quickstart →

r9g.large

The 2 vCPU tmpfs runner on arm64: Graviton5 with 8 GB per vCPU, /tmp, /home/runner and /var/lib/docker in memory (extras=tmpfs): the arm64 counterpart of runson-r8a-large-tmpfs.

available for self-hosted runners on RunsOn
single-thread · passmark
2,469pts
Neoverse-V3 (arm64)
one job · hardware suite · 14 Oct
1 run, ranked from 3
vcpus
2
2 cores
memory
16 GiB
ram
clock speed
3.3 GHz
max frequency
architecture
arm64
Graviton

Measured

1 job of 1 finished in the hardware suite, 14 Oct: the values are that job's, so one run decides them. Every job links its log on the open harness.

CPU Mark
2,840
PassMark, both vCPUs
cores of work
2.00
sysbench, 2 threads ÷ 1
sysbench 1 thread
3,766
events per second
7-zip
20,371
MIPS, all threads
datesuitesingle-threadcapacityjob log
14 Oct 2026 hardware suite 2,469 spot job log ↗
  • Cross-check, daily PassMark: 1,234 single-thread over 1 run (PassMark only, every day on RunsOn's 2 vCPU runners).
  • Also a runner in RunsOn's runner benchmarks: see it as a CI runner →

EC2 instance family: r9g

What the family is built for, the silicon underneath, and how to ask RunsOn for it.

r9g memory
cpu type
Neoverse-V3 (Graviton5)
architecture
arm64
notes
The 2 vCPU tmpfs runner on arm64: Graviton5 with 8 GB per vCPU, /tmp, /home/runner and /var/lib/docker in memory (extras=tmpfs): the arm64 counterpart of runson-r8a-large-tmpfs.
.github/workflows/ci.yml yaml
jobs:
  build:
    runs-on:
      - runs-on=${{ github.run_id }}
      - family=r9g
      - cpu=2
the label is a query — each line one constraint, resolved at launch. job labels →

Availability & price by region

Spot is the 7-day average of the cheapest zone's spot price in each region, from the RunsOn finder: an estimate of an hour in your own AWS account, where RunsOn adds no markup. The bar shows how far it runs under on-demand; the badge is AWS's spot-interruption band for the type.

US
N. Virginia
us-east-1
spot · 7-day average $0.0425/hr
on-demand$0.1284
savings vs on-demand67%
spot interrupt freq n/a
US
Oregon
us-west-2
cheapest spot
spot · 7-day average $0.0232/hr
on-demand$0.1284
savings vs on-demand82%
spot interrupt freq n/a
IE
Ireland
eu-west-1
spot · 7-day average —/hr
on-demand—
savings vs on-demand—
spot interrupt freq n/a
DE
Frankfurt
eu-central-1
spot · 7-day average $0.0705/hr
on-demand$0.1549
savings vs on-demand55%
spot interrupt freq n/a

Spot is the cheapest availability zone's spot price at each of the RunsOn finder's fetches (every 90 minutes), averaged over 7 days; on-demand is the Linux list price. Savings compare the two. Prices fetched 14 Oct 2026. The interruption band is AWS Spot Advisor's, fetched 14 Oct 2026; it lists none for us-east-1, us-west-2, eu-west-1 and eu-central-1 (“n/a”). No finder price for eu-west-1: “—”.

Other sizes measured

Larger r9g sizes the harness measured in RunsOn's runner benchmarks, each from the suite with the most runs.

typevCPUmemorysingle-threadCPU Markrunsspot $/hr · us-east-1
r9g.xlarge 4 32 GiB 2,474 5,624 2 runs · hardware suite $0.1071

About these benchmarks

The open benchmark harness launches r9g.large as a RunsOn runner and runs PassMark PerformanceTest (CPU Single Threaded and CPU Mark), sysbench on one thread and on every thread, and 7-Zip's built-in benchmark. These numbers come from the hardware suite, the suite that measured this type in the most runs; each is the value of its one job of the current workload definition in the last 30 days.

A type takes a place once it has 3 runs. A type within 5% of the type heading its group shares that place (“joint #2”); a lead under 5% is no lead.

Prices come from the RunsOn finder for the four regions shown; savings compare its 7-day average spot price with the on-demand price. RunsOn launches this type the same way for any job: ephemeral EC2 in your account, one CloudFormation stack. More in the methodology.

metricpassmark · single-thread
suitehardware suite
runs1 run · 1/1 jobs
measured14 Oct
regions4 · us-east-1, us-west-2, eu-west-1, eu-central-1
prices14 Oct 2026

Run r9g.large in your own account.

The benchmark runs on the same RunsOn anyone can deploy. One CloudFormation stack, your AWS account, ~10 minutes.