- Published on
Radxa X4L: A Pocket-Sized Powerhouse
- Authors
- Name
- Amrut Prabhu
- @amrutprabhu42
Table of Contents
Radxa has recently launched the X4L, the successor to their popular X4. Marketed as a portable PC that fits in your pocket, the X4L offers a suite of features aimed at enhancing functionality and performance for tech enthusiasts and professionals alike. Here’s an in-depth look at what this device provides, its benchmarks, and potential use cases.
Links to buy Radxa X4L:
- AliExpress : https://s.click.aliexpress.com/e/_DESXdqh
- Arace : https://arace.tech/products/palmshell-slim-x4l
Key Features and Specifications
The X4L is noticeably larger than its predecessor, being roughly twice the size of the X4 and weighing around 289 grams. Both models are powered by the Intel N100 processor, a quad-core x86 chip capable of clock speeds up to 3.4 GHz, making them compatible with a variety of operating systems, including Windows and Linux.
Improved Hardware:
-
Dual 32-bit memory bus: The X4L now utilizes dual 32-bit channels to fully leverage its LPDDR5 RAM, a significant improvement over the X4’s single-channel design.
-
Enhanced cooling: A temperature-controlled CPU fan has been added, with adjustable settings in the BIOS for optimized performance.
-
PCI Express 3.0 support: Confirmed through testing, this allows faster NVMe and peripheral performance.
Recovery and Installation: The Power of Roobi
The X4L comes pre-installed with Roobi, a recovery installer developed by Radxa. This network installer simplifies the setup process by eliminating the need for bootable disks.
- How it works: Connect a LAN cable, press the recovery button, and access the installer at
roobi.localfrom any device. - Supported OS options: OpenWRT, Ubuntu, Windows 10, and Fedora can be installed remotely.
- Out-of-the-box experience: Testing with Windows 10 revealed all drivers were pre-installed, streamlining the setup process.
Performance Testing
Memory and Storage:
-
Memory architecture: The X4L’s dual 32-bit channels boost memory performance by up to 35% for memory copy operations compared to the X4.
-
NVMe performance:
- Windows: 3400 MB/s (read speed).
- Ubuntu: 2400 MB/s (read speed).
Power Consumption
- At Idle, the X4L sips about 10 Watts of energy. While I was running a stress test, the power consumption went up to 21 Watts of energy.
Ethernet Speeds:
- The 2.5 Gigabit Ethernet port achieved speeds of up to 2300 MB/s during iPerf3 testing.
Stress Testing:
- Windows: Under load, temperatures reached 90°C, triggering minor CPU throttling. Prolonged stress pushed it to 100°C, primarily affecting the bottom metal plate.
- Ubuntu: Performed better, with temperatures peaking at 95°C during prolonged stress without throttling.
Benchmark Results:
-
Geekbench: Scored ~10% better than the X4 on Windows; no significant difference on Ubuntu.
-
Tiny MemBench: The X4L showed improvements across the board compared to the X4:
- C copy: +30%
- C fill: +16%
- Standard memcpy: +23%
- Standard memset: +21%
Tiny Membench Score Details (Expand)
tinymembench v0.4.9 (simple benchmark for memory throughput and latency) ========================================================================== == Memory bandwidth tests == == == == Note 1: 1MB = 1000000 bytes == == Note 2: Results for 'copy' tests show how many bytes can be == == copied per second (adding together read and writen == == bytes would have provided twice higher numbers) == == Note 3: 2-pass copy means that we are using a small temporary buffer == == to first fetch data into it, and only then write it to the == == destination (source -> L1 cache, L1 cache -> destination) == == Note 4: If sample standard deviation exceeds 0.1%, it is shown in == == brackets == ========================================================================== C copy backwards : 7110.9 MB/s (6.1%) C copy backwards (32 byte blocks) : 7100.4 MB/s (2.4%) C copy backwards (64 byte blocks) : 7128.2 MB/s C copy : 6816.3 MB/s (0.8%) C copy prefetched (32 bytes step) : 4136.5 MB/s (2.2%) C copy prefetched (64 bytes step) : 4280.4 MB/s C 2-pass copy : 6055.5 MB/s C 2-pass copy prefetched (32 bytes step) : 3142.8 MB/s C 2-pass copy prefetched (64 bytes step) : 3141.7 MB/s (0.1%) C fill : 8894.8 