A tidy multi-monitor workstation displays a web page, spreadsheet, and presentation around a compact desktop PC.
The GEEKOM A5 2027 Edition makes its strongest case through connectivity and low power consumption rather than benchmark fireworks. CNX Software’s Windows 11 Pro testing found working four-display output, fast USB transfers and 2.37 Gbps wired networking, alongside idle consumption below five watts. The limits were equally clear: modest graphics performance and effectively unusable YouTube playback at 8K and 60 frames per second.

These are CNX Software’s measurements, not WindowsForum hands-on results. The review builds on the publication’s earlier unboxing and teardown, using a manufacturer-supplied unit with 16 GB of memory and a 512 GB SSD. Its English article credits Suthinee Kerdkaew’s original review for CNX Software Thailand and discloses affiliate links.

A tidy multi-monitor workstation displays a web page, spreadsheet, and presentation around a compact desktop PC. A 2027 label, but a familiar AMD platform​

The tested configuration ran Windows 11 Pro 25H2, build 26200.9711, with two 8 GB Lexar DDR4-3200 SO-DIMMs. Networking came from Realtek RTL8125 2.5GbE and RTL8852BE Wi-Fi 6 controllers. CNX also successfully transferred a file over Bluetooth to a laptop. Those findings establish functionality on this particular Windows installation, rather than guaranteeing every accessory or application combination.

The processor deserves some context. AMD identifies the Ryzen 7 7730U as an eight-core, 16-thread Zen 3 chip launched in the first quarter of 2023, with a 2 GHz base clock, an official boost ceiling of 4.5 GHz and a 15 W default TDP. In other words, “2027 Edition” describes GEEKOM’s product, not a new CPU architecture. CNX’s reported 4.55 GHz utility reading should not be confused with AMD’s official specification.

That distinction is useful buyer context: assess this mini PC by its measured capabilities, not the year printed on the box.

Windows benchmarks: capable, not class-leading​

CNX selected Windows’ Best Performance power mode and tested at an ambient temperature of approximately 28–29°C. Those conditions matter when comparing results with a cooler room or a different power profile.

TestCNX Software result
PCMark 10 overall5,416
PCMark 10 Essentials8,992
PCMark 10 Productivity9,128
PCMark 10 Digital Content Creation5,253
PassMark PerformanceTest 11 overall4,023
3DMark Fire Strike3,044
Cinebench R23 single-core1,386
Cinebench R23 multi-core7,868
Unigine Heaven 4.0 at 1920 × 108033 FPS average

The results above are the publication’s reported measurements under those conditions.

The useful interpretation is that the A5 looks more convincing as a compact productivity desktop than as a graphics-focused machine. However, neither PCMark nor an older graphics benchmark establishes performance in a particular modern game, development project or video-editing workflow. The review does not provide those application-specific results.

Eight cores do not automatically win​

CNX compared the A5 with the GEEKOM A7 2026 Edition, A5 Pro 2026 Edition and Beelink EQi13 Pro. The A7’s six-core Ryzen 5 7545U system scored 6,503 in PCMark 10 and 9,578 in Cinebench R23 multi-core, ahead of the A5’s 5,416 and 7,868 respectively. The A5 nevertheless led both other GEEKOM systems in the table’s overall PassMark score. There is no single winner across every workload.

The Beelink comparison also needs boundaries. Its benchmarked Core i5-13500H configuration had 32 GB of memory and a 1 TB SSD, versus the A5’s 16 GB and 512 GB. CNX’s associated retail alternative used a different Core i7 processor. These results offer performance context, but they do not establish a clean, configuration-matched price comparison.

USB and storage: choose the right socket​

CNX tested all six USB ports, using an NVMe enclosure for the faster connections and a USB hard drive for the slower rear port. Five delivered approximately 929–967 MB/s reads and 818–840 MB/s writes. The bottom rear USB-A connection operated at USB 2.0 speeds, returning 43 MB/s reads and 36 MB/s writes with the test drive.

The practical takeaway is straightforward: reserve that lower rear socket for a low-bandwidth peripheral rather than an external SSD. The other five ports demonstrated substantially faster storage transfers, although the review does not establish their aggregate throughput when several drives are active together.

The internal NVMe SSD reached 3,731 MB/s sequential reads and 3,146 MB/s sequential writes in CrystalDiskMark. CNX also installed a 1 TB Toshiba hard drive in the 2.5-inch SATA bay and reported successful operation. That provides a tested expansion option without replacing the system drive.

Four displays worked—but keep the claim precise​

CNX connected four screens simultaneously through the two HDMI 2.0 outputs and two USB-C connections. The arrangement included one 32-inch 4K monitor and three portable displays; all four were detected. This is useful confirmation for a multi-screen Windows workspace. It is not evidence that four 4K displays were tested together at a particular refresh rate.

