Details of Nvidia’s fastest video card ever leak; RTX Pro 6000 Blackwell GPU will have 96GB GDDR7 ECC memory waynewilliams@onmail.com (Wayne Williams) | amznusa.com

• Details of Nvidia’s fastest ever workstation video card have been spotted online
• There are RTX Pro 6000 Blackwell and RTX Pro 6000 X Blackwell flavors
• The non-X version has 24,064 CUDA cores and 96GB of GDDR7 ECC memory

Online leaks appear to reveal details of Nvidia’s upcoming RTX Pro 6000 Blackwell GPU, which promises to be the company’s most powerful workstation card to date.

Information spotted on Leadtek’s website, and backed by NBD shipping data, point to the card rocking 24,064 CUDA cores and 96GB of GDDR7 memory.

These specs are significantly higher than both the current RTX 6000 Ada, which launched over two years ago, and the expected RTX 5090 gaming GPU, which reportedly features 2,304 fewer cores.

NBD’s shipping data shows two versions of the workstation GPU – the RTX Pro 6000 Blackwell and the RTX Pro 6000 X Blackwell – being sent to India for “testing purposes”. We’re not sure quite what the X signifies, or how different it will be from the other version. It’s notable however, that Nvidia seems to be identifying its new cards with a Pro label.

The information on one of these cards (the non-X version) was posted on X by Harukaze, showing that the GPU runs on a 512-bit memory interface, and GDDR7 supports ECC (error correction), making it a good choice for servers and systems where data accuracy is essential.

Power hungry beast

Power requirements are notably high for the new GPU, with a 600W Total Graphics Power delivered through a 16-pin PCIe 5.0 connector. This puts it among the most power-hungry GPUs in Nvidia’s lineup – over twice that required by RTX 6000 Ada – but that shouldn’t come as too much of a surprise. Unlike earlier workstation cards, this one will likely skip the traditional blower-style cooler and take its design cues from the RTX 5090, with a dual-flow-through cooling system.

Writing about the leak, Tom’s Hardware notes, “Workstation GPUs under Nvidia’s RTX lineup are tailor-made for professional applications. However, in most scenarios, 96GB of VRAM is overkill unless you’re looking for training or inferencing AI locally. While you wouldn’t typically use these GPUs for gaming, the added memory can be helpful in tasks involved with game development, content creation, ProViz, and computer-aided design.”

The GPU is expected to be officially introduced at Nvidia’s GPU Technology Conference (GTC) in March 2025. Pricing hasn’t been confirmed, but it will likely fall in the same high-end range as its predecessor, which can be had for between $6,000 and $8,000.

You might also like

• Check out our list of the best graphics cards for video editing you can buy
• Nvidia CEO Jensen Huang teases Blackwell Ultra reveal
• Nvidia has renamed its Blackwell Ultra product series – here’s why