TensorRT Acceleration for Stable Diffusion 3

AI PCs are here with NVIDIA RTX and GeForce RTX It also introduces new nomenclature that can confuse desktop and laptop buyers and a new approach to evaluate AI-accelerated job performance

[{"selector":"#anim-1437435f-0d6a-4712-9f83-69dd14c15442 [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-285076e0-752b-4ade-8173-71111ab8c6c9","keyframes":{"transform":["translate3d(115.18987%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":1000,"easing":"cubic-bezier(.2, 0, .8, 1)","fill":"both"}] [{"selector":"#anim-6aad10bf-1c12-41ab-ab70-880492993a25","keyframes":{"transform":["rotateZ(180deg)","rotateZ(0deg)"]},"delay":0,"duration":1000,"easing":"cubic-bezier(.2, 0, .5, 1)","fill":"forwards"}]

The AI Decoded series showcases new RTX PC hardware, software, tools, and accelerations while demystifying AI

[{"selector":"#anim-323173f4-9956-4526-8935-6be66cd69601 [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(31.794021900532332%, 0, 0)","translate3d(0%, 0, 0)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"both"}] [{"selector":"#anim-68388f17-e513-479e-91a3-562210ed74d6","keyframes":{"transform":["translate3d(-115.18988%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":1000,"easing":"cubic-bezier(.2, 0, .8, 1)","fill":"both"}] [{"selector":"#anim-489e43b0-23b5-4ac9-97a6-06331d99fbf2","keyframes":{"transform":["rotateZ(-180deg)","rotateZ(0deg)"]},"delay":0,"duration":1000,"easing":"cubic-bezier(.2, 0, .5, 1)","fill":"forwards"}]

Many generative AI jobs are harder. The NVIDIA RTX and GeForce RTX GPUs offer unprecedented speed for all generative activities; the GeForce RTX 4090 GPU has over 1,300 TOPS

[{"selector":"#anim-68791f05-07f8-46a9-9834-6cc2199069fb [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-66c91afe-a8f1-4aa6-bcb2-8a18f1a52ad9","keyframes":{"transform":["rotate(-540deg) scale(0.1)","none"],"opacity":[0,1]},"delay":0,"duration":1000,"fill":"both","iterations":1}]

AI super resolution in PC gaming, text/video visual synthesis, local LLM querying, and other activities require computer power

[{"selector":"#anim-dbf3fdab-2981-49fd-a820-f9f5b99f02b4 [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-116c5ca2-8e90-4975-9d54-62bd791736a2","keyframes":{"transform":["translate3d(-118.21191%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":1000,"easing":"cubic-bezier(.2, 0, .8, 1)","fill":"both"}] [{"selector":"#anim-31efc845-3162-4cd0-bb5b-b8edd33fc1e7","keyframes":{"transform":["rotateZ(-180deg)","rotateZ(0deg)"]},"delay":0,"duration":1000,"easing":"cubic-bezier(.2, 0, .5, 1)","fill":"forwards"}]

RTX GPUs are ideal for LLMs due to their large VRAM, Tensor Cores, and TensorRT-LLM software

[{"selector":"#anim-44819985-fbeb-4f51-922d-841389e8a897 [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(31.254577522972255%, 0, 0)","translate3d(0%, 0, 0)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"both"}] [{"selector":"#anim-e84ec516-8350-4091-bc99-978e18cbcbc0","keyframes":{"transform":["translate3d(-118.04512%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-b7b1e5b7-8bc6-4346-b88d-b71421a9bf97","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-8245ae49-fa04-4c15-9a3f-3b556741d624","keyframes":{"transform":["scale(0.15)","scale(1)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"forwards"}]

GeForce and NVIDIA RTX GPUs with 24GB and 48GB high-speed VRAM can support larger models and batch sizes

[{"selector":"#anim-a5e3537d-0114-4657-a29f-0d3fec55360a [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-f8e9ef42-a65c-4cc0-9bfa-69718d9b1fd2","keyframes":{"transform":["translate3d(121.33333%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":1000,"easing":"cubic-bezier(.2, 0, .8, 1)","fill":"both"}] [{"selector":"#anim-23ff3b01-1169-4d04-8d20-d4f3204fe6e3","keyframes":{"transform":["rotateZ(180deg)","rotateZ(0deg)"]},"delay":0,"duration":1000,"easing":"cubic-bezier(.2, 0, .5, 1)","fill":"forwards"}]

NVIDIA TensorRT SDK offers the highest-performance generative AI on over 100 million Windows PCs and workstations powered by RTX GPUs

[{"selector":"#anim-81ebdf8d-6734-46cb-aae2-3d332732d364 [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-f3c1a77f-1362-44a3-acc6-c489fb8f1a9a","keyframes":{"transform":["rotate(-540deg) scale(0.1)","none"],"opacity":[0,1]},"delay":0,"duration":1000,"fill":"both","iterations":1}]

processing the AI model using an RTX GPU instead of a CPU or NPU, these results are faster

[{"selector":"#anim-9a61ddd0-6c43-4dd5-8028-096cf72ccc4d [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-c4ba4dfc-09a9-4ccb-9029-84903dcc57be","keyframes":{"transform":["translate3d(-129.87013%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":1000,"easing":"cubic-bezier(.2, 0, .8, 1)","fill":"both"}] [{"selector":"#anim-981d89ad-5e0b-466d-a810-050a71839b5f","keyframes":{"transform":["rotateZ(-180deg)","rotateZ(0deg)"]},"delay":0,"duration":1000,"easing":"cubic-bezier(.2, 0, .5, 1)","fill":"forwards"}]

Performance is improved using TensorRT in Automatic1111. SDXL Base checkpoints allow RTX users to take pictures twice as fast, simplifying Stable Diffusion

[{"selector":"#anim-a56d6257-9c38-46cb-951d-8c13fb5e59a7 [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-9ee57968-42f3-4ac2-8e1f-2ff7dcbd8554","keyframes":{"transform":["translate3d(-120%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":1000,"easing":"cubic-bezier(.2, 0, .8, 1)","fill":"both"}] [{"selector":"#anim-f2884284-5768-47e9-8faa-16733e91275f","keyframes":{"transform":["rotateZ(-180deg)","rotateZ(0deg)"]},"delay":0,"duration":1000,"easing":"cubic-bezier(.2, 0, .5, 1)","fill":"forwards"}]

TensorRT acceleration speeds the UL Procyon AI Image Generation benchmark on a GeForce RTX 4080 SUPER GPU by 50% For more details Govindhtech.com

[{"selector":"#anim-3acf5cc7-d798-468c-a2b5-cf35dfcd8bc2 [data-leaf-element=\"true\"]","keyframes":{"transform":["translate3d(37.005537895833655%, 0, 0)","translate3d(0%, 0, 0)"]},"delay":0,"duration":2000,"easing":"cubic-bezier(.3,0,.55,1)","fill":"both"}] [{"selector":"#anim-1fd3320d-5bac-4cde-8852-77c78ee62c4e","keyframes":{"transform":["translate3d(-120.5298%, 0px, 0)","translate3d(0px, 0px, 0)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-7c837789-635a-43e1-bb32-406f0b0da9b5","keyframes":{"opacity":[0,1]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"both"}] [{"selector":"#anim-5bc753e8-c862-4b24-b76a-ccb4b5e4d6d6","keyframes":{"transform":["scale(0.15)","scale(1)"]},"delay":0,"duration":600,"easing":"cubic-bezier(0.4, 0.4, 0.0, 1)","fill":"forwards"}]