RunsOn m8a.xlarge · EBS gp3 vs Blacksmith 4 vCPU x64
RunsOn does each build for 2.9–5.8× less; on time, Blacksmith is faster on the Docker build and neither is clearly faster on the Rust and TypeScript builds. Both 4 vCPU x64.
All three builds 3.4× cheaper on RunsOn: $0.0218 vs $0.0741Blacksmith 1.4× faster: 8m 27s vs 12m 00s
Rust build 4.5× cheaper on RunsOn: $0.0034 vs $0.0155within run-to-run spread: 1m 49s · 1m 47s
TypeScript build 5.8× cheaper on RunsOn: $0.0021 vs $0.0121within run-to-run spread: 1m 03s · 1m 19s
Docker build 2.9× cheaper on RunsOn: $0.0162 vs $0.0465Blacksmith 1.7× faster: 5m 22s vs 9m 08s
RunsOn m8a.xlarge · EBS gp3
jobs ran from Amazon.com (US)
Blacksmith 4 vCPU
jobs ran from Secured Servers (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 | Blacksmith | Gap |
|---|---|---|---|
| Cost | |||
| Cost per Rust build | $0.0034 | $0.0155 | 4.5× RunsOn |
| Cost per TypeScript build | $0.0021 | $0.0121 | 5.8× RunsOn |
| Cost per Docker build | $0.0162 | $0.0465 | 2.9× RunsOn |
| Price per minute | $0.0013 | $0.0080 | 6.1× RunsOn |
| Real builds, wall time | |||
| Rust build (spotify-player) | 1m 49s | 1m 47s | ≈ even |
| TypeScript build (opencode) | 1m 03s | 1m 19s | ≈ even |
| Docker build (PostHog) | 9m 08s | 5m 22s | 1.7× Blacksmith |
| CPU & memory | |||
| PassMark single-thread (points) | 3,809 | 4,332 | 1.1× Blacksmith |
| PassMark CPU Mark, all cores (points) | 13,481 | 14,771 | 1.1× Blacksmith |
| sysbench CPU, 1 thread (events/s) | 5,012 | 6,493 | 1.3× Blacksmith |
| 7-Zip compression (MIPS) | 38,485 | 28,802 | 1.3× RunsOn |
| Memory read bandwidth | 36.6 GiB/s | 41.7 GiB/s | 1.1× Blacksmith |
| Measurement | RunsOn | Blacksmith | Gap |
|---|---|---|---|
| Workspace disk (fio) | |||
| Random read, 4k QD32 (IOPS) | 2,994 | 728,035 | 243.2× Blacksmith |
| Random write, 4k QD32 (IOPS) | 2,991 | 218,915 | 73.2× Blacksmith |
| Random read, 4k QD1 (IOPS) | 1,825 | 71,728 | 39.3× Blacksmith |
| QD1 read latency, p99 | 807 µs | 16 µs | 50.7× Blacksmith |
| Sequential read | 401 MiB/s | 7.0 GiB/s | 18.0× Blacksmith |
| Sequential write | 401 MiB/s | 1.7 GiB/s | 4.3× Blacksmith |
| Sync write, 4k + fdatasync (IOPS) | 540 | 17,925 | 33.2× Blacksmith |
| Network, cache & Docker | |||
| Download from GitHub | 159 MiB/s | 76 MiB/s | 2.1× RunsOn |
| Cache restore, 4 GB (actions/cache) | 16s | 11s | 1.5× Blacksmith |
| Cache save, 4 GB (actions/cache) | 16s | 7s | 2.5× Blacksmith |
| docker pull | 4.8s | 7.9s | 1.6× RunsOn |
| Database | |||
| pgbench read/write (TPS) | 3,256 | 12,408 | 3.8× Blacksmith |
| pgbench read-only (TPS) | 142,376 | 109,979 | 1.3× RunsOn |
| Queue | |||
| Queue under burst load | 17s | 10s | 1.7× Blacksmith |
- ≈ 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); 9 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