LTX-2.3 video gen vs GPU power cap sweep — 31 caps × 3 runs#16
Merged
Conversation
Tower2 RTX PRO 6000 Blackwell GPU0, ComfyUI two-stage T2V workflow (640x352 / 4-step → upsampler → 1280x704 / 8-step), 121-frame 5s clip, identical "founder gets an idea at sunrise" booth prompt across all 93 generations. Power cap swept 600W → 300W in 10W steps via nvidia-smi -pl. Headline: cap genuinely binds for diffusion (peak draw within 1-2W of every cap tested), unlike LLM single-stream which plateaus at ~510W on the same hardware. Going 500W → 600W buys +11.0% throughput (4.54s saved on a 41.35s gen). The V/f knee sits around 360-400W; below that, each 10W removed costs ~3% per-gen time. Contains: README, full findings.md with per-cap table + slope analysis + LLM cross-comparison + booth-day operating-point recommendation, raw combined CSV (94 rows), per-phase CSVs, sweep harness source, render- table renderer, and the full stdout from both phase runs. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
vllm-power-sweep-2026-04-29/— same hardware, fundamentally different workload behavior. Diffusion's cap binds (peak GPU draw within 1-2W of every cap); LLM single-stream plateaus at ~510W regardless of cap.Test plan
render-table.mjs(reproducible)🤖 Generated with Claude Code