Observed results
| Metric | Accelerated path | Origin path |
|---|---|---|
| Probe attempts / reported errors | 5,432 / 6 | 5,439 / 24 |
| Checks with no reported error | 99.89% | 99.56% |
| Time to First Screen, P50 | 1.50 s | 4.83 s |
| Time to First Screen, P95 | 3.86 s | 14.82 s |
| First-screen runs over 10 s | 31 / 5,398 (0.57%) | 670 / 5,389 (12.43%) |
| DNS lookup time, P50 | 12 ms | 48 ms |
| TCP connect time, P50 | 47 ms | 211 ms |
| SSL/TLS handshake time, P50 | 59 ms | 296 ms |
| First-byte time, P50 | 796 ms | 826 ms |
The Accelerated path reduced the reported Time to First Screen by 3.33 seconds at the median and 10.96 seconds at P95. P50 is the median; P95 is the value at or below which 95% of the included first-screen measurements fall.
The Accelerated path had a lower median on every monitored date: its daily P50 ranged from 1.47 to 1.58 seconds, compared with 4.41 to 5.31 seconds for the Origin path.
Time to First Screen over the monitoring period
What we compared
The test site was hosted outside mainland China. The Accelerated path routed requests through CN-Infra Hub; the Origin path used the site's origin domain, which may itself use other delivery infrastructure.
During multi-day monitoring, the median transferred size was approximately 1.85 MB on each path, with a difference of less than 0.1%.
Faster across all three major carriers
Across the monitored locations in major Chinese cities, the Accelerated path recorded a lower median Time to First Screen on each of the three major carrier networks.
| Carrier | Accelerated path, P50 | Origin path, P50 | Reduction |
|---|---|---|---|
| China Telecom | 1.36 s | 5.02 s | 73% |
| China Unicom | 1.53 s | 3.96 s | 61% |
| China Mobile | 1.69 s | 5.51 s | 69% |
Where the time was saved
Median DNS lookup, TCP connection, and TLS handshake times were lower on the Accelerated path. Median first-byte times were close—796 ms on the Accelerated path and 826 ms on the Origin path—while median Time to First Screen was 1.50 and 4.83 seconds, respectively.
How the comparison was measured
Two synthetic monitoring tasks ran from 15 to 21 October 2026. Both covered major Chinese cities across China Telecom, China Unicom, and China Mobile. The tasks used the same monitoring window, with measurements collected separately for each path.
Time to First Screen measures the interval from entering the URL until the rendered page reaches the configured first-screen height, 600 pixels by default. For a shorter page, measurement ends at document complete. This is a synthetic monitoring metric; the browser test below reports FCP and LCP separately.
All timing percentiles and the share over 10 seconds use successful measurements with a positive recorded first-screen time.
A browser view of the page
In a browser test from Shanghai, First Contentful Paint (FCP) and Largest Contentful Paint (LCP) were both 1.252 seconds on the Accelerated path and 2.311 seconds on the Origin path. Both measurements used the same desktop browser and network settings.
| Displayed metric | Accelerated path | Origin path |
|---|---|---|
| First Contentful Paint | 1.252 s | 2.311 s |
| Largest Contentful Paint | 1.252 s | 2.311 s |
| Page weight / requests | 636 KB / 43 | 649 KB / 47 |
Accelerated path
Origin path
A multi-node HTTP snapshot
In simultaneous HTTP response tests, the Accelerated path averaged 0.499 seconds across 314 participating nodes; the Origin path averaged 2.988 seconds across 317 nodes. The table shows average response times by carrier.
| Node group | Accelerated path average | Origin path average |
|---|---|---|
| All participating nodes | 0.499 s (314 nodes) | 2.988 s (317 nodes) |
| China Telecom | 0.354 s | 1.850 s |
| China Unicom | 0.343 s | 6.619 s |
| China Mobile | 0.807 s | 2.445 s |
Origin path · 317 nodes · 2.988 s average
Accelerated path · 314 nodes · 0.499 s average
What these results mean
For this test site, the Accelerated path reduced the median Time to First Screen from 4.83 to 1.50 seconds. The P95 result fell from 14.82 to 3.86 seconds, indicating a shorter wait for slower measurements as well as a faster typical result. These figures describe the monitored site and locations during the test period.
The Accelerated path also recorded a lower median across all three major carrier networks in the monitored cities. This matters when an audience reaches a site through different providers: an overall average alone can hide differences between networks.
The browser test provides an additional view of when content appeared on screen, while the HTTP response test shows a snapshot across multiple network nodes. Together, they help inform an evaluation of page experience and access performance. Results for another site will depend on its content, cache configuration, origin, and user locations.
To evaluate your own site, we can establish a baseline from the locations and carrier networks that matter to your users, compare the two paths, and review the results with your team.