Compact workstation comparison: bottom line for AEC pros
A compact workstation comparison for AEC professionals weighs the HP Z2 Mini G1a and Lenovo ThinkStation P3 Ultra SFF G2 as powerful, space‑saving systems that approach performance and expandability in very different ways, helping architects, engineers, and contractors balance constrained desks with demanding 3D, BIM, and visualization workloads.
If you want a small form factor workstation that breaks from the old desktop‑versus‑mobile compromises, the HP Z2 Mini G1a is the bold option built around a shared memory architecture and powerful integrated graphics, brought over from the same AMD Ryzen AI Max PRO platform used in the HP ZBook Ultra G1a. By contrast, the Lenovo ThinkStation P3 Ultra SFF Gen 2 is a professional mini PC that behaves more like a trimmed‑down tower, using Intel Core Ultra CPUs plus discrete NVIDIA graphics and multiple low‑profile PCIe slots for expansion. For most BIM‑heavy AEC teams in tight spaces, the HP design is cleaner and easier to deploy; for firms that need certified discrete GPUs, extra cards and long‑term upgrade paths, Lenovo’s slightly larger chassis is more practical.
HP Z2 Mini G1a: shared memory and integrated GPU for BIM and AI
The HP Z2 Mini G1a is part of a new generation of mainstream compact workstations, alongside the HP ZBook Ultra G1a, that moves beyond incremental updates to deliver a “genuine shift — in GPU memory capacity, in integrated graphics performance, and in what compact form factors can now deliver for professional AEC work”. It uses the AMD Ryzen AI Max PRO processor with an integrated AMD Radeon 8060S GPU, and in this desktop form runs that chip at a higher 150W thermal envelope for more sustained performance. In CAD, BIM and entry‑level visualization, this integrated GPU competes with mid‑range discrete GPUs while avoiding the frame‑rate collapses that occur when small fixed VRAM buffers overflow.
The key to this behaviour is unified system memory. The integrated GPU can address up to 96 GB from a 128 GB RAM configuration, letting it keep large visualization scenes responsive where 8–12 GB discrete laptop GPUs would falter or even crash. According to the cited independent testing, scenes that break some fixed‑memory GPUs run without issue on this platform. For AEC firms exploring local AI, the same architecture allows sizeable language models and autonomous agents to run fully on‑premise, so sensitive project data never has to leave the network. The trade‑off is that if your work revolves around huge, highly detailed models or high‑end real‑time rendering, a workstation with a high‑tier discrete GPU will still offer a more complete experience.
Lenovo ThinkStation P3 Ultra SFF G2: compact yet expandable mini tower
The Lenovo ThinkStation P3 Ultra SFF Gen 2 approaches the professional mini PC idea from the opposite direction: start with a small tower concept and shrink it until it fits under a monitor without giving up expansion. Billed as an AI workstation, it bridges the gap between a mini‑PC and a small tower, with a 3.9‑litre chassis (202 x 87 x 223 mm) that remains compact but is “large enough to house multiple low-profile PCIe cards,” providing real expansion options while still counting as a mini‑PC. This design is aimed at users who want a small form factor workstation but still need add‑in cards or replaceable discrete GPUs over time.
Inside, the P3 Ultra SFF G2 in the reviewed configuration combines an Intel Core Ultra 9 285 processor with 64 GB of DDR5‑6400 memory, a 1 TB PCIe Gen5 x4 SSD, and an NVIDIA RTX 4000 Ada SFF GPU with 20 GB of GDDR6, plus Intel integrated graphics. Two low‑profile PCIe slots (one Gen4 x16, one Gen4 x4) and a 330 W external PSU underpin that expandability. The heavily vented front panel and airflow‑focused chassis allow the system to support a dual‑slot GPU while staying reasonably small. This makes it well suited to more demanding visualization, GPU‑accelerated simulation, or other professional‑grade applications that benefit from certified NVIDIA drivers and upgradeable PCIe hardware, provided you can accommodate a slightly larger box on or under the desk.