Video playback produced a similarly specific boundary:

  • 4K at 30 FPS: smooth, with no dropped frames reported.
  • 4K at 60 FPS: smooth, with no dropped frames reported.
  • 8K at 30 FPS: smooth, with no dropped frames reported.
  • 8K at 60 FPS: 1,429 dropped frames out of 1,474.

The successful playback runs lasted approximately five minutes in Firefox on YouTube. These findings should not be generalized to every codec, local media player or output resolution. Smooth playback of an 8K stream also does not, by itself, demonstrate native 8K monitor output. The sensible conclusion is narrower: 4K60 YouTube playback worked in this test; 8K60 did not.

Networking delivers a clear wired advantage​

Using 60-second iperf3 runs, CNX measured 2.37 Gbps in both directions over Ethernet. Wi-Fi 6 testing on the 5 GHz band with a Xiaomi Mi Router AX6000 produced 879 Mbps download and 864 Mbps upload. These are local-network throughput measurements, not internet speed-test results.

For readers moving large files, the wired result is the stronger finding. Nevertheless, neither measurement guarantees matching Windows file-copy speeds: the review’s iperf results do not include the complete storage and file-sharing workload.

An independently verified Bluetooth check​

CNX interpreted the adapter’s LMP 13 firmware identifier as Bluetooth 5.4. Microsoft’s support documentation confirms that mapping and provides a way to inspect it on Windows 11:

  1. Right-click Start and open Device Manager.
  2. Expand Bluetooth.
  3. Open Properties for the actual radio—such as Realtek Bluetooth Adapter, not a Microsoft Bluetooth Enumerator.
  4. Select Advanced and locate the LMP number beside the firmware information.

An LMP value beginning with 13 maps to Bluetooth Core Specification 5.4. This independently supports CNX’s identification and gives owners a practical verification method beyond a product listing.

Cooling and noise depend on the workload​

CNX recorded a maximum CPU temperature of 83.5°C during Fire Strike, without reported thermal throttling. Cinebench R23 multi-core initially peaked at 69.9°C; repeated testing produced maxima around 72°C, again without reported thermal throttling. These observations support the cooling system’s behavior in the tested workloads, not an unlimited sustained-performance claim.

Acoustics told a more nuanced story. Measured approximately five centimeters above the case, noise was 37.1–37.3 dBA at idle, 39.1–40.0 dBA during Cinebench, and as high as 47.4–48.1 dBA under heavy Fire Strike load. The room baseline was already 36.5–37.1 dBA, so the idle reading sat close to the measurement environment’s floor. These are near-case measurements, not desk-distance listening levels.

The important distinction is that CPU-only rendering remained relatively quiet in CNX’s testing, while combined CPU-and-GPU work could make the fan considerably more noticeable. “Quiet under load” needs an asterisk—and the workload’s name.

Low idle power is a genuine strength​

CNX measured consumption at the wall using the supplied power adapter:

Operating stateReported consumption
Powered off0.7–0.8 W
Idle3.9–4.8 W
Firefox YouTube playback, 4K6011.5–14.3 W
Fire Strike combat scene26.9–28.1 W
Cinebench multi-core, initial seconds36.6–37.0 W
Cinebench multi-core, approximately first minute30.1–31.8 W
Cinebench multi-core, sustained23.7–24.0 W

The setup used Wi-Fi, a keyboard receiver, a wired mouse and an HDMI-connected 4K monitor. These readings describe the mini PC’s adapter draw, not the electricity consumption of an entire workstation.

The decline from an initial peak near 37 W to roughly 24 W sustained is also important context for the CPU scores. Together with the absence of reported thermal throttling, it is consistent with a constrained sustained power budget rather than evidence that temperature alone explains performance. It does not prove that changing firmware power limits would be safe or worthwhile.

The verdict: connectivity over spectacle​

CNX Software’s testing presents the GEEKOM A5 2027 Edition as a functional Windows 11 Pro mini PC whose most persuasive qualities are low idle consumption, useful expansion, working four-screen connectivity and strong wired networking. Its benchmark results are mixed rather than exceptional, and the graphics and 8K60 playback findings set clear limits.

For a multi-monitor productivity setup, those practical strengths may matter more than a benchmark trophy. For sustained rendering, graphics-heavy applications or demanding media playback, the evidence argues for comparing the exact workload carefully. The A5’s appeal is not that it makes an older platform new again; it is that, within the boundaries actually tested, it puts that platform to useful work.

 

References

  1. GEEKOM A5 2027 Edition review – Part 2: A mid-range AMD Ryzen 7 7730U mini PC tested with Windows 11 Pro - CNX Software CNX Software 2026-11-16T12:00:46+00:00
  2. Facebook th.cnx-software.com
  3. AMD Ryzen™ 7 7730U Drivers and Downloads | Latest Version amd.com