Leaks have surfaced revealing details about Nvidia’s upcoming RTX Pro 6000 Blackwell GPU, which is shaping up to be the company’s most powerful workstation graphics card.
According to information spotted on Leadtek’s website and backed by NBD shipping data, this powerhouse GPU boasts 24,064 CUDA cores and a massive 96GB of GDDR7 ECC memory. That’s a serious spec leap compared to its predecessor, the RTX 6000 Ada, and even Nvidia’s expected gaming flagship, the RTX 5090.
RTX Pro 6000 and the Mystery “X” Variant
NBD 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.“ However, the exact differences between the standard and “X“ versions remain unclear.
It’s also notable that Nvidia appears to be using a “Pro“ branding for its new workstation GPUs, possibly indicating a shift in positioning.
A 600W Power-Hungry Beast
The RTX Pro 6000 Blackwell is an absolute power monster, with a 600W Total Graphics Power (TGP) delivered through a 16-pin PCIe 5.0 connector. This is over twice the power requirement of the RTX 6000 Ada, which suggests Nvidia is pushing the limits of workstation performance.
Unlike traditional workstation cards that feature blower-style cooling, this one is expected to adopt a dual-flow-through cooling design similar to the RTX 5090. Effective cooling will be crucial, given its massive core count and high power draw.
Built for AI, Rendering, and Pro Visualization
According to Tom’s Hardware, workstation GPUs under Nvidia’s RTX lineup are designed for professional applications. However, 96GB of VRAM is likely overkill unless users work on AI training, game development, high-end content creation, or CAD workflows.
The RTX Pro 6000 Blackwell runs on a 512-bit memory interface with ECC-enabled GDDR7, making it a strong candidate for AI workloads and scientific computing. This positions it as Nvidia’s closest equivalent to a Titan-class Blackwell GPU for prosumers.
How Does It Stack Up?
For perspective, the RTX Pro 6000 Blackwell has 24,064 CUDA cores—just 512 cores shy of the full GB202 configuration, which has 24,576 CUDA cores. This slight cut is likely due to yield limitations during manufacturing.
Interestingly, it also outpaces the GeForce RTX 5090, which is expected to feature 21,760 CUDA cores—2,304 fewer cores than the RTX Pro 6000. However, due to its 600W TGP, the Pro 6000 may run at lower clocks to balance power consumption, especially with power-hungry GDDR7 memory in the mix.
Specifications at a Glance
- Process: 5nm
- GPU: GB202
- CUDA Cores: 24,064
- Tensor Cores: 752
- RT Cores: 188
- Memory: 96GB GDDR7 ECC
- Memory Bus: 512-bit
- Clock Speeds: Base 2017 MHz / Boost 2407 MHz
- TGP: 600W
- Cooling: Dual-flow-through (RTX 5090-style)
- Display Outputs: 4x DisplayPort 2.1
- Interface: PCIe 5.0 x16
- Size: 304mm x 137mm x 40mm (dual-slot)
Expected Announcement at GTC 2025
Nvidia’s GTC 2025 conference on March 17 will likely bring official details about the RTX Pro 6000 Blackwell and insights into its potential applications in AI, simulation, and professional visualization.
With the most CUDA cores ever in a workstation GPU, cutting-edge GDDR7 memory, and a 600W power envelope, this Blackwell beast is shaping up to be a true game-changer for professionals pushing the limits of compute performance.
🚀 Stay tuned for more updates as we approach GTC 2025!