Specs, thermals and trade‑offs: which compact workstation fits your desk?
| Spec | HP Z2 Mini G1a | Lenovo ThinkStation P3 Ultra SFF G2 |
|---|---|---|
| CPU platform | AMD Ryzen AI Max PRO (integrated AMD Radeon 8060S GPU) | Intel Core Ultra 9 285 (8P + 16E, up to 5.6 GHz) |
| GPU approach | Integrated graphics using unified system memory, up to 96 GB GPU‑accessible | NVIDIA RTX 4000 Ada SFF 20 GB + Intel Xe integrated graphics |
| Max memory (reviewed config) | Up to 128 GB system RAM, 96 GB accessible by GPU | 64 GB DDR5‑6400 (2×32 GB) in reviewed spec |
| Storage (reviewed config) | Noted for handling large datasets via unified RAM; specific SSD config not detailed | 1 TB PCIe Gen5 x4 M.2 2280 SSD |
| Expansion slots | Focus on integrated GPU and shared memory; no discrete PCIe GPU slots described | 1× low‑profile PCIe Gen4 x16, 1× low‑profile PCIe Gen4 x4 |
| Form factor & volume | Compact mini workstation form factor, designed to shift expectations of small workstations | Mini‑PC, 3.9 L chassis, 202×87×223 mm, 1.8 kg |
| Power & thermals | Integrated GPU runs at 150 W thermal envelope; designed to keep performance up under load | 330 W external PSU, chassis optimized for strong front‑to‑back airflow |
| Ports (front, reviewed config) | Not detailed in source; emphasis on internal architecture and memory | 2× USB‑C 20 Gbps, 1× USB‑A 10 Gbps, 3.5 mm audio jack |
| Target workloads | CAD, BIM, entry‑level visualization, large datasets and local AI in space‑constrained environments | AI workstation, demanding visualization and workflows needing discrete GPU and PCIe expansion |
Both machines are small form factor workstations aimed at serious professional workflows where desk space is scarce. Yet they embody opposite philosophies. HP bets on a high‑power integrated GPU and shared memory to sidestep traditional VRAM ceilings in BIM and entry‑level visualization, while also supporting sizeable local AI workloads. Lenovo offers a slightly larger but more flexible box that can hold a powerful dual‑slot discrete GPU and an extra PCIe card, making it easier to adapt to changing software or hardware needs over time.
For most everyday AEC production—Revit or similar BIM work, CAD drafting, and light real‑time walkthroughs—the HP Z2 Mini G1a’s integrated Radeon 8060S is a credible professional tool that can outlast some mid‑range discrete GPUs when scenes grow beyond their VRAM. However, both platforms share one caveat: if your practice routinely works with very large, complex models or expects top‑tier real‑time rendering performance, you will still want to keep a high‑end discrete GPU tower in the mix for those heaviest tasks.
Buy if / Skip if
- Buy the HP Z2 Mini G1a if you want a compact workstation comparison winner for BIM‑first AEC work where unified memory and integrated graphics can keep large scenes smooth without juggling VRAM limits.
- Skip the HP Z2 Mini G1a if your primary workloads are massive models or high‑fidelity real‑time visualization that clearly demand a top‑tier discrete GPU workstation for the full experience.
- Buy the HP Z2 Mini G1a if running local AI agents and language models on sensitive project data is important and you need them to stay inside your own network.
- Skip the Lenovo ThinkStation P3 Ultra SFF G2 if absolute minimal footprint or ultra‑clean, appliance‑like deployment matters more to you than add‑in cards and upgrade paths.
- Buy the Lenovo ThinkStation P3 Ultra SFF G2 if you need a professional mini PC with discrete NVIDIA RTX graphics, PCIe slots for extra cards, and room to refresh GPUs over time.
- Skip the Lenovo ThinkStation P3 Ultra SFF G2 if your workflows rarely stretch beyond CAD and BIM and you would not make use of its discrete GPU or expansion capabilities.
