DeskGearHQ

AI & Smart Office

HP Z2 Mini G1a Workstation (Ryzen AI Max+ PRO 395, 64GB, 1TB)

By Nick Miles ยท Updated 2026-10-03 ยท 4 expert sources

HP Z2 Mini G1a Workstation (Ryzen AI Max+ PRO 395, 64GB, 1TB)

Pros

  • +PCMag rated it 4/5 with an Editors' Choice award, and Creative Bloq gave it 8/10
  • +Three-year warranty including onsite service (PCMag)
  • +The Register found its Strix Halo chip trades blows with DGX Spark on single-batch token generation
  • +Windows 11 Pro workstation with HP's Wolf Pro Security (Amazon listing, PCMag)
  • +Compact and mountable, with plentiful connectivity for its size (PCMag)

Cons

  • โˆ’This listing has 64GB; both rated reviews tested 128GB, and no 128GB version was found on Amazon
  • โˆ’Fans get loud under load (PCMag, Creative Bloq)
  • โˆ’Memory is not upgradable, and the chassis is not highly expandable (PCMag, Creative Bloq)
  • โˆ’Creative Bloq calls it expensive, and The Register measured DGX Spark roughly 2-3x faster on time-to-first-token

Score breakdown

  • Expert consensus8.0
  • Effectiveness6.5
  • Build safety9.0
  • Durability7.0
  • Value7.5

Expert Consensus is the mean of PCMag (4/5) and Creative Bloq (8/10); both tested 128GB units, so Effectiveness is reduced by 1.0 for this 64GB listing. Build safety is provisional: our 2026-10-03 sources showed no recall, safety, heat or failure reports for this product, but we have not yet verified its UL/ETL or equivalent safety certification, so the 9.0 is a placeholder, not a certified rating.

What the experts say

โ€œHPโ€™s Z2 Mini G1a compact desktop is a tiny powerhouse, packing a serious wallop for workstation tasks and AI dev work with AMDโ€™s Ryzen AI Max+ processor and 128GB of unified memory.โ€

โ€” PCMag (8.0/10)

โ€œHP has created a potent rival to the Mac Studio, equipping it with its Wolf security software, to make a PC that will appeal to FX pros and AI developers alike.โ€

โ€” Creative Bloq (8.0/10)

โ€œIn this respect, HP's Z2 Mini G1a is one of the better options out there, especially if you're mostly interested in running single-batch LLM inference as opposed to fine-tuning or image gen.โ€

โ€” The Register