A Principled Technologies report: Hands-on testing. Real-world results.

A woman in an office working in front of a monitor

Do more with the HP Z2 Mini G1a

Compared to both the Apple Mac mini and the Apple Mac Studio, this AMD Ryzen AI Max+ PRO processor-based desktop system delivered stronger performance across a series of benchmark tests

PC users in today’s digital environment need hardware that lets them multitask effectively and run demanding workloads such as AI without interruption. Teams that prefer powerful desktop PCs in a small form factor may be torn between Windows PCs such as the HP Z2 Mini G1a Workstation with AMD Ryzen AI Max+ PRO 395 processor and Mac devices such as the Apple® Mac mini and the Mac Studio, but lack good information on how they compare in terms of performance.

To help buyers make a more informed decision, we compared those three systems using a range of industry-standard benchmarks to assess their general as well as AI performance. Across the board, we found that the HP Z2 Mini G1a workstation powered by an AMD Ryzen AI Max PRO processor delivered superior benchmark scores, indicating that it could boost productivity for everyday tasks and help maximize the impact of AI compared to the Apple systems we tested.

Boost productivity with up to 64% stronger general performance. Maximize the impact of AI with up to 3.2 times the AI performance.

To assess general productivity, graphics, on-device AI, and creative workload performance, we ran the following benchmarks:

  • PassMark PerformanceTest 11.1
  • Cinebench R23
  • GeekBench AI CPU and GeekBench AI GPU
  • Chaos V-Ray CPU

HP provided the systems. The results we report reflect the specific configurations we tested. Any difference in the configurations you test, as well as network traffic or software additions, can affect these results. For a deeper dive into our testing parameters, and more details about the systems we tested, see the science behind the report.

What we learned: Performance & productivity tests

To assess the general performance of the systems, we used two benchmarks: PassMark PerformanceTest 11.1 and Cinebench R23. Both of these tests use strenuous tasks related to graphics and rendering, which can indicate performance users with demanding workloads might see.

PassMark PerformanceTest 11.1 benchmark suite stresses CPU, 2D and 3D graphics, storage, and memory. The CPU test executes complex mathematical calculations involving compression, encryption, and physics simulations,4 and the score reflects raw processing power.

Cinebench R23 tests how fast a CPU can render a 3D scene.5 The Cinebench multi-core score indicates how a processor might handle demanding, multi-threaded tasks in everyday use in addition to rendering capability.

Higher is better. Chart title Up to 54% better processor performance. Bar chart showing PassMark PerformanceTest 11.1 — CPU Mark score (median) results. Bars compare three systems. HP Z2 Mini G1a with AMD Ryzen AI Max+ PRO 395 scored 60,877; Apple Mac Studio with Apple M2 Ultra scored 53,610; Apple Mac mini with Apple M4 Pro scored 39,403.
PassMark PerformanceTest 11.1 CPU Mark score. Source: PT
Higher is better. Chart title Up to 64% better CPU Multi Core performance.Bar chart showing Cinebench R23 — CPU Multi Core (median). Bars compare three systems: HP Z2 Mini G1a with AMD Ryzen AI Max+ PRO 395 scored 38,663; Apple Mac Studio with Apple M2 Ultra scored 29,014; Apple Mac mini with Apple M4 Pro scored 23,495.
Cinebench R23 CPU Multi-Core score. Source: PT

What we learned: AI tests

These days, desktop users aren’t only doing basic productivity or even strenuous creative work—they’re also exploring how to use AI to accelerate their work. That’s why it’s worthwhile to invest in systems that also leverage powerful GPUs and NPUs in addition to CPUs. Better AI capabilities could pay dividends in productivity boosts.

We used multiple tests from the Geekbench AI benchmark suite to test the systems. Geekbench AI evaluates CPU, GPU, and NPU performance using real-world ML tasks.6 We used both the Open Neural Network Exchange (ONNX) and the Intel OpenVINO inference engines in our testing of the Windows-based HP Z2 Mini G1a, with the DirectML GPU backend for both engines. For the Apple systems, we used the Core ML inference engine with the GPU backend. We ran tests in many formats and present single-precision (FP32) and half-precision (FP16) formats and a quantized score. FP32 results are more relevant for workloads that demand accuracy—such as medical imaging or advanced deep learning—while FP16 results are useful for scenarios where speed and efficiency take priority, such as training smaller datasets or models.

