Top single-processor Blackwell accelerator for an AI workstation
96 GB of GDDR7 with ECC and 1.8 TB/s of bandwidth (512-bit bus) make it possible to keep multi-billion-parameter language models and billion-polygon scenes entirely in video memory - without moving the weights out to system RAM. This is the defining parameter for local LLM fine-tuning, generative AI and photorealistic rendering, where memory capacity limits the size of the task more than clock speeds do.
At its core is the NVIDIA Blackwell architecture: 24064 CUDA cores, 752 5th-generation Tensor cores with FP4 support (up to 4000 AI TOPS) and 188 4th-generation RT cores with RTX Mega Geometry. Universal MIG technology splits one graphics processor into isolated instances (up to 4x 24 GB, 2x 48 GB or 1x 96 GB), which is convenient for parallel tasks and VDI profiles on a single card.
- 96 GB GDDR7 ECC, 512-bit bus, 1792 GB/s
- 24064 CUDA, 752 Tensor (5th gen., FP4), 188 RT (4th gen.)
- PCIe 5.0 x16 interface, 1x 16-pin power (CEM5), TGP 600 W
- 4x DisplayPort 2.1b, double-flow-through cooling, 2 slots
- 9th-generation NVENC and 6th-generation NVDEC video engines with 4:2:2 support
Part number VCNRTXPRO6000B-PB is supplied in PNY branded packaging with a full manufacturer warranty and ISV-certified Enterprise drivers (Autodesk, Dassault/SolidWorks, Adobe). This is a scarce item - check availability and the shipment date with a specialist before ordering.
Frequently Asked Questions
What power supply does this card need?
Auxiliary power connector: 1x16-pin. With two of these cards in one system, size the supply with headroom and check the circuit rating.
Will this card fit my case?
The card takes 2 slots, length 305 mm. Measure the clearance from the rear panel to the drive cage before ordering: that is where a few millimetres usually go missing.
How is this card cooled?
Active cooling, fans: 2. It is built for a workstation. For rack mounting next to other cards, choose the passively cooled server variant.
What model size fits into 96 GB of memory?
Roughly a 70B model at 4-bit with room for context, or two 30B models side by side. The exact figure depends on context length: longer conversations consume more cache. Memory: 96 GB GDDR7, 512 bit bus.
How many displays can I connect?
Up to 4 at once. For multi-screen setups this is the limiting factor, not the card's performance.
Which GPU does this card use?
NVIDIA Quadro RTX PRO 6000, 24064 compute cores, 4 nm process. The GPU generation decides which compute formats are accelerated in hardware, and that affects inference speed more than raw memory capacity does.
What warranty is provided and what is in the box?
Warranty: 36 months. Ships in the manufacturer's original packaging. In the box: install guide, 1x CEM5 16-pin power cable, board extender with screws.
Not sure this card fits your system?
We will check compatibility with your platform, calculate the power draw for your configuration and tell you whether the memory is enough for your models.
The model you need is out of stock? ETE is listed among official PNY partners, so we bring professional NVIDIA cards to order with the manufacturer warranty.
Message us on Telegram or leave a request.
Read the review and measurements: RTX PRO 6000 Blackwell: 96 GB GDDR7, MIG and 600 W — review →
