diff --git a/website/Caddyfile b/website/Caddyfile
index 1ae4639f5..1571566ad 100644
--- a/website/Caddyfile
+++ b/website/Caddyfile
@@ -20,7 +20,7 @@
Referrer-Policy strict-origin-when-cross-origin
X-Frame-Options DENY
Permissions-Policy "interest-cohort=(), browsing-topics=(), camera=(), microphone=(), geolocation=()"
- Content-Security-Policy "default-src 'self'; script-src 'self' {$CSP_SCRIPT}; style-src 'self' {$CSP_STYLE}; img-src 'self' data: https://avatars.githubusercontent.com; font-src 'self'; connect-src 'self' https://api.github.com https://raw.githubusercontent.com; frame-ancestors 'none'; base-uri 'self'; form-action 'self'"
+ Content-Security-Policy "default-src 'self'; script-src 'self' {$CSP_SCRIPT}; style-src 'self' {$CSP_STYLE}; img-src 'self' data: https://avatars.githubusercontent.com; font-src 'self'; connect-src 'self' https://api.github.com https://raw.githubusercontent.com https://telemetry.playhorizon.dev; frame-ancestors 'none'; base-uri 'self'; form-action 'self'"
-Server
}
@assets path /assets/*
@@ -31,9 +31,11 @@
redir /server /docs/hosting permanent
redir /server.html /docs/hosting permanent
rewrite /privacy /privacy.html
+ rewrite /releases /releases.html
+ rewrite /benchmarks /benchmarks.html
@doc path_regexp doc ^/docs/([a-z-]+)$
rewrite @doc /docs/{re.doc.1}.html
- @pages path / /index.html /404.html /privacy.html /docs /docs/* /sitemap.xml
+ @pages path / /index.html /404.html /privacy.html /releases /releases.html /benchmarks /benchmarks.html /benchmarks.json /docs /docs/* /sitemap.xml
header @pages Cache-Control "no-cache"
@legal path /LEGAL.md /fonts/*.txt
header @legal Content-Type "text/plain; charset=utf-8"
diff --git a/website/README.md b/website/README.md
index 36c76c331..8ac6e3910 100644
--- a/website/README.md
+++ b/website/README.md
@@ -2,12 +2,12 @@
Live at **https://playhorizon.dev**.
-A landing page and a documentation section in Horizon's design language (`docs/design-system.md`), with the tokens of
+A landing page, a Releases page, a Benchmarks page and a documentation section in Horizon's design language (`docs/design-system.md`), with the tokens of
`crates/horizon-theme` (generated into `src/tokens.css` by `cargo run -p xtask -- tokens`):
the Blocks look: Pixelify Sans headings over Geist text, Horizon's purple in a light and a dark
appearance (the system's, or the toggle's), square panels with an outline, a bevel and a lip, a
-stone texture and a pixel landscape drawn in code (`scripts/art.mjs`), and short, eased motion. A headline and the download, then a scroll through what it is, the screenshots, the parity grid, the launcher, the patch
-notes, the community links and the contributors. Plain HTML, CSS and JavaScript; no framework and
+stone texture and a pixel landscape drawn in code (`scripts/art.mjs`), and short, eased motion. A headline and the download, then a scroll through what it is, the screenshots, the parity grid, the launcher, the newest
+releases, the community links and the contributors. Plain HTML, CSS and JavaScript; no framework and
no runtime dependencies.
