ABS AI Workstation - 96GB VRAM (NVIDIA RTX PRO 6000 Blackwell Max-Q), AMD EPYC 4565P, 128GB ECC DDR5 UDIMM, 2x 2TB NVMe M.2, Ubuntu (Zaurion ZRE4565P-RP6000MQ)
- Item #: 9B-59-991-055
- Mfr. Part #: ZRE4565P-RP6000MQ
- Compliance Documents Required for Purchase
- Positioning: Large-Model AI Workstation
- AI engineers running large-model inference (70B-class)
- Teams building RAG platforms and AI agent systems
- Operating System: Ubuntu (Request a Quote for Windows)
- CPU: AMD EPYC 4565P 16 Core, Liquid cooling
- GPU: RTX Pro 6000 MaxQ Blackwell (96 GB GDDR7)
- Motherboard: ASUS Pro WS B850M-ACE SE AMD Micro-ATX
- Memory: 128GB DDR5 Server RAM - ECC, Unbuffered
- SSD: 2 x 2TB M.2 NVMe
- LAN: 1 x 10GB/s LAN, 1 x 2.5GB/s LAN, 1 x 10/100/1000 Mbps Management LAN
- PSU: Single 1000W 80 PLUS Platinum power supply
- Optional: Pre-engineered AI environment with LLM, RAG, automation, submit a quote for pricing
- AI Agent Ready: Pre-installed OpenClaw (enterprise AI agent framework)
- Assembled in USA
- TAA Compliant
96 GB, quietly.
The RTX PRO 6000 MaxQ Blackwell keeps the flagship's entire 96 GB of ECC GDDR7 and trades peak wattage for grace: this whole system runs on a 1000 W 80 PLUS Platinum supply, cool and composed, in an office where people think. Built for 70B-class inference, RAG platforms, and agent systems — OpenClaw ships pre-installed on Ubuntu.
The Reading Room: Two Processors, Examined
NVIDIA RTX PRO 6000 MaxQ Blackwell
- All the memory, none of the roar — MaxQ keeps the full 96 GB of GDDR7 with ECC; large open-weight models load whole, with context to spare, on one card.
- ECC video memory — a service that answers questions for months can't afford a silent bit-flip in its weights. Corrected, not served.
- Efficiency as a feature — the power-tuned profile is why this entire system needs only a 1000 W 80 PLUS Platinum supply, where the full-power sibling ships with a 1300 W supply.
- Universal MIG — partition the card into isolated instances; a RAG platform, an agent fleet, and an experiment can share it politely.
- Blackwell 5th-gen Tensor Cores with FP4 — the low-precision math of modern inference, at architecture-native speed.
AMD EPYC 4565P, Liquid-Cooled
- Server pedigree at a desk — EPYC-class cores handle retrieval, chunking, embedding preprocessing, and OpenClaw agent orchestration beside the GPU.
- Up to 5.7 GHz max turbo — low single-thread latency keeps agent decisions and query pipelines feeling instant.
- 64 MB of L3 cache — tokenization and rerankers stay fed close to the silicon.
- ECC UDIMM platform + liquid loop — data integrity end to end, clocks that hold around the clock, acoustics that respect the room.
Large-model inference is the job description.
Not our claim — the listing's own positioning: AI engineers running large-model inference (70B-class); teams building RAG platforms and AI agent systems. 96 GB of ECC GDDR7 on a single card is what makes that sentence casual.
Same Library. Smaller Power Bill.
Both editions live on this listing — the honest side-by-side:
Three Rooms It Belongs In
The Study · 70B-Class Inference
Large open-weight models with long context, served locally — private, fixed-cost, and quiet enough to share a room with deep work.
The Archive · RAG Platforms
Embed, index, and answer over your own corpus with citations. Want it turnkey? The optional pre-engineered LLM/RAG/automation environment comes by quote.
The Front Desk · Agent Systems
OpenClaw ships pre-installed: triage, drafting, and multi-step automations running around the clock on ECC-protected hardware.
Scene photos are AI-generated illustrations of typical deployments, not photographs of this exact product.
As Configured
| ZRE4565P-RP6000MQ | |
|---|---|
| GPU | NVIDIA RTX PRO 6000 MaxQ Blackwell — power-tuned edition, 96 GB GDDR7 with ECC |
| CPU | AMD EPYC 4565P — 16-core, liquid-cooled, 4.3 GHz base / 5.7 GHz max turbo, 64 MB L3, Socket AM5 |
| Memory | 128 GB DDR5 ECC UDIMM — expandable to 256 GB across 4 DIMM slots |
| Storage | 2× 2 TB M.2 NVMe (4 TB total) |
| Network | 1× 10 Gb/s + 2.5 Gb/s LAN + dedicated management port |
| Expansion | PCIe 5.0 x16 + PCIe 4.0 x16 · ASUS Pro WS B850M-ACE SE |
| Power | 1000 W 80 PLUS Platinum |
| OS | Ubuntu pre-installed, OpenClaw included (Windows available by quote) |
Quiet Answers Before You Buy
What does "MaxQ" mean on this GPU?
The power-tuned edition of the flagship: same 96 GB of GDDR7 ECC, tuned for efficiency and acoustics — which is why this system runs on a 1000 W 80 PLUS Platinum supply while the full-power sibling ships with a 1300 W supply.
Can it really run 70B-class models?
That's the listing's own stated positioning — 96 GB of ECC GDDR7 on one card holds large open-weight models with context to spare, no multi-GPU sharding.
MaxQ or full power — how do I choose?
Same memory, different priorities: MaxQ for offices, acoustics, and efficiency; full power (same listing) when maximum sustained tokens-per-second is the metric.
What does "OpenClaw pre-installed" mean?
The OpenClaw enterprise AI agent framework comes ready on Ubuntu — agentic workflows on day one, no setup. Windows is available by quote.
Government / public-sector eligible?
TAA compliant and assembled in the USA, with compliance documents available for purchase review.
How is it supported?
1-year parts and labor limited warranty backed by ABS manufacturer support. Build to Order — custom built and stress-tested after purchase.
Warranty & Returns
Warranty, Returns, And Additional Information
Warranty
- Limited Warranty period (parts): 1 year
- Limited Warranty period (labor): 1 year
Return Policies
- Return for refund within: non-refundable
- Return for replacement within: non-replaceable
- This item is covered by NeweggBusiness.com's Manufacturer Only Return Policy.
- Read full return policy for details.
Manufacturer Contact Info
- Manufacturer Product Page
- Manufacturer Website
- Support Phone: 1-800-685-3471
- Support Email: [email protected]
- Support Website
- View other products from ABS















