Skip to content
Superluminal

Superluminal vs GarageFarm

By Superluminal Engineering

GarageFarm came second overall, and Superluminal still had the shorter render pipeline in every scene. Classroom: 90.5 against 163.6 seconds.

Published 2026-08-04 · Updated 2026-08-08

Across all four scenes, Superluminal put the complete output on disk first and recorded the shorter render pipeline.

Superluminal and GarageFarm in the benchmark ranking

Both services are selected. The other six farms stay visible but muted for context.

Superluminal and GarageFarm selected on the 2026 render farm score scatterplot
undefined Download chart PNG

GarageFarm ranked second overall at 126.2, with a Speed Score of 138.2 and a Cost Score of 102.2. The benchmark method scores render-pipeline time and price across the completed scenes.

Render pipeline by scene

SceneFastest Superluminal render pipelineGarageFarm render pipelineGarageFarm charge
Classroom90.5 s163.6 s$7.55
Junk Shop64.5 s127.3 s$3.92
Monster Under the Bed59.9 s88.6 s$2.73
Fox (EEVEE)66.5 s, 10 nodes114.0 s$1.12

Superluminal recorded workflow

Superluminal submission terminal confirming the queued job after packaging and upload
Submission ends in the add-on's terminal: the job is packaged, uploaded, and queued before any browser opens. Open image
Superluminal Classroom job starting active rendering on 30 nodes
Active rendering started after the upload and queue phases in the same recording. Open image
Superluminal Classroom job after all 145 frames completed
Provider completion showed all 145 Classroom frames finished. Open image
Superluminal Classroom job after all 145 frames reached local disk
The final boundary confirms the complete local frame set 106.7 seconds after upload began. Open image

GarageFarm recorded workflow

GarageFarm WebManager job settings for Classroom
The operator configures and starts the uploaded job in WebManager. Open image
GarageFarm job page showing active nodes
WebManager shows active nodes while work runs. Open image
Completed GarageFarm Classroom job with cost
The completed page reports progress and dollar spend. Open image
GarageFarm renderBeamer delivering completed output
renderBeamer continues retrieving originals after provider completion. Open image
GarageFarm completed benchmark jobs with average time per frame for each worker group
Done Frames reports average frame time, worker time, node type, and cost for each group. Open image
GarageFarm process results separating pre-render time from render duration for a five-frame worker group
Process Results separates seven seconds of pre-render work from two minutes and 33 seconds of rendering. Open image

One upload, three programs

GarageFarm's desktop client compresses each package before transfer, for an effective 8.2 MB/s. That is compression, not raw network speed.

Submission and monitoring live in different places. The client uploads, then hands off to a separate web manager. On Junk Shop the manager's Done status lagged behind the last completed frame, so a finished job still read as unfinished.

The background downloader also fell behind the farm. Classroom frames were still arriving while the next scene rendered. That is where the 16.2-to-301.3-second retrieval tails come from. The client's test-job mode also produces an auxiliary job, which forced a full-job resubmit of Fox (EEVEE).

Superluminal keeps the loop in one tool. The add-on uploads, streams finished frames back while later frames render, and the web job page mirrors everything live. Nothing to hand off, no second program to check.

Capacity and billing

Each job ran at High priority. GarageFarm advertised a 60-node cap. The interface showed a peak of 29 active nodes.

GarageFarm charges a prepaid balance at $0.012 per OctaneBench-hour and shows the dollar charge after the job ends. You learn the price when it is over.

Superluminal shows the rate before you submit and the accruing charge while frames finish. Each job settles as exact dollars, at the published rates.

GarageFarm completed every scene without a defect and still finished behind in every one. That is the engineering gap when hardware is not the problem.

Sources

  1. Render farm benchmark dataset, July to August 2026