## Layout
@@ -20,14 +20,24 @@ serve.mjs a tiny static server for local checks
src/index.html the page, with {{placeholders}} build.mjs fills
src/404.html the not-found page
src/privacy.html what the launcher can send when players allow it, retention, deletion (/privacy)
+src/releases.html every release, newest first, with what it adds and fixes, its notes and downloads (/releases)
+src/benchmarks.html Horizon against vanilla 26.3 on our machines, the hardware picker, players' results (/benchmarks)
src/docs.html the template of every documentation page (/docs/...), with its left navigation
src/tokens.css the design tokens, generated from crates/horizon-theme (do not edit)
src/style.css every style, on top of the tokens
-src/main.js download buttons, patch notes, links, contributors, nav and menu, tabs, lightbox, motion
+src/main.js the home page: download buttons, recent releases, links, contributors, nav, tabs, lightbox, motion
+src/chrome.js the light/dark switch, the phone menu and GitHub fetching, shared by the pages
+src/releases-page.js the Releases page: the channel filter and the refresh from GitHub
+src/benchmarks-page.js the Benchmarks page: players' results from the telemetry server, the picker
+src/bench.js the Benchmarks tables, shared by build and page
+src/hardware.js the spec table estimates use (GPU TFLOPS and bandwidth, CPU single-thread score)
+src/estimate.js estimates for unmeasured hardware from the nearest measured machine
src/markdown.js a small Markdown renderer (escapes everything), shared by build and page
-src/releases.js the patch notes' markup, shared by build and page
+src/releases.js the home page's "Recently added" and the Releases page's markup, shared by build and page
src/contributors.js the contributors' filtering and markup, shared by build and page
-public/ fonts, screenshots, icons, og.png, robots.txt: copied as is (sitemap.xml is written by the build)
+public/ fonts, screenshots, icons, og.png, robots.txt, benchmarks.json: copied as is (sitemap.xml is written by the build)
+data/benchmarks/ our raw benchmark reports (tools/perf_bench.py) and fleet.json, the list scripts/benchmarks.mjs reads
+scripts/benchmarks.mjs benchmark reports -> public/benchmarks.json (ours, and players' groups of 3+ installs)
scripts/assets.mjs rebuilds public/ from the repo (fonts to WOFF2, images to WebP, the pixel art)
scripts/art.mjs draws the stone tile and the hero's landscape (PNG, one pixel per texel)
scripts/shots.mjs Playwright screenshots at every width, and og.png
@@ -46,8 +56,9 @@ shots/ (gitignored) full-page screenshots at 360–1920 px, a 1250×
the checkout; on Railway, which builds only this folder, from `main` on GitHub. The build
fails if the README has no parity table, so a broken README never deploys.
- **Fetched by the page**: the releases (the download button picks the newest
- `launcher-v*` release's asset for the visitor's OS, and the three latest game releases'
- notes are rendered), the contributors (avatars come from `avatars.githubusercontent.com`),
+ `launcher-v*` release's asset for the visitor's OS, the home page lists the newest game releases, and
+ the Releases page reloads its list when GitHub has a newer release), players' benchmark results
+ (`GET https://telemetry.playhorizon.dev/v1/benchmarks/summary`, 3-second timeout), the contributors (avatars come from `avatars.githubusercontent.com`),
and `links.json` from `main`. Each
falls back to what was built in, and the download button to the releases page.
@@ -107,3 +118,18 @@ GitHub's anonymous rate limit is hit.
- Parity is worded as the README words it: 100% of what the harnesses compare.
- No cookies, analytics, trackers or third-party fonts and scripts.
- Motion stops under `prefers-reduced-motion`; nothing hijacks the scroll.
+
+## Benchmarks
+
+`public/benchmarks.json` is what the Benchmarks page shows: `{ format: "horizon-benchmark-summary", version: 1,
+generated, min_installs, groups }`, one group per GPU, CPU, render distance and game (`horizon`, `vanilla`) with
+`medians` (`avg_fps`, `low_1pct_fps`, `p95_ms`, `world_ms`, `chunk_load_ms`, `rss_peak_mb`, `rss_steady_mb`).
+Groups `"source": "us"` are our runs (with the machine, date, version and the four scenes); groups `"source":
+"players"` come from players' `benchmark` uploads (docs/telemetry.md) and need at least `min_installs` (3)
+different installs. The telemetry server's `GET /v1/benchmarks/summary` answers in the same shape; the page asks it
+for players' groups and keeps the baked ones when it doesn't answer. The rules are in `scripts/benchmarks.mjs`, with
+a fixture set (`scripts/fixtures/benchmarks/`) the server's test shares.
+
+To publish new measurements: run `tools/perf_bench.py` for Horizon and vanilla on a machine (docs/benchmark.md), copy
+the reports into `data/benchmarks/`, list them in `data/benchmarks/fleet.json`, then
+`node scripts/benchmarks.mjs --ours data/benchmarks/fleet.json`.
diff --git a/website/build.mjs b/website/build.mjs
index acb121538..698079dc1 100644
--- a/website/build.mjs
+++ b/website/build.mjs
@@ -11,7 +11,9 @@ import { createHash } from "node:crypto";
import { join, dirname } from "node:path";
import { fileURLToPath } from "node:url";
import { markdown, inline, section, escape } from "./src/markdown.js";
-import { releasesHtml } from "./src/releases.js";
+import { recentHtml, releasesPageHtml } from "./src/releases.js";
+import { machinesHtml, playersHtml } from "./src/bench.js";
+import { CPUS, GPUS, singleThread } from "./src/hardware.js";
import { isGame, isLauncher, launcherVersion, newestFirst } from "./src/versions.js";
import { contributorsFrom, contributorsHtml } from "./src/contributors.js";
@@ -41,11 +43,18 @@ try {
// An optional token (a Railway variable, or `gh auth token` locally) lifts the rate limit.
