NVIDIA RTX PRO 6000 “Blackwell” 96 GB GPU Outperforms 32 GB RTX 5090 by 5% in Pure Path Tracing Benchmark

NVIDIA has introduced the RTX PRO 6000 “Blackwell”GPU, marking a significant advancement in graphics technology. Recent benchmarks reveal that this cutting-edge graphics card significantly ups the ante with more cores and triple the memory of its predecessor, the RTX 5090, particularly in pure path tracing tests.

Unveiling the RTX PRO 6000 “Blackwell”: Performance Benchmarks Compared to RTX 5090

The RTX PRO 6000 “Blackwell”has been positioned as NVIDIA’s flagship prosumer graphics card, boasting superior specifications compared to the RTX 5090, which is primarily aimed at gaming enthusiasts. Although both GPUs share the same architecture and utilize the identical GB202 chipset, the RTX PRO 6000 clearly outshines its gaming counterpart in several key aspects.

NVIDIA RTX PRO 6000 Blackwell — Image Source: NVIDIA

To begin with, let’s delve into the specifications. The NVIDIA RTX PRO 6000 is equipped with a remarkable 24, 064 cores distributed across 188 Streaming Multiprocessors (SMs), which is an impressive 10.5% increase compared to the 21, 760 cores housed in the 170 SMs of the RTX 5090. Notably, the full potential of the GB202 GPU encompasses 192 SMs; thus, while the RTX PRO 6000 does not fully utilize the maximum capabilities of Blackwell architecture, it still represents the highest core count available in consumer graphics cards to date. Moreover, this powerhouse operates at a total thermal design power (TDP) of 600W, 25W above the RTX 5090’s 575W, indicating a significant potential for enhanced performance.

Performance Metrics

The performance figures for the RTX PRO 6000 are compelling:

AI TOPS: 4000 (+19% over the RTX 5090’s 3352 AI TOPS)
FP32 Performance: 125 TFLOPS (+19% compared to RTX 5090’s 105 TFLOPS)
Ray Tracing Performance: 380 TFLOPS (+19% more than the 318 TFLOPS delivered by RTX 5090)

In addition to its robust core architecture, the RTX PRO 6000 features a generous 96GB of GDDR7 memory, which is three times the memory capacity of the 32GB RTX 5090. This memory is delivered over a 512-bit bus, maintaining the same 28 Gbps speed for a remarkable bandwidth of 1792 GB/s. The dual-slot form factor of this GPU remains unchanged along with its dual-flow-through design, and it offers Multi-Instance-GPU (MIG) capabilities in configurations of 4x 24 GB, 2x 48 GB, or a single 96 GB instance.

Recent test results conducted via GameTechBench have shown that the RTX PRO 6000 excels in pure path tracing (CGI rendering).In these tests, it leads the 4K performance category, exhibiting a 5% advantage over the RTX 5090. However, at 1440p resolution, results indicate the RTX PRO 6000 falls slightly behind the RTX 5090, though both cards shine when executing real-time ray tracing.

In offline Path Tracing rendering mode, the RTX PRO 6000 delivers a mere 2% performance increase over the RTX 5090, which may suggest that the full capabilities of the hardware are not yet fully optimized. Potential reasons for the modest gains include hardware underutilization or the need for driver improvements. Nevertheless, the introduction of 96GB VRAM and enhanced specifications will likely appeal to professionals in the fields of workstation graphics, prosumers, and AI innovators. The RTX PRO 6000 is set to become a benchmark for visualization, enabling unprecedented levels of realism through its AI-driven hardware enhancements.

Source & Images