This is a synthetic lab comparison, not a production benchmark or installer. The same isolated Nginx front was used for direct and cached traffic. The origin was a tiny PHP JSON response with public max-age=120; the Varnish policy capped lifetime at 60 seconds. No artificial origin delay was added. Requests used loopback HTTP with a new TCP connection; DNS, TLS, browser rendering and page assets were excluded. There were 80 timed requests per case, three rounds, at concurrency 1 and 4. Phase order alternated between rounds. Five untimed warm-up requests preceded each batch. Cold samples used 80 distinct URLs; warm samples repeatedly requested a prefilled URL. RPS is achieved batch throughput from this small client, not maximum server capacity. Error count is non-200 or invalid fixture response. Origin counts derive from the fixture's locked request counter and observed DIRECT/MISS/HIT outcomes. Warm-up and initial fill are excluded from timed origin counts. All 1,440 timed requests succeeded. Settings were preserved, 29 existing vhost files were unchanged during the run, and the fixture was stopped afterward. The old Varnish engine needs a supported security-maintained update before public live-site beta testing. No real WordPress or Laravel application was benchmarked.