NVIDIA RTX PRO 6000 “Blackwell” 96 GB GPU Is Up To 5% Faster Than 32 GB RTX 5090 In Pure Path Tracing Benchmark

Hassan Mujtaba
NVIDIA's 96GB RTX PRO 6000 Matches Four RTX 5090s on a 230B AI Model While Drawing a Quarter the Power 1

We have the first benchmarks of NVIDIA's RTX PRO 6000 "Blackwell" GPU with more cores and three times the memory of the RTX 5090 in a pure path tracing test.

NVIDIA's Latest Titan Behemoth, The RTX PRO 6000 "Blackwell" GPU, Has Been Tested In A Pure Path Tracing Benchmark & Performs Nearly Similar To An RTX 5090 Despite Having More Cores & Memory

NVIDIA has launched its flagship prosumer graphics card, the RTX PRO 6000 "Blackwell". This card is way better on paper than the RTX 5090, the top GeForce gaming graphics card. While both cards are based on the same architecture and even the same GB202 GPU, the specs for the RTX PRO 6000 are better.

Related Story Never-Released GeForce RTX 3050 Ti Desktop Card Appears Online, Featuring GA106 Die With 3328 CUDA Cores
Image Source: NVIDIA

To sum up the specs first, the NVIDIA RTX PRO 6000 graphics card comes with 24,064 cores within 188 SMs while the RTX 5090 features 21,760 cores within 170 SMs. That's 10.5% more than the RTX 5090. The full GB202 GPU packs 192 SMs, so while the RTX PRO is still not the best that Blackwell offers, it is the biggest core count we have seen yet on a consumer graphics card. It also makes use of the full 600W TDP that the single 12V-2x6 connector can offer, 25W more than 575W on the RTX 5090. All of this core power leads to some big performance figures, which are as follows:

  • 4000 AI TOPS (+19% more than 5090's 3352 AI TOPS)
  • 125 TFLOPS FP32 (+19% more than 5090's 105 TFLOPS)
  • 380 TFLOPS RT (+19% more than 5090's 318 TFLOPS)

In addition to the larger core specs, the NVIDIA RTX PRO 6000 "Blackwell" also comes with 96 GB of GDDR7 memory, which is 3x the frame buffer of the 32 GB RTX 5090. This is offered along the same 512-bit bus with the same 28 Gbps memory speed for 1792 GB/s of total bandwidth. The card will launch in the same double-flow-through design with a dual-slot form factor and will support MIG (Multi-Instance-GPU) capability with 4x 24 GB, 2x 48 GB, or 1x 96 GB modes.

Now this flagship solution should offer a nice uplift over the already impressive GeForce RTX 5090. We obtained some new benchmarks that have been carried out within the GameTechBench, which features full path tracing (CGI rendering) and should be a nice demonstration of how powerful the new card is.

_rtx-6000_uhd
_rtx-6000_qhd

The NVIDIA RTX PRO 6000 Blackwell leads the 4K performance chart and ends up with a 5% lead over the RTX 5090. At 1440p, the RTX PRO 6000 seems to be slightly behind the RTX 5090, but these are both real-time ray tracing tests.

In the offline Path Tracing rendering mode, the NVIDIA RTX PRO 6000 "Blackwell" sits 2% faster than the RTX 5090.

So you can see that the gains aren't that big for NVIDIA's newest behemoth despite being 19% better on paper. This could be due to underutilization of such hefty hardware or drivers needing optimization work, but overall, the 96 GB VRAM and the upgraded specs will definitely entice the vast majority of workstation, prosumers, and AI enthusiasts out there. This card will also be the benchmark within the visualization segment, achieving new levels of realism using its AI-tuned hardware.

Hassan Mujtaba Photo

About the author: A Software Engineer by training and a PC enthusiast by passion, Hassan Mujtaba serves as Wccftech's Senior Editor for hardware section. With years of experience in the industry, he specializes in deep-dive technical analysis of next-generation CPU and GPU architectures, motherboards, and cooling solutions. His work involves not only breaking news on upcoming technologies but also extensive hands-on reviews and benchmarking.

Follow Wccftech on Google to get more of our news coverage in your feeds.

Button