\u0e2a\u0e2d\u0e07\u0e40\u0e27\u0e2d\u0e23\u0e4c\u0e0a\u0e31\u0e19\u0e04\u0e37\u0e2d SXM (\u0e40\u0e1e\u0e34\u0e48\u0e21\u0e1b\u0e23\u0e30\u0e2a\u0e34\u0e17\u0e18\u0e34\u0e20\u0e32\u0e1e\u0e2a\u0e39\u0e07\u0e2a\u0e38\u0e14) \u0e41\u0e25\u0e30 PCIe (\u0e43\u0e0a\u0e49\u0e07\u0e32\u0e19\u0e17\u0e31\u0e48\u0e27\u0e44\u0e1b) \u0e1b\u0e23\u0e30\u0e2a\u0e34\u0e17\u0e18\u0e34\u0e20\u0e32\u0e1e\u0e2a\u0e39\u0e07\u0e2a\u0e38\u0e14\u0e16\u0e36\u0e07 ~4 \u0e1e\u0e31\u0e19 TFLOPS \u0e43\u0e19 FP8, \u0e23\u0e2d\u0e07\u0e23\u0e31\u0e1a MIG (Multi\u2010Instance GPU) \u0e41\u0e1a\u0e48\u0e07\u0e01\u0e32\u0e23\u0e4c\u0e14\u0e40\u0e1b\u0e47\u0e19\u0e2b\u0e25\u0e32\u0e22\u0e2b\u0e19\u0e48\u0e27\u0e22, \u0e43\u0e0a\u0e49 PCIe Gen5, NVLink \u0e04\u0e27\u0e32\u0e21\u0e40\u0e23\u0e47\u0e27\u0e2a\u0e39\u0e07 \u0e41\u0e25\u0e30\u0e21\u0e35\u0e2b\u0e19\u0e48\u0e27\u0e22\u0e04\u0e27\u0e32\u0e21\u0e08\u0e4d\u0e32 HBM3 \u0e02\u0e19\u0e32\u0e14 80 GB \u0e1e\u0e23\u0e49\u0e2d\u0e21\u0e41\u0e1a\u0e19\u0e14\u0e4c\u0e27\u0e34\u0e14\u0e17\u0e4c\u0e2a\u0e39\u0e07\u0e16\u0e36\u0e07 3.35 TB\/s (SXM) \u0e2b\u0e23\u0e37\u0e2d 3.9 TB\/s (NVL) \u0e40\u0e2b\u0e21\u0e32\u0e30\u0e01\u0e31\u0e1a\u0e23\u0e30\u0e1a\u0e1a\u0e23\u0e30\u0e14\u0e31\u0e1a\u0e2d\u0e07\u0e04\u0e4c\u0e01\u0e23\u0e17\u0e35\u0e48\u0e21\u0e35\u0e04\u0e27\u0e32\u0e21\u0e15\u0e49\u0e2d\u0e07\u0e01\u0e32\u0e23\u0e1b\u0e23\u0e30\u0e21\u0e27\u0e25\u0e1c\u0e25\u0e2a\u0e39\u0e07<\/td>\n<\/tr>\n<tr>\n<td><strong>Technical detail<\/strong><\/td>\n<td>The NVIDIA H100 is a data\u2010center GPU built on the advanced Hopper architecture, engineered for full-scale AI, HPC, and data analytics workloads. Featuring 4th\u2010generation Tensor Cores, HBM3 memory, and high-speed interconnects, it accelerates training and inference of large models and scientific compute tasks with exceptional performance.<\/td>\n<\/tr>\n<tr>\n<td><strong>The NVIDIA H100 is available in two variants<\/strong><\/td>\n<td>SXM for maximum performance and PCIe for broader compatibility. It offers up to ~4,000 TFLOPS (FP8) peak performance, supports MIG for instance partitioning, and features PCIe Gen5 and high-speed NVLink. With 80 GB of HBM3 memory delivering up to 3.35 TB\/s (SXM) or 3.9 TB\/s (NVL) bandwidth, it\u2019s ideal for enterprise-level compute-intensive tasks.<\/td>\n<\/tr>\n<tr>\n<td><strong>Technical detail<\/strong><\/td>\n<td>The NVIDIA H100 stands as a pinnacle of data-center GPUs, powered by the Hopper architecture. Its Transformer Engine and 4th-gen Tensor Cores dramatically accelerate AI model training and inference, especially for LLMs and scientific compute. With MIG for workload partitioning, vast HBM3 memory, and high-speed interconnects like NVLink, the H100 delivers an exponential leap in performance. It\u2019s the go-to choice for enterprises pushing towards exascale AI compute\u2014combining power, flexibility, and robust security.<\/td>\n<\/tr>\n<tr>\n<td><strong>Technical detail<\/strong><\/td>\n<td>Condition &amp; Warranty<\/td>\n<\/tr>\n<tr>\n<td><strong>Technical detail<\/strong><\/td>\n<td>\u2714 Fully tested and 100% working \u2714 Secure packaging for safe shipping \u2714 Warranty available (subject to store policy)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u0e02\u0e49\u0e2d\u0e21\u0e39\u0e25\u0e2d\u0e49\u0e32\u0e07\u0e2d\u0e34\u0e07\u0e08\u0e32\u0e01\u0e1c\u0e39\u0e49\u0e1c\u0e25\u0e34\u0e15: <a