- The details of the fastest workstation video card in NVIDIA have been online
- There are flavors RTX Pro 6000 Blackwell and RTX Pro 6000 X Blackwell
- The No X version has 24,064 CUDA and 96 GB of memory ECCDR7 memory
Online leaks seem to reveal details of the next RTX Pro 6000 Blackwell RTX Pro 6000 GPU, which promises to be the company’s most powerful work card to date.
The information detected on the Leadtek website, and backed by NBD shipping data, points to the card that balances 24,064 CUDA cores and 96 GB of GDDR7 memory.
These specifications are significantly higher than the current RTX 6000 ADA, which was launched more than two years ago, and the expected RTX 5090 games, which, according to reports, presents 2,304 less cores.
NBD shipping data show two versions of the workstation GPU, the RTX Pro 6000 Blackwell and the RTX Pro 6000 X Blackwell, are sent to India for “trial purposes.” We are not sure what the X means, or how different the other version will be. However, it is remarkable that Nvidia seems to be identifying her new cards with a professional label.
Harukaze published the information about one of these cards (the version NO X) in X in X, which shows that the GPU is executed in an 512 -bit memory interface, and GDDR7 admits ECC (error correction), which makes it a good option for servers and systems where the accuracy of the data is essential.
Hungry beast of power
The power requirements are remarkably high for the new GPU, with a total graphic power of 600 W delivered through a PCIE 5.0 connector of 16 pins. This places it among the most famous energy GPUs in the Nvidia alignment, more than twice than required by RTX 6000 ADA, but that should not be a surprise too much. Unlike the previous work station cards, the traditional blowing style refrigerator is likely to save and take your design signs from RTX 5090, with a double flow cooling system.
Writing about the escape, Tom hardware Notes: “The GPUs of the workstation under the NVIDIA RTX line are tailored to professional applications. However, in most scenarios, 96 GB of Vram is excessive unless you are looking for training or inference of the locally. Although I would not normally use these GPUs for games, additional memory can be useful in the tasks involved with the development of games, the creation of content, demand and computer -assisted design. “
The GPU is expected to be officially introduced to the NVIDIA GPU Technology Conference (GTC) in March 2025. The price has not been confirmed, but it will probably fall into the same high -end range as its predecessor, which can be between $ 6,000 and $ 8,000.