MB/s C fill (shuffle within 16 byte blocks) : 8928.3 MB/s (0.2%) C fill (shuffle within 32 byte blocks) : 8919.1 MB/s (3.2%) C fill (shuffle within 64 byte blocks) : 8906.1 MB/s (0.2%) --- standard memcpy : 9620.9 MB/s standard memset : 9253.6 MB/s (0.3%) --- MOVSB copy : 7101.1 MB/s MOVSD copy : 7097.0 MB/s SSE2 copy : 7100.3 MB/s SSE2 nontemporal copy : 9798.7 MB/s SSE2 copy prefetched (32 bytes step) : 5836.6 MB/s SSE2 copy prefetched (64 bytes step) : 6093.5 MB/s SSE2 nontemporal copy prefetched (32 bytes step) : 6809.2 MB/s SSE2 nontemporal copy prefetched (64 bytes step) : 7234.2 MB/s (0.2%) SSE2 2-pass copy : 5725.8 MB/s (0.1%) SSE2 2-pass copy prefetched (32 bytes step) : 4133.2 MB/s (0.2%) SSE2 2-pass copy prefetched (64 bytes step) : 4267.9 MB/s SSE2 2-pass nontemporal copy : 2868.6 MB/s SSE2 fill : 9249.8 MB/s (0.3%) SSE2 nontemporal fill : 19631.3 MB/s (0.3%) ========================================================================== == Memory latency test == == == == Average time is measured for random memory accesses in the buffers == == of different sizes. The larger is the buffer, the more significant == == are relative contributions of TLB, L1/L2 cache misses and SDRAM == == accesses. For extremely large buffer sizes we are expecting to see == == page table walk with several requests to SDRAM for almost every == == memory access (though 64MiB is not nearly large enough to experience == == this effect to its fullest). == == == == Note 1: All the numbers are representing extra time, which needs to == == be added to L1 cache latency. The cycle timings for L1 cache == == latency can be usually found in the processor documentation. == == Note 2: Dual random read means that we are simultaneously performing == == two independent memory accesses at a time. In the case if == == the memory subsystem can't handle multiple outstanding == == requests, dual random read has the same timings as two == == single reads performed one after another. == ========================================================================== block size : single random read / dual random read, [MADV_NOHUGEPAGE] 1024 : 0.0 ns / 0.0 ns 2048 : 0.0 ns / 0.0 ns 4096 : 0.0 ns / 0.0 ns 8192 : 0.0 ns / 0.0 ns 16384 : 0.0 ns / 0.0 ns 32768 : 0.0 ns / 0.0 ns 65536 : 2.5 ns / 3.7 ns 131072 : 3.8 ns / 4.7 ns 262144 : 5.1 ns / 6.0 ns 524288 : 6.4 ns / 7.2 ns 1048576 : 7.1 ns / 7.5 ns 2097152 : 7.9 ns / 8.2 ns 4194304 : 14.0 ns / 17.3 ns 8388608 : 35.5 ns / 53.1 ns 16777216 : 95.7 ns / 138.6 ns 33554432 : 132.3 ns / 173.5 ns 67108864 : 152.0 ns / 187.9 ns block size : single random read / dual random read, [MADV_HUGEPAGE] 1024 : 0.0 ns / 0.0 ns 2048 : 0.0 ns / 0.0 ns 4096 : 0.0 ns / 0.0 ns 8192 : 0.0 ns / 0.0 ns 16384 : 0.0 ns / 0.0 ns 32768 : 0.0 ns / 0.0 ns 65536 : 2.5 ns / 3.7 ns 131072 : 3.8 ns / 4.7 ns 262144 : 4.4 ns / 4.9 ns 524288 : 4.7 ns / 5.0 ns 1048576 : 4.9 ns / 5.0 ns 2097152 : 5.3 ns / 5.3 ns 4194304 : 11.3 ns / 14.4 ns 8388608 : 30.7 ns / 46.5 ns 16777216 : 87.4 ns / 128.3 ns 33554432 : 121.6 ns / 156.4 ns 67108864 : 137.7 ns / 165.2 ns
-
Whisper Speech To Text Conversion
Practical Use Cases
The X4L excels in versatility, making it suitable for a range of applications:
- Home Automation: Running Home Assistant with local voice assistants like Whisper and Piper demonstrated noticeable performance improvements over the X4 and Raspberry Pi 5.
- Portable PC: Ideal for running Windows or Linux as a compact workstation.
- Virtualization: Supports Proxmox for managing virtual machines.
- Network Storage: Use OpenMediaVault for creating a network-attached storage (NAS) solution.
Hardware Expansion
The board features GPIO pin holes, and Radxa may release an optional RP2040 module for GPIO support, enabling broader functionality in IoT and automation projects.
Final Thoughts and Benchmarks
The Radxa X4L builds on its predecessor’s strengths with improved memory bandwidth, enhanced thermal performance, and user-friendly recovery tools.
If you’re interested in this device, drop your suggestions for tests you’d like to see in the video comments section. Don’t forget to like, subscribe, and support my YouTube channel through Patreon or by buying me a coffee.
Stay tuned for more content on home automation and tech reviews.