NVIDIA GB300 NVL72: 50x Faster AI with Blackwell Ultra GPUs

NVIDIA GB300 NVL72 is a test-time scaling inference platform with 36 Arm-based NVIDIA Grace CPUs and 72 NVIDIA Blackwell Ultra GPUs

[{"selector":"#anim-77651741-37fc-418b-aa22-6395d44512c6","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.2, 0.6, 0.0, 1)","fill":"both"}] [{"selector":"#anim-f0a01f2a-f51c-4558-8941-7e8d40f50031","keyframes":{"transform":["translate3d(0px, 433.03509%, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.2, 0.6, 0.0, 1)","fill":"both"}] [{"selector":"#anim-844e5d83-1368-40e2-bbd9-f69346ee9113 [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(-12.514648210040342%, 0, 0)","translate3d(0%, 0, 0)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"both"}]

NVIDIA Hopper platform, AI factories with GB300 NVL72, Quantum-X800 InfiniBand or Spectrum-X Ethernet, and ConnectX-8 SuperNICS infer reasoning models 50x more

[{"selector":"#anim-d235acba-dc8a-4f5c-a2fd-1115bd0c5a6f","keyframes":{"transform":["translate3d(114.92064%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":1000,"easing":"cubic-bezier(.2, 0, .8, 1)","fill":"both"}] [{"selector":"#anim-c39e4dea-f8cd-4c79-8f98-66910c980049","keyframes":{"transform":["rotateZ(180deg)","rotateZ(0deg)"]},"delay":0,"duration":1000,"easing":"cubic-bezier(.2, 0, .5, 1)","fill":"forwards"}] [{"selector":"#anim-404a5f70-ad2f-4fa6-af67-5fe19dfca4db [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(12.514648210040342%, 0, 0)","translate3d(0%, 0, 0)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"both"}]

Compared to NVIDIA Blackwell GPUs, NVIDIA Blackwell Ultra's Tensor Cores have 1.5 times more AI computation FLOPS and 2 times the attention-layer acceleration

[{"selector":"#anim-7d38d373-3f12-444b-9607-d0ec01c294bb [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(-12.514648210040342%, 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-23e355e5-d69e-42aa-90c2-1fbb7d859090","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.2, 0.6, 0.0, 1)","fill":"both"}] [{"selector":"#anim-c69ea2f8-3d9e-4d81-81e5-b9d760b95332","keyframes":{"transform":["translate3d(0px, 143.647%, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.2, 0.6, 0.0, 1)","fill":"both"}]

NVIDIA Blackwell Ultra GPUs provide 1.5x bigger HBM3e memory, increasing AI reasoning throughput for the longest context lengths

[{"selector":"#anim-ed28122f-c8f9-41cb-85f4-597d34c72856","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.2, 0.6, 0.0, 1)","fill":"both"}] [{"selector":"#anim-e1462b63-3adc-4bf7-9944-a1dac4f2d244","keyframes":{"transform":["translate3d(0px, -157.86729%, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.2, 0.6, 0.0, 1)","fill":"both"}] [{"selector":"#anim-5f11dcb1-a344-4662-be9d-ec44dafc3cd7 [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(12.514648210040342%, 0, 0)","translate3d(0%, 0, 0)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"both"}]

The NVIDIA ConnectX-8 SuperNIC's input/output (IO) module houses two ConnectX-8 devices, giving each GPU in the NVIDIA GB300 NVL72 800 Gb/s of network access

[{"selector":"#anim-eb882cd9-69b3-4f74-b9be-f5f6d45cee68 [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(-12.514648210040342%, 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-acb7747e-3fd6-4bb9-aff1-0adae2a6081f","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.2, 0.6, 0.0, 1)","fill":"both"}] [{"selector":"#anim-9980f3f8-4006-4e5a-84e0-129dc55d743a","keyframes":{"transform":["translate3d(0px, 313.98194%, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.2, 0.6, 0.0, 1)","fill":"both"}]

It offers twice the energy efficiency of the top server processors of today along with exceptional performance and memory bandwidth

[{"selector":"#anim-95a41248-8080-46a0-82fc-588741e24d32 [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(-12.514648210040342%, 0, 0)","translate3d(0%, 0, 0)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"both"}] [{"selector":"#anim-2395b8f0-3942-4a93-9c34-7b4fce5823c3","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.2, 0.6, 0.0, 1)","fill":"both"}] [{"selector":"#anim-32425069-7e13-4f7b-9b62-656f304a6106","keyframes":{"transform":["translate3d(0px, 351.5208%, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.2, 0.6, 0.0, 1)","fill":"both"}]

AI reasoning models can achieve faster performance to the fifth-generation NVIDIA NVLink scale-up interconnect

[{"selector":"#anim-a887417e-da34-4663-bac4-f1b4ac101ff2 [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-f4676908-d0d9-45e5-988a-20aedcf0bab6","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}]

A massive GPU designed for the era of artificial intelligence reasoning is created by combining 18 superchips using NVIDIA NVLink Switch technology and NVIDIA BlueField-3 DPUs

[{"selector":"#anim-023314ea-74c0-4196-95f4-ed95c6d15bde [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(-12.514648210040342%, 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-b60f87bb-f784-4ce8-90f2-dd166840dd22","keyframes":{"opacity":[0,1]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"both"}] [{"selector":"#anim-3b6476fc-6812-4e16-b5ad-d4fe9a5f514f","keyframes":{"transform":["scale(0.3333333333333333)","scale(1)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"forwards"}]

AI reasoning models can achieve faster performance to the fifth-generation NVIDIA NVLink scale-up interconnect Read more on Govindhtech.com

[{"selector":"#anim-3f83afda-d443-47cd-9979-6be6dd300ebc","keyframes":{"opacity":[1,1]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"both"}] [{"selector":"#anim-42257eb8-888c-4afc-863e-675640145d0d","keyframes":{"transform":["scale(3)","scale(1)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"forwards"}] [{"selector":"#anim-c06b5305-21e1-4895-afb8-4889bd7b2b55 [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(12.514648210040342%, 0, 0)","translate3d(0%, 0, 0)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"both"}]