docs(gpu-support): add HEVC encode column and popular AI GPU section

[Copilot]

HEVC (H.265) is an important transcoding capability on the Livepeer network today. The GPU support reference pages lacked explicit HEVC support visibility and had no dedicated guidance for operators selecting GPUs for AI workloads.

Changes

v2/orchestrators/resources/reference/gpu-support.mdx + v2/orchestrators/resources/gpu-support.mdx

• HEVC Encode column added to the "Supported GPU Families" matrix. All listed families support HEVC via NVENC; notable callouts added inline:
  • RTX 40xx: AV1 and HEVC 10-bit encode
  • RTX 20xx: HEVC B-frames supported
• New section: Popular GPUs for AI Workloads — 10-row table covering consumer and data centre options (RTX 4090, RTX 3090, RTX 4070 Ti Super, RTX 4080 Super, T4, A10G, L4, L40S, A100, H100) with VRAM, HEVC encode flag, best-fit use case, and notes. Includes an <Info> callout recommending 24 GB cards for combined AI + transcoding.
• HEVC and H.265 added to frontmatter keywords.
• lastVerified updated to 2026-04-07 in both files.

Scope

• v2/orchestrators/resources/reference/gpu-support.mdx
• v2/orchestrators/resources/gpu-support.mdx

Out of scope: v1/references/go-livepeer/gpu-support.mdx (frozen), docs.json (no routing changes).

Validation

• node operations/scripts/validators/content/structure/lint-structure.js --check — no files flagged
• node operations/scripts/validators/governance/compliance/check-agent-docs-freshness.js --json — passed
• Pre-commit hooks — passed

Follow-up Tasks

none

Type of Change

• Bug fix (fixes an issue)
• New content (adds new documentation)
• Content update (improves existing content)
• Style/formatting fix
• Information architecture change
• Other (please describe)

Related Issues

Related to #

Changes Made

• Added HEVC Encode column to Supported GPU Families table in both active v2 gpu-support files
• Added "Popular GPUs for AI Workloads" section with 10-row reference table
• Updated frontmatter keywords and lastVerified date

Testing

• Tested locally with npm run dev
• Verified all links work
• Checked formatting and style
• Reviewed against style guides
• Screenshots (if UI changes)

Screenshots (if applicable)

N/A — table content changes only.

Checklist

• My changes follow the style guides
• I've reviewed the Component Library for available components
• I've updated related pages if needed
• I've checked for broken links
• My changes are clear and easy to understand
• I've tested locally
• I've added/updated keywords and metadata if needed

Additional Notes

The hyphen used in the GTX 16xx "No Tensor cores" note is intentional — the file's own style TODO explicitly states "No em-dashes are used (instead use standard -)". Pre-existing em dashes in other rows are untouched as out-of-scope cleanup.

Governance Approval

Not required for this PR.

Copy Governance Gate - L5

Auto-checked by CI (lint-copy.js + lint-structure.js)

• No Tier 1 banned words or phrases
• No unresolved {/* REVIEW: */} flags in rendered content
• No non-USD currency outside declared regional scope
• No empty cells in decision tables
• pageType, audience, lastVerified present in frontmatter

Human review required

Sequence (L2)

• Page opens with operator outcome - value kept, no conditionals
• Majority operator path is featured before variants
• Introduced paths are developed on-page or linked

Completeness (L3)

• Decision aids (tables, matrices, flowcharts) are complete - no placeholders
• Every warning or consequence is followed by a prevention action and link
• Primary actions and URLs appear in body copy (not only in accordions)

Tone (L4)

• No sentence ends on a hedge, disclaimer, or restatement
• No section ends with a Note that apologises rather than forward-points

Brief compliance

• L0 (Q1-Q4) was completed before drafting
• L1 persona mapping table was completed before drafting

[github-actions]

📊 V2 Browser Sweep Results

❌ Test report not found. The test may have failed before report generation.

• Likely failed in setup/cache/server-start steps before npm run test:v2-pages.