RunsOn m8id.xlarge · local NVMe vs GitHub 4-core x64
RunsOn does each build for 6.1–8.5× less; on time, RunsOn is faster on the Rust and TypeScript builds and neither is clearly faster on the Docker build. Both 4 vCPU x64.
All three builds 7.8× cheaper on RunsOn: $0.0316 vs $0.2461RunsOn 1.3× faster: 13m 19s vs 17m 52s
Rust build 6.1× cheaper on RunsOn: $0.0074 vs $0.0457RunsOn 1.3× faster: 2m 44s vs 3m 38s
TypeScript build 7.3× cheaper on RunsOn: $0.0035 vs $0.0258RunsOn 1.3× faster: 1m 20s vs 1m 44s
Docker build 8.5× cheaper on RunsOn: $0.0206 vs $0.1746within run-to-run spread: 9m 15s · 12m 30s
RunsOn m8id.xlarge · local NVMe
jobs ran from Amazon.com (US)
GitHub 4-core
jobs ran from Microsoft (US)
Side by side
Every measurement both runners share, cost first. The dot marks the better value; gap is the larger over the smaller.
| Measurement | RunsOn | GitHub | Gap |
|---|---|---|---|
| Cost | |||
| Cost per Rust build | $0.0074 | $0.0457 | 6.1× RunsOn |
| Cost per TypeScript build | $0.0035 | $0.0258 | 7.3× RunsOn |
| Cost per Docker build | $0.0206 | $0.1746 | 8.5× RunsOn |
| Price per minute | $0.0016 | $0.0120 | 7.3× RunsOn |
| Real builds, wall time | |||
| Rust build (spotify-player) | 2m 44s | 3m 38s | 1.3× RunsOn |
| TypeScript build (opencode) | 1m 20s | 1m 44s | 1.3× RunsOn |
| Docker build (PostHog) | 9m 15s | 12m 30s | ≈ even |
| CPU & memory | |||
| PassMark single-thread (points) | 3,264 | 2,895 | 1.1× RunsOn |
| PassMark CPU Mark, all cores (points) | 8,033 | 7,511 | 1.1× RunsOn |
| sysbench CPU, 1 thread (events/s) | 3,079 | 4,076 | 1.3× GitHub |
| 7-Zip compression (MIPS) | 24,076 | 18,743 | 1.3× RunsOn |
| Memory read bandwidth | 21.7 GiB/s | 24.1 GiB/s | 1.1× GitHub |
| Measurement | RunsOn | GitHub | Gap |
|---|---|---|---|
| Workspace disk (fio) | |||
| Random read, 4k QD32 (IOPS) | 67,090 | 19,575 | 3.4× RunsOn |
| Random write, 4k QD32 (IOPS) | 33,526 | 11,216 | 3.0× RunsOn |
| Random read, 4k QD1 (IOPS) | 10,866 | 7,286 | 1.5× RunsOn |
| QD1 read latency, p99 | 249 µs | 241 µs | ≈ even |
| Sequential read | 451 MiB/s | 394 MiB/s | 1.1× RunsOn |
| Sequential write | 216 MiB/s | 244 MiB/s | 1.1× GitHub |
| Sync write, 4k + fdatasync (IOPS) | 7,903 | 4,163 | 1.9× RunsOn |
| Network, cache & Docker | |||
| Download from GitHub | 187 MiB/s | 95 MiB/s | 2.0× RunsOn |
| Cache restore, 4 GB (actions/cache) | 30s | 41s | 1.4× RunsOn |
| Cache save, 4 GB (actions/cache) | 24s | 34s | 1.4× RunsOn |
| docker pull | 5.7s | 16.4s | 2.9× RunsOn |
| Database | |||
| pgbench read/write (TPS) | 7,066 | 4,414 | 1.6× RunsOn |
| pgbench read-only (TPS) | 60,641 | 40,172 | 1.5× RunsOn |
| Queue | |||
| Queue under burst load | 19s | 3s | 6.3× GitHub |
- ≈ even: under 5% apart, or a build inside both runners' run-to-run spread
- Queue: identical jobs queued at the same instant (burst suite)
- RunsOn at EC2 spot, its default (7-day average of the cheapest us-east-1 zone's spot price, fetched 16 Nov 2026); 5 of its jobs fell back to on-demand (AWS had no spot capacity for the type) and are priced at the spot rate over the same billed time
- Cost: a managed SaaS runner against RunsOn self-hosted on your AWS account; the effort of running it was not measured