Character sheet, 12 poses

One hero render, then twelve consistent variations, the workflow behind every game and animation pitch.

The pipeline

Render the hero image, flux-1-dev
Generate 12 consistent poses, flux-1-kontext-dev

Fastest card where every stage is a real measurement: NVIDIA B300, 2.2 min for the whole job.

Every GPU, slowest job to fastest

GPU	Total time	Compute	Energy	Basis
NVIDIA B300	2.2 min	73 s	20.73 Wh	all 2 stages measured
NVIDIA B100	2.7 min	2.7 min	n/a	anchored estimate (0/2 stages measured)
NVIDIA GH200 Grace Hopper	2.8 min	2.8 min	n/a	anchored estimate (0/2 stages measured)
NVIDIA B200	2.9 min	2.2 min	32.04 Wh	all 2 stages measured
NVIDIA H100 NVL	2.9 min	2.9 min	n/a	anchored estimate (0/2 stages measured)
NVIDIA H800 80GB	2.9 min	2.9 min	n/a	anchored estimate (0/2 stages measured)
NVIDIA H100 PCIe	3.4 min	3.4 min	n/a	anchored estimate (0/2 stages measured)
NVIDIA H100 80GB HBM3	3.7 min	2.9 min	33.26 Wh	all 2 stages measured
NVIDIA H200	3.9 min	2.8 min	32.58 Wh	all 2 stages measured
NVIDIA RTX PRO 6000 Blackwell Server Edition	4.6 min	4.1 min	37.55 Wh	all 2 stages measured
NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition	5.2 min	5.2 min	n/a	anchored estimate (0/2 stages measured)
NVIDIA RTX PRO 6000 Blackwell Workstation Edition	6.1 min	4 min	40.17 Wh	all 2 stages measured
NVIDIA A100 40GB SXM4	6.3 min	6.3 min	n/a	anchored estimate (0/2 stages measured)
NVIDIA A800 80GB	6.3 min	6.3 min	n/a	anchored estimate (0/2 stages measured)
NVIDIA A100 40GB PCIe	6.8 min	6.8 min	n/a	anchored estimate (0/2 stages measured)
NVIDIA RTX PRO 5000 Blackwell	7.6 min	6.9 min	34.5 Wh	all 2 stages measured
NVIDIA A100 80GB SXM4	7.6 min	6.3 min	41.23 Wh	all 2 stages measured
NVIDIA A100 80GB PCIe	7.7 min	6.8 min	33.46 Wh	all 2 stages measured
NVIDIA L40S	8.4 min	7.3 min	41.61 Wh	all 2 stages measured
NVIDIA L40	12.1 min	11.2 min	55.81 Wh	all 2 stages measured
NVIDIA RTX 6000 Ada Generation	12.6 min	12.1 min	60.12 Wh	all 2 stages measured
NVIDIA RTX PRO 4500 Blackwell	14.6 min	14.4 min	35.78 Wh	all 2 stages measured
NVIDIA RTX A6000	15.2 min	11.9 min	58.73 Wh	all 2 stages measured
NVIDIA RTX 5000 Ada Generation	16 min	15.9 min	46.88 Wh	all 2 stages measured
NVIDIA RTX 5880 Ada Generation	18.3 min	18.3 min	n/a	anchored estimate (0/2 stages measured)
AMD Radeon Pro W7900	24.2 min	24.2 min	n/a	anchored estimate (0/2 stages measured)
AMD Radeon Pro W7800	36.6 min	36.6 min	n/a	anchored estimate (0/2 stages measured)
AMD Radeon Pro W6800	39.4 min	39.4 min	n/a	anchored estimate (0/2 stages measured)

Cards that can't run this pipeline

Every one of these fails on the same kind of wall, a stage that will not fit in VRAM.

