GB200 NVL72: Next-Gen AI scientific discovery tool

At rack scale, GB200 NVL72 connects 36 Grace CPUs and 72 Blackwell GPUs

[{"selector":"#anim-e172a5a2-f623-4c8b-97c3-5d7e4f60d20e [data-leaf-element=\"true\"]","keyframes":{"transform":["translate(-7.105427357601002e-15%, 0%) scale(1.5)","translate(0%, 0%) scale(1)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"forwards"}] [{"selector":"#anim-82eb54f8-3733-44c8-b76f-14871f749427","keyframes":{"transform":["rotate(-540deg) scale(0.1)","none"],"opacity":[0,1]},"delay":0,"duration":1000,"fill":"both","iterations":1}]

The NVIDIA GB200 Grace Blackwell Superchip uses NVLink-C2C to connect two powerful NVIDIA Blackwell Tensor Core GPUs and a Grace CPU

[{"selector":"#anim-f0bafd4f-efc9-4a32-bfcb-bb104c924267 [data-leaf-element=\"true\"]","keyframes":{"transform":["translate(0%, 0%) scale(1.5)","translate(0%, 0%) scale(1)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"forwards"}] [{"selector":"#anim-37c44ace-8bb5-4240-a7fb-69006baff460","keyframes":{"transform":["rotate(-540deg) scale(0.1)","none"],"opacity":[0,1]},"delay":0,"duration":1000,"fill":"both","iterations":1}]

This progress is enabled by a new generation of Tensor Cores with high precision, throughput, and microscaling formats

[{"selector":"#anim-d0d4bb04-bea4-4296-aacd-aa14581cabc2","keyframes":{"transform":["translate3d(140.54053%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":1000,"easing":"cubic-bezier(.2, 0, .8, 1)","fill":"both"}] [{"selector":"#anim-f40024d7-7143-4694-a1f7-27107fb922ed","keyframes":{"transform":["rotateZ(180deg)","rotateZ(0deg)"]},"delay":0,"duration":1000,"easing":"cubic-bezier(.2, 0, .5, 1)","fill":"forwards"}] [{"selector":"#anim-62fc7d9f-99d0-429e-9176-15cafbd031ec [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(-31.049465125650606%, 0, 0) translate(-25%, 0%) scale(1.5)","translate3d(0%, 0, 0) translate(0%, 0%) scale(1)"]},"delay":0,"duration":2000,"fill":"forwards"}]

Data centres with liquid-cooled GB200 NVL72 racks use less energy and have a smaller carbon footprint

[{"selector":"#anim-a77b519f-9acc-4a01-b66e-935c560443a8 [data-leaf-element=\"true\"]","keyframes":{"transform":["translate(0%, 0%) scale(1.5)","translate(0%, 0%) scale(1)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"forwards"}] [{"selector":"#anim-7d63e00f-9c19-44c0-9873-415427d78ff6","keyframes":{"transform":["translate3d(-130.65135%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-2cf3770b-084b-471c-90b2-003f31f6f1c5","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-6149c2ee-18a4-4062-89c2-9dac535d74ff","keyframes":{"transform":["scale(0.15)","scale(1)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"forwards"}]

GB200 leverages the NVIDIA Blackwell architecture’s high-bandwidth memory performance, NVLink-C2C

[{"selector":"#anim-5de27764-a7d8-408a-bc77-449d93c4bd0c [data-leaf-element=\"true\"]","keyframes":{"transform":["translate(0%, 0%) scale(1.5)","translate(0%, 0%) scale(1)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"forwards"}] [{"selector":"#anim-1f287595-1f37-4906-8c1e-a13c20e49e7c","keyframes":{"transform":["translate3d(-124.72727%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-72860e8e-f750-4436-9fb8-c844b6b067bd","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-2302a932-4e9f-4746-9598-7be2b2c0b6dd","keyframes":{"transform":["scale(0.15)","scale(1)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"forwards"}]

One such system that is capable of the job is the new NVIDIA GB200 NVL72

[{"selector":"#anim-87adcbc0-97a1-4737-91cc-0326f6f33454 [data-leaf-element=\"true\"]","keyframes":{"transform":["translate(0%, 0%) scale(1.5)","translate(0%, 0%) scale(1)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"forwards"}] [{"selector":"#anim-3ca9cd59-933b-46ff-89b4-badca3ffba5d","keyframes":{"transform":["rotate(-540deg) scale(0.1)","none"],"opacity":[0,1]},"delay":0,"duration":1000,"fill":"both","iterations":1}]

The GB200 NVL72 has 72 GPUs in one rack or two racks with 18 dual or single GB200 compute nodes

[{"selector":"#anim-62504eca-3361-41a4-9d84-85840b9e1592 [data-leaf-element=\"true\"]","keyframes":{"transform":["translate(0%, 0%) scale(1.5)","translate(0%, 0%) scale(1)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"forwards"}] [{"selector":"#anim-bb0d3573-7254-4e78-bdc7-c84616a238e0","keyframes":{"transform":["translate3d(-123.55213%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-b8224366-7866-440e-81fa-10f73ff766b8","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-625296d3-8aae-457d-8a54-3b5c652a06a0","keyframes":{"transform":["scale(0.15)","scale(1)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"forwards"}]

With unmatched speed, efficiency, and scalability, the NVIDIA Blackwell architecture ushers in a new era of computing

[{"selector":"#anim-9332ee3d-adb5-44c1-b3d2-84920fa6e69b [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(-11.238223183368396%, 0, 0) translate(-25%, 0%) scale(1.5)","translate3d(0%, 0, 0) translate(0%, 0%) scale(1)"]},"delay":0,"duration":2000,"fill":"forwards"}] [{"selector":"#anim-e189e5d6-a8c7-4ca8-afb4-f3a89a9a0cce","keyframes":{"transform":["translate3d(-122.86822%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":1000,"easing":"cubic-bezier(.2, 0, .8, 1)","fill":"both"}] [{"selector":"#anim-7d5a4c2d-0bd6-4702-ba15-58888a54fcfb","keyframes":{"transform":["rotateZ(-180deg)","rotateZ(0deg)"]},"delay":0,"duration":1000,"easing":"cubic-bezier(.2, 0, .5, 1)","fill":"forwards"}]

NVIDIA Grace CPU is a cutting-edge processor for AI, cloud, and HPC applications in modern data centres For more details Govindhtech.com

[{"selector":"#anim-0946ff3a-60aa-44b3-9794-4602646cb857","keyframes":{"transform":["translate3d(129.99998%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":1000,"easing":"cubic-bezier(.2, 0, .8, 1)","fill":"both"}] [{"selector":"#anim-a45ad503-c4f4-4539-a085-92f1cfdc503a","keyframes":{"transform":["rotateZ(180deg)","rotateZ(0deg)"]},"delay":0,"duration":1000,"easing":"cubic-bezier(.2, 0, .5, 1)","fill":"forwards"}] [{"selector":"#anim-6c71019f-e899-4b8f-aad2-e412f7b65c56 [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(-21.962890454871893%, 0, 0) translate(-25%, 0%) scale(1.5)","translate3d(0%, 0, 0) translate(0%, 0%) scale(1)"]},"delay":0,"duration":2000,"fill":"forwards"}]