Skip to content

docs: HW-vs-SW Lumen measured — +20 ms/frame on the 760M#110

Merged
proggeramlug merged 1 commit into
mainfrom
docs/hw-lumen-measured
Jul 16, 2026
Merged

docs: HW-vs-SW Lumen measured — +20 ms/frame on the 760M#110
proggeramlug merged 1 commit into
mainfrom
docs/hw-lumen-measured

Conversation

@proggeramlug

@proggeramlug proggeramlug commented Jul 16, 2026

Copy link
Copy Markdown
Contributor

Answers the question EN-023 and EN-058(c) have carried since the DXC work: granting EXPERIMENTAL_RAY_QUERY costs 14.4 vs 20.3 fps (~+20 ms/frame) on the shooter, same build, PT off — because SSGI/WSRC silently flip to the HW trace once the TLAS exists, and every skinned draw starts paying per-frame pre-skin + BLAS builds. The hiz-screen boot line is init-time only and misleading. Whether HW Lumen's quality is worth 20 ms is the product owner's call (needs GI screenshot A/B); BLOOM_FORCE_SW_GI=1 is the per-box escape hatch.

Summary by CodeRabbit

  • Documentation
    • Added measured performance comparisons between hardware and software Lumen GI paths on the 760M.
    • Documented the observed quality-versus-frame-rate tradeoff to support default configuration decisions.
    • Clarified startup logging behavior and noted the available per-device option to force software GI.

@proggeramlug
proggeramlug merged commit 580dc17 into main Jul 16, 2026
@proggeramlug
proggeramlug deleted the docs/hw-lumen-measured branch July 16, 2026 11:31
@coderabbitai

coderabbitai Bot commented Jul 16, 2026

Copy link
Copy Markdown

Review Change Stack

Caution

Review failed

The pull request is closed.

ℹ️ Recent review info
⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: b00c41d6-8639-4b25-b68a-9d09d266a41c

📥 Commits

Reviewing files that changed from the base of the PR and between 314546d and b34c1f3.

📒 Files selected for processing (1)
  • docs/tickets.md

📝 Walkthrough

Walkthrough

Adds an EN-052 follow-on note documenting measured hardware versus software Lumen performance on a 760M, backend logging timing, DXC-based enablement, and the BLOOM_FORCE_SW_GI=1 override.

Changes

Lumen measurements

Layer / File(s) Summary
Measured Lumen comparison and follow-on findings
docs/tickets.md
Documents the hardware and forced software GI PERFTEST results, DXC DLL enablement behavior, pre-TLAS backend logging, and the software GI workaround.

Estimated code review effort: 1 (Trivial) | ~2 minutes

Possibly related PRs

✨ Finishing Touches
🧪 Generate unit tests (beta)
  • Create PR with unit tests
  • Commit unit tests in branch docs/hw-lumen-measured

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

Comment @coderabbitai help to get the list of available commands.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant