Trust layer for browser GPU benchmarks
Trusted live GPU benchmark results are now live.
Volume Shader BM now connects public result pages, fresher leaderboards, anonymous analytics, stricter sample quality filters, and visible dataset status into one verifiable benchmark workflow.
Positioning
From test toy to benchmark system
Promise
Every claim gets a clearer evidence path
Dataset
Live status now sits on the homepage
What shipped
Five trust features, released as one product moment
MVP 2.5 is not a cosmetic update. It adds the public surfaces and internal feedback loop needed for a browser GPU benchmark to be believed, shared, and improved.
Public result pages
Finished benchmark runs can resolve to durable pages that are easier to revisit, compare, cite, and share.
Fresh live leaderboards
Ranking pages now render from the current production benchmark pool instead of feeling like a stale snapshot.
Anonymous funnel analytics
The product can see where benchmark flows succeed or fail without turning the test into an identity product.
Stricter sample quality
Incomplete, noisy, and misleading records are filtered more aggressively before they influence public rankings.
Visible dataset status
Homepage counters now make the live benchmark corpus easier to evaluate before trusting a claim.
Before / after
The benchmark now explains why users should trust it
More rows are useful only when the product explains what is fresh, what is eligible, and what can be verified. This release makes those trust signals visible.
Result sharing
Before MVP 2.5
Single-run score screens were hard to cite
Now
Public result URLs create a stable evidence trail
Leaderboard trust
Before MVP 2.5
Rankings could feel detached from live submissions
Now
Server-rendered rankings stay closer to current data
Product learning
Before MVP 2.5
Growth work depended too much on guesswork
Now
Anonymous events show where the funnel needs work
Data quality
Before MVP 2.5
More rows did not always mean better rankings
Now
Eligibility filters make comparisons more defensible
| Surface | Before MVP 2.5 | Now |
|---|---|---|
| Result sharing | Single-run score screens were hard to cite | Public result URLs create a stable evidence trail |
| Leaderboard trust | Rankings could feel detached from live submissions | Server-rendered rankings stay closer to current data |
| Product learning | Growth work depended too much on guesswork | Anonymous events show where the funnel needs work |
| Data quality | More rows did not always mean better rankings | Eligibility filters make comparisons more defensible |
How it works
A cleaner path from browser run to public benchmark record
The new workflow makes the journey easier to understand: run the test, qualify the sample, publish the result, then use anonymous product signals to improve the next run.
Run
A user completes a browser GPU test with WebGPU or WebGL context.
Qualify
The sample passes completeness and ranking eligibility checks.
Publish
The result page and leaderboard can expose the record in a clearer public context.
Learn
Anonymous events reveal where the product flow should improve next.
Bottom line
Volume Shader BM is no longer just a browser GPU test page. It is now a live benchmark product with shareable evidence, fresher rankings, and visible data quality signals.
Keep reading
Read the 32,604-submission browser GPU reportThe fixed report remains the editorial snapshot. This launch adds the live trust layer on top of that dataset story.
FAQ
Quick answers for benchmark users and AI search
These answers summarize the update in short, extractable blocks while keeping the details visible to human readers.
Is this only a leaderboard redesign?
No. The release combines public result pages, live leaderboard rendering, analytics persistence, quality filters, and dataset status. The visual layer is only the surface of a broader trust update.
Why do the report count and live count differ?
The 32,604-submission report is a fixed editorial snapshot. The live homepage status reflects the current production benchmark pool, so it changes as new runs arrive and filters improve.
Does analytics collect personal identities?
No. The instrumentation is focused on anonymous benchmark funnel events: starts, completions, failures, and routes that help improve reliability.