AMD Radeon RX 6700, blocked at "Render the hero image" (FLUX.1 dev)
AMD Radeon RX 6800, blocked at "Render the hero image" (FLUX.1 dev)
AMD Radeon RX 6800 XT, blocked at "Render the hero image" (FLUX.1 dev)
AMD Radeon RX 6900 XT, blocked at "Render the hero image" (FLUX.1 dev)
AMD Radeon RX 6950 XT, blocked at "Render the hero image" (FLUX.1 dev)
AMD Radeon RX 7600, blocked at "Render the hero image" (FLUX.1 dev)
AMD Radeon RX 7700 XT, blocked at "Render the hero image" (FLUX.1 dev)
AMD Radeon RX 7800 XT, blocked at "Render the hero image" (FLUX.1 dev)
AMD Radeon RX 7900 XT, blocked at "Render the hero image" (FLUX.1 dev)
AMD Radeon RX 7900 XTX, blocked at "Render the hero image" (FLUX.1 dev)
AMD Radeon RX 9070, blocked at "Render the hero image" (FLUX.1 dev)
AMD Radeon RX 9070 XT, blocked at "Render the hero image" (FLUX.1 dev)
GeForce GTX 1080 Ti, blocked at "Render the hero image" (FLUX.1 dev)
GeForce RTX 4060, blocked at "Render the hero image" (FLUX.1 dev)
GeForce RTX 4080 Super, blocked at "Render the hero image" (FLUX.1 dev)
GeForce RTX 5060 Ti, blocked at "Render the hero image" (FLUX.1 dev)
GeForce RTX 5070, blocked at "Render the hero image" (FLUX.1 dev)
GeForce RTX 5070 Ti, blocked at "Render the hero image" (FLUX.1 dev)
GeForce RTX 5080, blocked at "Render the hero image" (FLUX.1 dev)
Intel Arc A750, blocked at "Render the hero image" (FLUX.1 dev)
Intel Arc A770 Limited Edition, blocked at "Render the hero image" (FLUX.1 dev)
Intel Arc B580, blocked at "Render the hero image" (FLUX.1 dev)
Intel Arc Pro A60, blocked at "Render the hero image" (FLUX.1 dev)
NVIDIA A10G, blocked at "Render the hero image" (FLUX.1 dev)
NVIDIA GeForce GTX 1070 Ti, blocked at "Render the hero image" (FLUX.1 dev)
NVIDIA GeForce GTX 1080, blocked at "Render the hero image" (FLUX.1 dev)
NVIDIA GeForce GTX 1660, blocked at "Render the hero image" (FLUX.1 dev)
NVIDIA GeForce GTX 1660 Super, blocked at "Render the hero image" (FLUX.1 dev)
NVIDIA GeForce GTX 1660 Ti, blocked at "Render the hero image" (FLUX.1 dev)
NVIDIA GeForce RTX 2060, blocked at "Render the hero image" (FLUX.1 dev)
NVIDIA GeForce RTX 2060 Super, blocked at "Render the hero image" (FLUX.1 dev)
NVIDIA GeForce RTX 2070, blocked at "Render the hero image" (FLUX.1 dev)
NVIDIA GeForce RTX 2070 SUPER, blocked at "Render the hero image" (FLUX.1 dev)
NVIDIA GeForce RTX 2080 Founders Edition, blocked at "Render the hero image" (FLUX.1 dev)
NVIDIA GeForce RTX 2080 Super, blocked at "Render the hero image" (FLUX.1 dev)
NVIDIA GeForce RTX 2080 Ti Founders Edition, blocked at "Render the hero image" (FLUX.1 dev)
NVIDIA GeForce RTX 3050, blocked at "Render the hero image" (FLUX.1 dev)
NVIDIA GeForce RTX 3060, blocked at "Render the hero image" (FLUX.1 dev)
NVIDIA GeForce RTX 3060 Ti, blocked at "Render the hero image" (FLUX.1 dev)
NVIDIA GeForce RTX 3070 Founders Edition, blocked at "Render the hero image" (FLUX.1 dev)

…and 31 more.

How these numbers are built

Each stage time is the quantity of work divided by that card's measured throughput for that model, from our own bench. The pipeline is assumed to run batched, every image, then every clip, so each model loads once. Model load time is added where we recorded it; our text-generation runs don't carry a load measurement yet, so pipelines with a language-model stage are slightly optimistic. Nothing here is a single timed run of the whole pipeline, and we don't present it as one.