NVIDIA Unveils Blackwell RTX: Showcasing Neural Rendering & Gaming with RTX PRO 6000 Running Four Instances of Cyberpunk 2077 via MIG

NVIDIA continues to make waves in the tech industry with its latest advancements in neural rendering and gaming, particularly through its Blackwell RTX GPU series, which includes the RTX 5090 and RTX PRO 6000.

NVIDIA Showcases Blackwell RTX Innovations at Hot Chips 2025

Launched in January 2023, the Blackwell RTX architecture has unveiled groundbreaking features that primarily focus on artificial intelligence (AI).This innovation is not a new venture for NVIDIA, as its journey began with the launch of CUDA in 2006, marking a significant advancement in accelerated computing and AI.

AI development timeline from 2006 to 2025, highlighting key NVIDIA innovations and achievements.

NVIDIA asserts that advancements in compute density have surpassed Moore’s Law scaling, achieved through techniques such as Sparsity, a new Instruction Set Architecture (ISA), and optimized lower precision formats. The 2018 introduction of real-time ray tracing and the subsequent launch of DLSS the following year represent critical milestones in this evolution.

NVIDIA Blackwell architecture diagram with GPUs, CPUs, and memory specs detailed for performance scaling.

These innovations have been achieved using advanced technologies, including RT cores and Tensor cores, and Blackwell now amplifies these capabilities to a new level.

RTX ecosystem framework featuring Blackwell architecture tools for AI and rendering enhancements.

In the data center space, NVIDIA introduced FP4 precision, providing a 4x boost for workloads that require dense scaling. Jensen Huang, NVIDIA’s CEO, aims to reaffirm AI’s pivotal role in graphics, facilitating the emergence of the neural rendering era with Blackwell RTX. The RTX brand signifies innovation in simulation, content creation, and gaming, paving the way to extend data center technologies to consumer RTX GPUs.

Neural rendering promises enhanced gaming with AI visuals, smooth responsiveness, and adaptive experiences.

So, what does the Blackwell architecture contribute? It boasts advancements like DLSS 4, MFG, ACE, and enhanced Path Tracing, all designed to accelerate performance and improve visual fidelity. NVIDIA claims Blackwell RTX can achieve a “10x amplification in performance, footprint, and design cycle.”DLSS 4 strategically uses AI to render 100% of the pixels following the initial frame, ultimately leading to faster rendering times and extended battery life for mobile devices.

RTX Blackwell chip with design principles: optimized neural workloads, reduced memory, efficiency.

The key design principles of the RTX Blackwell GPU include:

Optimization for new neural workloads
Minimizing memory footprint
Ensuring quality service for neural and graphics tasks
Scalable energy efficiency

NVIDIA RTX Blackwell chip with AI, Tensor Cores, G7 Memory, advanced rendering architecture.

On a technical scale, RTX Blackwell is an engineering powerhouse, offering 4000 AI Tera Operations Per Second (TOPS) and High-Speed FP4 support based on 5th Generation Tensor Cores. It provides up to 360 RT TFLOPs targeted at Mega Geometry with the 4th Gen RT Cores, while the AI Management Process (AMP) effectively manages AI models alongside graphics processing.

Diagram of RTX Blackwell SM architecture for neural shaders, featuring shader and tensor cores.

The architecture of the RTX Blackwell Streaming Multiprocessor (SM) diverges significantly from its Data Center counterpart. One notable improvement is the integration of FP32 and INT32 units that were previously separate, enhancing processing efficiency.

Diagram showing RTX Blackwell SM enhancing Shader Execution Reordering by 2X with neural shaders and cores.

Furthermore, RTX Blackwell enhances Shader Execution Reordering (SER), doubling the efficiency of shader execution.

Blackwell 5th Gen Tensor Cores with FP4, showcasing throughput advancements over predecessors.

The 5th Generation Tensor Core introduces FP4 support and includes MFG mode in DLSS 4, allowing the GPU to use AI to render four frames simultaneously.

DLSS4 + RTX Blackwell boosts gaming with faster, efficient frame rendering and GPU power savings.

As a result, utilizing DLSS 4 along with Frame Generation significantly reduces individual frame rendering time by providing a tenfold increase in core rail gating speed, while achieving 100 times faster DRAM self-refresh rates. Mobile platforms can experience a reduction of up to 2x in GPU power consumption, greatly enhancing battery life.

GDDR7 vs. GDDR6x: Enhanced data rates and efficiency in new graphics DRAM standard.

The introduction of GDDR7 enables the RTX Blackwell to achieve speeds of up to 30 Gbps, effectively doubling the data rate of its predecessor GDDR6. This new memory standard further enhances efficiency on mobile platforms.

Thanks to NVIDIA’s AMP unit, simultaneous execution of AI and graphics workloads is a reality, leading to smoother frame delivery and quicker model responses.

RTX PRO 6000 overview displaying Universal MIG and GPU instance optimization details.

Transitioning from gaming applications to professional uses, NVIDIA is featuring novel capabilities in the RTX PRO 6000, such as Universal MIG. This allows for up to four instances of the RTX PRO GPUs, each equipped with 24 GB VRAM, to operate simultaneously with consistent latency and throughput.

An impressive demonstration showcased the RTX PRO 6000’s capability by running four instances of Cyberpunk 2077 at 1080p on maximum settings, a task manageable for this powerful GPU.

RTX PRO 6000 Multi-tenant Scaling Bar Chart, Cyberpunk 2077 at 1080p Performance Metrics.

With the use of a standard time-sliced instance as a baseline, comparisons against MIG 2x and 4x modes revealed a remarkable 60% increase in scalability. The RTX PRO 6000 Blackwell GPU is indeed well-suited for managing multiple instances of demanding applications like Cyberpunk 2077.

Graph comparing Blackwell FP4 performance and VRAM usage with image samples.

Overall, NVIDIA’s Blackwell GPU architecture has been making strides since its release, continuously evolving for both consumer and professional applications. As more games and content creation tools begin to incorporate the extensive AI and neural enhancements offered by Blackwell, the anticipation surrounding future developments in this space is palpable.

Source & Images