Nvidia's next-generation RTX Pro 6000 Blackwell Workstation Edition graphics card, featuring 96 GB of ECC memory, has been benchmarked in Geekbench, offering a sneak peek into its potential performance. Despite the impressive hardware, the results have raised some eyebrows due to the card’s surprisingly similar scores compared to its gaming counterpart, the GeForce RTX 5090. The RTX Pro 6000 Blackwell Workstation Edition is powered by the GB202 GPU, which boasts 24,064 CUDA cores spread across 188 streaming multiprocessors. This GPU can reportedly operate at a clock speed of up to 2,617 MHz. The card also carries 96 GB of ECC memory, designed to ensure data integrity in demanding workloads such as those found in professional visualization (ProViz) applications. In contrast, the GeForce RTX 5090, currently one of the best graphics cards for high-end gaming, also uses the GB202 GPU but with slightly fewer CUDA cores—21,760—and operates at a clock speed of up to 2,410 MHz. It comes equipped with 32 GB of GDDR7 memory, which is standard for gaming applications. Based on the Geekbench OpenCL benchmark results, the new RTX Pro 6000 Blackwell Workstation Edition scored 368,219 points, while the GeForce RTX 5090 achieved a slightly higher score of 376,858 points. This 2.3% performance gap is minimal, especially considering the substantial hardware differences between the two cards. Here's a deeper look at some specific benchmarks: Background Blur: The RTX Pro 6000 scored 63,762 points, or 263.9 images per second, compared to the RTX 5090's 75,075 points (310.7 images per second). Face Detection: Scores were 60,254 for the RTX Pro 6000 (196.7 images per second) and 73,968 for the RTX 5090 (241.5 images per second). Horizon Detection: The RTX Pro 6000 outperformed with 684,753 points (21.3 gigapixels per second) versus the RTX 5090's 637,294 points (19.8 gigapixels per second). Edge Detection: The RTX Pro 6000 scored 864,739 points (32.1 gigapixels per second) while the RTX 5090 reached 838,261 points (31.1 gigapixels per second). Gaussian Blur: The RTX Pro 6000 scored 832,815 points (36.3 gigapixels per second) compared to the RTX 5090's 795,994 points (34.7 gigapixels per second). Feature Matching: Scores were very close, with 57,199 points (2.25 gigapixels per second) for the RTX Pro 6000 and 57,464 points (2.27 gigapixels per second) for the RTX 5090. Stereo Matching: The RTX Pro 6000 and RTX 5090 achieved nearly identical scores of 2,797,728 and 2,802,350 points respectively (2.66 terapixels per second). Particle Physics: The RTX Pro 6000 scored 1,114,648 points (49,056.6 frames per second) while the RTX 5090 scored 1,069,886 points (47,086.6 frames per second). However, there are several factors to consider when interpreting these results. Firstly, the RTX Pro 6000 Blackwell Workstation Edition is still in the pre-release stage and being tested with pre-release drivers. Performance can significantly improve once the card is launched with finalized drivers. Secondly, the power limits of the RTX Pro 6000 might be set artificially low for testing, which could cap its compute performance. The GeForce RTX 5090, on the other hand, has a maximum TGP (Total Graphics Power) of 575W. Lastly, the current drivers limit the amount of memory visible to OpenCL applications to just 23.8 GB, far below the 96 GB the card actually carries. This further indicates that the driver software is not in its final form. Additionally, the RTX Pro 6000 Blackwell is not expected to be Nvidia's top-tier ProViz offering. According to earlier reports, the flagship model will likely be named the RTX Pro 6000 X Blackwell, suggesting that this tested card is one step below the highest-performance option. For the latest updates and insights into Nvidia's upcoming products, make sure to follow Tom's Hardware on Google News. Click the Follow button to stay informed with their comprehensive coverage of news, analysis, and reviews.