href=\"https:\/\/www.nvidia.com\/en-us\/data-center\/h100\/\" target=\"_blank\" rel=\"nofollow noopener\">NVIDIA<\/a><\/p>\n<p>\u0e2a\u0e19\u0e43\u0e08\u0e2a\u0e2d\u0e1a\u0e16\u0e32\u0e21\u0e23\u0e32\u0e04\u0e32 \u0e2a\u0e40\u0e1b\u0e04 \u0e2b\u0e23\u0e37\u0e2d\u0e02\u0e2d\u0e43\u0e1a\u0e40\u0e2a\u0e19\u0e2d\u0e23\u0e32\u0e04\u0e32 \u0e15\u0e34\u0e14\u0e15\u0e48\u0e2d\u0e44\u0e14\u0e49\u0e17\u0e35\u0e48 <strong>LINE: @metaxr<\/strong><\/p>","protected":false},"excerpt":{"rendered":"<p><strong>NVIDIA H100<\/strong> \u0e40\u0e1b\u0e47\u0e19 Data Center GPU \/ Accelerator \u0e08\u0e32\u0e01 <strong>NVIDIA<\/strong> \u0e40\u0e2b\u0e21\u0e32\u0e30\u0e2a\u0e33\u0e2b\u0e23\u0e31\u0e1a\u0e07\u0e32\u0e19 AI training, AI inference, HPC, virtual workstation \u0e23\u0e2d\u0e07\u0e23\u0e31\u0e1a\u0e01\u0e32\u0e23\u0e43\u0e0a\u0e49\u0e07\u0e32\u0e19\u0e23\u0e30\u0e14\u0e31\u0e1a\u0e2d\u0e07\u0e04\u0e4c\u0e01\u0e23 \u0e28\u0e39\u0e19\u0e22\u0e4c\u0e02\u0e49\u0e2d\u0e21\u0e39\u0e25 \u0e41\u0e25\u0e30\u0e07\u0e32\u0e19\u0e27\u0e34\u0e08\u0e31\u0e22\u0e14\u0e49\u0e32\u0e19 AI\/HPC<\/p>\n<div style=\"width: 100%; text-align: center; padding: 24px 0;\"><a style=\"display: inline-block; background-color: #06c755; color: #ffffff; font-size: 16px; font-weight: bold; text-decoration: none; padding: 14px 32px; border-radius: 50px; box-shadow: 0 4px 12px rgba(6,199,85,0.4); white-space: nowrap;\" href=\"https:\/\/page.line.me\/682mqqmh?openQrModal=true\" target=\"_blank\" rel=\"noopener\">\u0e2a\u0e2d\u0e1a\u0e16\u0e32\u0e21\u0e23\u0e32\u0e04\u0e32 \/ \u0e2a\u0e31\u0e48\u0e07\u0e0b\u0e37\u0e49\u0e2d LINE: @metaxr<\/a><\/div>","protected":false},"featured_media":18683,"comment_status":"closed","ping_status":"closed","template":"","meta":{"_acf_changed":false,"site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"default","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","ast-disable-related-posts":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}}},"product_brand":[],"product_cat":[388,384],"product_tag":[237,397,396],"class_list":{"0":"post-18682","1":"product","2":"type-product","3":"status-publish","4":"has-post-thumbnail","6":"product_cat-data-center-gpu","7":"product_cat-gpu-ai-server","8":"product_tag-ai-","9":"product_tag-gpu","10":"product_tag-nvidia","11":"desktop-align-left","12":"tablet-align-left","13":"mobile-align-left","15":"first","16":"instock","17":"shipping-taxable","18":"product-type-simple"},"acf":[],"_links":{"self":[{"href":"https:\/\/droneth.or.th\/ja\/wp-json\/wp\/v2\/product\/18682","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/droneth.or.th\/ja\/wp-json\/wp\/v2\/product"}],"about":[{"href":"https:\/\/droneth.or.th\/ja\/wp-json\/wp\/v2\/types\/product"}],"replies":[{"embeddable":true,"href":"https:\/\/droneth.or.th\/ja\/wp-json\/wp\/v2\/comments?post=18682"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/droneth.or.th\/ja\/wp-json\/wp\/v2\/media\/18683"}],"wp:attachment":[{"href":"https:\/\/droneth.or.th\/ja\/wp-json\/wp\/v2\/media?parent=18682"}],"wp:term":[{"taxonomy":"product_brand","embeddable":true,"href":"https:\/\/droneth.or.th\/ja\/wp-json\/wp\/v2\/product_brand?post=18682"},{"taxonomy":"product_cat","embeddable":true,"href":"https:\/\/droneth.or.th\/ja\/wp-json\/wp\/v2\/product_cat?post=18682"},{"taxonomy":"product_tag","embeddable":true,"href":"https:\/\/droneth.or.th\/ja\/wp-json\/wp\/v2\/product_tag?post=18682"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}