CPU hits 90%+ with 4 phones on Appium device farm - workarounds for Intel x86-64 and ARM?

[amashasaraph]

Hi,

We are running an Appium device farm and tested the same setup across two different machines:

• Machine 1: Intel x86-64 (MSI laptop) - CPU hits 90%+ with just 3 phones connected
• Machine 2: ARM Cortex-A76 (Raspberry Pi 5) - CPU hits 90%+ at 4 phones, but gets there much more slowly

So the RPi handles it better, but both machines eventually struggle when we push to 4 devices in parallel. We are seeing lag and test instability once CPU crosses 90%.

Our setup:

• Appium 2.x
• Node.js [your version]
• 4 Android devices connected via USB
• Tests running in parallel across all devices

Any advice on how to scale beyond 4 devices without hitting this CPU ceiling would be really appreciated. Thanks!

[saikrishna321]

@amashasaraph I have a Mac Pro 16 Core and i'm able to run 8 devices without any issues.

[amashasaraph]

In our case, it seems to be working fine on the Mac Mini as well. However, the limitation is more noticeable when running on the Raspberry Pi.

Do you happen to know if there’s any workaround or optimization that could help improve performance on the Raspberry Pi?

[amashasaraph]

We only need the capability to remotely control mobile devices, and we were not running tests using Appium device farm.

[saikrishna321]

@amashasaraph

Why scaling fails on Raspberry Pi

1. Hardware limitations

Raspberry Pi has limited CPU cores and RAM
Running multiple Appium sessions = heavy load (Node.js + drivers + device communication)
You quickly hit bottlenecks → tests slow down or fail

2. I/O and USB bandwidth constraints

Device farms rely on USB-connected devices or emulators
Pi’s USB controller can’t handle many devices simultaneously with stable throughput

3. Network & process overhead

Appium Device Farm manages sessions, routing, logs, etc.
Each session adds overhead → Pi struggles to manage concurrent sessions

4. Not designed for distributed scaling

Proper scaling usually means:
multiple machines (cluster/grid)
load balancing
Raspberry Pi is typically used as a single-node setup, not a scalable grid controller