const githubHeaders = { Accept: "application/vnd.github+json", ...(process.env.GITHUB_TOKEN ? { Authorization: `Bearer ${process.env.GITHUB_TOKEN}` } : {}) };
+// Every release (the Releases page lists them all), a page of 100 at a time.
let releases = [];
try {
- const res = await fetch(`https://api.github.com/repos/${REPO}/releases?per_page=100`, { headers: githubHeaders, signal: AbortSignal.timeout(15000) });
- if (res.ok) releases = (await res.json()).filter((r) => !r.draft);
- else console.warn(`releases: ${res.status} from GitHub; the page will load them itself`);
+ let res;
+ for (let page = 1; page <= 5; page++) {
+ res = await fetch(`https://api.github.com/repos/${REPO}/releases?per_page=100&page=${page}`, { headers: githubHeaders, signal: AbortSignal.timeout(15000) });
+ if (!res.ok) break;
+ const batch = await res.json();
+ releases.push(...batch.filter((r) => !r.draft));
+ if (batch.length < 100) break;
+ }
+ if (!res.ok) console.warn(`releases: ${res.status} from GitHub; the pages will load them themselves`);
// A day of betas can push the stable release out of the newest 100: it is GitHub's "latest" (betas never are).
if (res.ok && !releases.some((r) => isGame(r.tag_name) && !r.prerelease)) {
const latest = await fetch(`https://api.github.com/repos/${REPO}/releases/latest`, { headers: githubHeaders, signal: AbortSignal.timeout(15000) });
@@ -110,14 +119,21 @@ const parityCells = body
const matching = body.filter(([, score]) => Number(score) === 1).length;
const parityHeadline = matching === body.length ? "Every subsystem matches vanilla." : `${matching} of ${body.length} subsystems match vanilla.`;
-// --- The latest notes -------------------------------------------------
+// --- The releases ----------------------------------------------------------------------
-// Stable releases only: betas become stable 6 hours later (docs/release-flow.md, "Channels"). Newest first by
-// tag: every 0.2 tag is newer than the dated ones (src/versions.js).
+// Newest first by tag: every 0.2 tag is newer than the dated ones (src/versions.js). Alpha builds are GitHub
+// pre-releases, Stable the others (docs/release-flow.md, "Channels").
const byTag = (a, b) => newestFirst(a.tag_name, b.tag_name);
-const game = releases.filter((r) => isGame(r.tag_name) && !r.prerelease).sort(byTag).slice(0, 3);
+const ours = releases.filter((r) => isGame(r.tag_name) || isLauncher(r.tag_name)).sort(byTag);
+const game = ours.filter((r) => isGame(r.tag_name));
const launcher = releases.filter((r) => isLauncher(r.tag_name) && !r.prerelease).sort(byTag)[0];
+// --- The benchmarks ----------------------------------------------------------------------
+
+// public/benchmarks.json: our measurements (scripts/benchmarks.mjs from data/benchmarks/fleet.json) and, once the
+// telemetry server has some, players' groups. The page shows it baked and asks the server for newer player results.
+const benchSummary = JSON.parse(await readFile(join(site, "public", "benchmarks.json"), "utf8"));
+
// --- The page --------------------------------------------------------------------------
const fingerprint = (text) => createHash("sha256").update(text).digest("hex").slice(0, 10);
@@ -152,7 +168,9 @@ await writeFile(join(dist, "favicon.ico"), Buffer.concat([ico, png32]));
// The page's modules, fingerprinted; each import of a shared module is rewritten to its name.
const names = {};
const rewrite = (code) => code.replace(/"\.\/(\w+\.js)"/g, (all, file) => (names[file] ? `"./${names[file]}"` : all));
-for (const file of ["markdown.js", "releases.js", "contributors.js", "versions.js", "main.js"]) {
+// Dependencies before what imports them.