Two bar charts showing Geekbench AI CPU results (Intel OpenVINO) at two precision levels. Higher is better. Bars compare three systems. For FP32 single precision level: HP Z2 Mini G1a with AMD Ryzen AI Max+ PRO 395 scored 7,114; Apple Mac Studio with Apple M2 Ultra scored 4,590; Apple Mac mini with Apple M4 Pro scored 5,789. For quantized precision level: HP Z2 Mini G1a scored 18,326; Apple Mac Studio scored 5,661; Apple Mac mini scored 6,994. Above the chart is a banner reading Up to 3.2 times the AI performance.
Geekbench AI CPU FP32 and quantized scores. Source: PT.
Three bar charts showing Geekbench AI GPU — ONNX framework results at three precision levels. Higher is better. Bars compare three systems. For FP32 single precision level: HP Z2 Mini G1a with AMD Ryzen AI Max+ PRO 395 scored 21,135; Apple Mac Studio with Apple M2 Ultra scored 18,008; Apple Mac mini with Apple M4 Pro scored 14,199. For FP16 Half precision level: HP Z2 Mini G1a scored 27,828; Apple Mac Studio scored 19,323; Apple Mac mini scored 16,050. For quantized precision level: HP Z2 Mini G1a scored 16,788; Apple Mac Studio scored 17,588; Apple Mac mini scored 14,342. Above the chart is a banner reading Up to 73% better AI inference performance.
Geekbench AI GPU FP32, FP16, and quantized scores. Source: PT.

What we learned: Creative tests

The workflows of creative professionals, such as video editing, image processing, and digital content creation, may combine heavy CPU, GPU, and memory usage. Faster performance of any or all of these resources could directly affect the lifecycle of creative projects, from concept to completion. Benchmarking tasks from creative applications helps show how well systems can support creative workflows, helping boost productivity and responsiveness.

Bar chart showing Chaos V Ray CPU Benchmark — Overall score rendering performance results. Higher is better. Bars compare three systems: HP Z2 Mini G1a with AMD Ryzen AI Max+ PRO 395 scored 40,754; Apple Mac Studio with Apple M2 Ultra scored 37,632; Apple Mac mini with Apple M4 Pro scored 27,711. Above the chart is a banner reading Up to 47% better rendering performance.
Chaos V-Ray CPU Benchmark Overall score. Source: PT.

The Chaos V-Ray Benchmark measures system rendering performance using CPUs, GPUs, or both together. It evaluates rendering speed with V-Ray 6 engines powered by RTX and CUDA technologies.7

Conclusion

When it comes to choosing a compact yet powerful desktop PC for demanding workloads, our benchmark results clearly show that the HP Z2 Mini G1a workstation with AMD Ryzen AI Max+ PRO 395 processor stands out against the Apple Mac mini and Apple Mac Studio. Its overall superior performance in both general applications and AI-focused tasks suggests it can help modern professionals be more productive while giving them the tools they need to leverage new AI workflows. For teams seeking high productivity and strong AI capabilities in a small form factor, the HP Z2 Mini G1a presents a compelling, performance-driven choice.

  1. AMD, “AMD Ryzen AI Max+ PRO 395,” accessed October 3, 2025, https://www.amd.com/en/products/processors/laptop/ryzen-pro/ai-max-pro-300-series/amd-ryzen-ai-max-plus-pro-395.html.
  2. TechPowerUp, “AMD Strix Halo,” accessed October 3, 2025, https://www.techpowerup.com/gpu-specs/amd-strix-halo.g1096.
  3. HP, “HP Z2 Mini G1a Workstation Desktop PC – AI PC | HP® Official Site, accessed September 24, 2025, https://www.hp.com/us-en/workstations/z2-mini-a.html.
  4. PassMark Software, “PerformanceTest,” accessed September 24, 2025, https://www.passmark.com/products/performancetest/index.php.
  5. TechSpot, “MAXON Cinebench R23,” accessed September 24, 2025, https://www.techspot.com/downloads/7579-cinebench-r23.html.
  6. Geekbench, “Introducing Geekbench AI,” accessed September 23, 2025, https://www.geekbench.com/ai/.
  7. Chaos, “How fast is V-Ray on your hardware?” accessed September 23, 2025, https://www.chaos.com/vray/benchmark?srsltid=AfmBOooSmUTW7VCzZdD-wc_w4egZ-scIiEPgjPtJz3JuaOuSZvSKUV3Q.

This project was commissioned by HP.

October 2025

Principled Technologies is a registered trademark of Principled Technologies, Inc.

All other product names are the trademarks of their respective owners.

Forgot your password?