NVIDIA Quadro RTX 6000 Ada Generation
NVIDIA Quadro RTX 6000 Ada Generation
48GB GDDR6 Graphics Card
900-5G133-2250-000
A full-height, dual-slot professional GPU built on the Ada Lovelace architecture — 48GB of ECC memory, 18,176 CUDA cores, and enough tensor and RT throughput to handle serious 3D rendering, simulation, and mid-scale AI workloads in a single card.
Not every AI or visualization workload needs a data-center-class accelerator locked behind a rack of liquid cooling. A large chunk of professional work — CAD, VFX, medical imaging, digital twins, and small-to-mid-size model fine-tuning — runs perfectly well on a single well-specced workstation card, as long as that card has enough memory headroom to hold the dataset or scene without constant swapping. That’s exactly the niche the Quadro RTX 6000 Ada was built to fill.
Part number 900-5G133-2250-000 identifies this as a factory-new, full-height NVIDIA Quadro RTX 6000 Ada Generation card with 48GB of GDDR6 memory and ECC (error-correcting code) protection — the kind of reliability feature that matters when a single bit-flip could silently corrupt a rendering job or a training run hours into execution. It sits in NVIDIA’s professional lineup as the flagship 48GB Ada-generation workstation card, one generation behind the newer Blackwell-based RTX PRO 6000 series, but still a very capable and considerably more affordable option for teams that don’t need the latest silicon.
Ada Lovelace, not Blackwell — and why that’s often fine
It’s worth being upfront about generational placement: this is an Ada-architecture card, not the newer Blackwell-based RTX PRO 6000. In practice that means somewhat lower raw throughput and no FP4 tensor precision support — but it also means a meaningfully lower price point for teams whose workloads (rendering, CAD, mid-size inference, multi-display professional visualization) don’t require the newest architecture to run well.
What 48GB of ECC memory actually buys you
Large-scene headroom
48GB is enough to hold complex CAD assemblies, VFX scenes, or medical imaging datasets entirely in VRAM, avoiding the performance cliff that comes from spilling into system memory.
ECC memory protection
Error-correcting code memory catches and corrects bit-level errors automatically — critical for long-running simulation, rendering, or training jobs where silent corruption is costly.
Multi-display professional output
Four DisplayPort 1.4 outputs support up to 7680×4320 across dual displays, built for multi-monitor CAD, editorial, and visualization workstations.
Mid-scale AI capable
568 third-gen Tensor Cores and 1,457 TFLOPS of tensor performance make this a workable card for fine-tuning and inference on models that fit comfortably within 48GB.
Specifications overview
| Field | Detail |
|---|---|
| Part number | 900-5G133-2250-000 |
| GPU | NVIDIA RTX 6000 Ada Generation |
| CUDA cores | 18,176 |
| Tensor / RT cores | 568 3rd-gen Tensor Cores / 142 2nd-gen RT Cores |
| Memory | 48GB GDDR6 with ECC, 384-bit bus, 960 GB/s bandwidth |
| Performance | 91.1 TFLOPS FP32 · 210.6 TFLOPS RT · 1,457 TFLOPS Tensor |
| Interface | PCI Express 4.0 x16 |
| Outputs | 4x DisplayPort 1.4, up to 7680×4320 @ 60Hz (dual display) |
| Form factor | Full-height, dual-slot |
| Power | 300W operational |
| OS support | Windows 10, Linux |
Specs sourced from reseller listing data — confirm final figures against NVIDIA’s official RTX 6000 Ada datasheet before finalizing a purchase or build spec.
Where this card sits among 48GB options
48GB professional GPUs span a wide price range depending on generation and condition. Here’s how this listing compares to related cards pulled from the same reseller catalog.
Prices from ServerSupply.com’s catalog as of this writing. A 5% checkout discount plus an additional 1% mobile-app discount applies on this listing, bringing the effective price closer to ~$10,246. Refurbished units carry a shorter warranty window than new stock — confirm terms before comparing purely on price.
Buying tip: This listing is marked NCNR (no cancel, no return) with a 90-day warranty — standard for surplus/channel GPU inventory, but worth factoring into your decision versus buying through a standard authorized NVIDIA/PNY channel with longer warranty coverage.
FAQ
Is this the same as the newer RTX PRO 6000 Blackwell card?
No. This is the previous-generation RTX 6000 Ada card. The RTX PRO 6000 Blackwell Workstation Edition is a newer, higher-performance card with 96GB of GDDR7 memory and a different architecture entirely — the two are easy to confuse by name but are not interchangeable.
What does ECC memory actually protect against?
ECC memory detects and corrects single-bit errors automatically, which matters for long simulation, rendering, or training runs where an undetected bit-flip could silently corrupt output hours into a job.
Is 300W power draw a concern for existing workstations?
Most professional workstation chassis and power supplies built for GPU work in the last several years comfortably support a 300W card, but confirm your PSU headroom and case airflow before purchase, especially in dense multi-GPU builds.
What does “NCNR” mean on this listing?
No Cancel, No Return — a common term in surplus/channel electronics sales meaning the order can’t be canceled or returned once placed, aside from standard manufacturer warranty claims.
Get a firm quote on 900-5G133-2250-000
Pricing on professional GPUs shifts fast across the channel and refurb market. Send us your quantity and timeline and we’ll return current pricing, including any available volume or trade-in options.
Email us for a quote →
