NVIDIA RTX 50 Series: A Quantum Leap in Graphics and AI Performance

The NVIDIA RTX 50 series has arrived! Powered by NVIDIA Blackwell architecture, these cards boast significant advancements over their predecessors. On January 30th, the GeForce RTX 5090 and GeForce RTX 5080 arrive on store shelves (if you can find them!) The GeForce RTX 5070 Ti and GeForce RTX 5070 will be available starting in February.

Let’s look at an overview of the top-of-line RTX 5090 versus the previous generation heavy-hitter, the RTX 4090:

    •    CUDA Cores: 21,760 (vs 16,384)

    •    Memory: 32GB GDDR7 (vs 24GB GDDR6X)

    •    Memory Interface: 512-bit (vs 384-bit)

    •    Ray Tracing Cores: 4th Gen (vs 3rd Gen)

    •    Tensor Cores: 5th Gen, 3352 AI TOPS (vs 4th Gen, 1321 AI TOPS)

    •    TDP: 575W (vs 450W)

More comparisons can be found at: https://www.nvidia.com/en-gb/geforce/graphics-cards/50-series/

Key Features:

The architectural core of the RTX 50 series, Blackwell, introduces 5th-generation Tensor Cores optimised for FP8 and DLSS 4, delivering exceptional AI performance as well as all-new Streaming Multiprocessors that enhance lighting and other visual effects through AI-powered techniques.

DLSS 4 introduces Multi-Frame Generation, an AI technique that can generate up to three additional frames for every rendered frame. This translates to significantly higher frame rates and smoother gameplay. Despite being designed for gaming, DLSS 4 can be leveraged for inference tasks and translates to faster processing times for AI models.

Blackwell also features brand new RT Cores designed to ray trace massive amounts of detailed geometry. The RT Cores have 2X the ray triangle intersection rate of the previous generation, and enhanced compression designed to reduce memory footprint. This allows for the rendering of incredibly detailed environments with millions of polygons, while neural rendering techniques utilise AI to enhance lighting, reflections, and shadows in real-time. Applications that rely heavily on highly realistic simulations and visualisations will benefit greatly from this new technology. 

The adoption of GDDR7 memory provides a significant boost in bandwidth, enabling faster data transfer and smoother performance. This, combined with increased memory capacities, allows for higher resolutions and more demanding game settings.

Blackwell has also been enhanced with PCIe Gen5 and DisplayPort 2.1b UHBR20, driving displays up to 8K 165Hz.

Considerations:

The new RTX 50 Series cards set a new precedent in the market and at retail price are great value for money. Though as a result, they will be in significant demand, leading to inevitable supply issues. This will, in turn, raise the price of these cards significantly above their retail price. ($1999 for the RTX 5090).

Another factor to consider would be the increased power requirements - these cards draw significantly more power than the previous generation. This may lead to issues integrating these new cards if the infrastructure cannot support them without upgrading cooling and power supply.

To conclude, the NVIDIA RTX 50 series represents a major technological leap forward, offering significant capabilities for a range of gaming and enterprise applications. Its overall performance will make it a valuable asset for organisations working in the rapidly growing industries of AI of HPC.