+const modules = ["markdown.js", "versions.js", "chrome.js", "hardware.js", "estimate.js", "releases.js", "contributors.js", "bench.js", "main.js", "releases-page.js", "benchmarks-page.js"];
+for (const file of modules) {
const code = rewrite(await readFile(join(site, "src", file), "utf8"));
names[file] = `${file.replace(".js", "")}.${fingerprint(code)}.js`;
await writeFile(join(dist, "assets", names[file]), code);
@@ -180,9 +198,53 @@ const baked = {
})),
};
-const notesHtml =
- releasesHtml(game) ||
- `
Read them on GitHub`;
+
+// The top bar and the footer of the pages beside the home page (Releases, Benchmarks).
+const navLinks = [["/#about", "About"], ["/releases", "Releases"], ["/benchmarks", "Benchmarks"], ["/docs/", "Docs"]];
+const siteNav = (current) => `
+
We run Horizon and vanilla Minecraft: Java Edition 26.3 one after the other on the same computer: a fresh world from the same seed, the same settings, and the same four flights over a village, a forest, the ocean and a mountain. Every figure comes from the same code for both games. Where Java does better, the tables say so.
+
+
+
Measured by us
+
The project's own machines, at render distance 12 (vanilla's default) and 32. VSync off, no frame cap, a 1280x720 window, simulation distance 12, everything else at vanilla's defaults. Each figure is the median over the runs; the frame figures are averaged over the four flights.
+
+{{MACHINES}}
+
+
+
Not yet measured against Java: the desktop (RTX 4080, Core i7-13700KF). It is in daily use; the comparison runs on it once it is free. The last figures from it are Horizon alone, from an older build (2026-10-13, offscreen, tools/graphics_bench.py): over the village, 2,409 FPS at render distance 8, 1,041 at 16 and 48 at 32, where that build still stuttered while loading. They are not comparable with the tables above; the picker estimates the RTX 4080 from the measured machines instead.
+
+
+
+
+
Your hardware
+
Pick a graphics card and a processor. If we or at least three players have measured that pair, you see what was measured; otherwise an estimate from the nearest measured machine, as a range.
+
+
The picker needs JavaScript; the tables above and the method below work without it.
+
+
+
+
From players
+
Players who allow benchmark uploads in the launcher's data collection settings share the report of horizon-client --benchmark standard. Results are grouped by graphics card, processor and render distance, and a group shows only once at least 3 different installs are in it, so no one can be picked out. Only medians are published: no install ID, date, driver, memory size or screen.
+
{{PLAYERS_SOURCE}}
+
+{{PLAYERS}}
+
+
+
+
+
How the estimates work
+
+
An estimate starts from the measured machine closest to the one you picked, then scales its figures with a small table of specifications we wrote ourselves (no benchmark database):
+
+
GPU factor: the geometric mean of the two graphics cards' ratio of FP32 shader throughput (TFLOPS) and of memory bandwidth (GB/s), as their makers state them.
+
CPU factor: the ratio of the two processors' single-thread scores: top boost clock times a rough factor per chip generation (Zen 4 = 1).
+
Frame rate: when the measured run split each frame into CPU and GPU time, the slower side is scaled by its factor; otherwise by the geometric mean of both. Whatever the split, the real figure lies between the measured one scaled by the smaller and by the larger factor; we widen that by 15% for the roughness of the table, and show it as the range.
+
Load times are CPU work (generating, lighting and meshing chunks), scaled by the CPU factor. Memory is not scaled: it follows the world and the settings, not the speed.
+
+
Estimates ignore drivers, the graphics API, thermal limits and the operating system, which can matter as much as the chip. Treat them as a rough idea, and run horizon-client --benchmark standard on your own machine for the real figure.
+
+
+ The specification table
+
{{SPEC_TABLE}}
+
+
+
+
+
How we measure
+
+
tools/perf_bench.py starts each game from a fresh world (seed 1234, creative), waits until every chunk around the spawn is loaded and drawn, then for each scene teleports, waits for the chunks, holds for 2 seconds and flies a straight 192-block path in 12 seconds. Every frame is recorded. Vanilla runs as players get it (your own client, the official launcher's Java options, Java 25), driven by a small development-only mod that only calls the game's own code; nothing of Minecraft is shipped. The full method and its caveats are in docs/benchmark.md, and every raw report is in the repository.
+
Measure your own computer with horizon-client --benchmark standard --render-distance 12: it writes a report to the benchmarks folder of the game folder.
+
+
+
Not an official Minecraft product. Not approved by or associated with Mojang or Microsoft.