Nvidia's RTX 50 Series: A Sideways Look at the Next-Generation Graphics Card Design and Performance
Nvidia's RTX 5000 series GPUs, based on the Blackwell architecture, have officially arrived, marking a significant leap in graphics technology. The new series, announced at CES 2025, promises enhanced performance, improved ray tracing, and advancements in AI-driven features.
Architecture and Key Features
The RTX 5000 series is powered by Nvidia's Blackwell GPU architecture, a significant overhaul from the previous Ada Lovelace architecture used in the RTX 4000 series. The Blackwell architecture brings several enhancements, including a new unified cache design for increased memory efficiency and bandwidth, and redesigned Tensor Cores that triple performance in AI workloads. Ray tracing cores have also been upgraded to deliver twice the throughput.
The RTX 5090, the series' flagship, features the GB202 chip with 92.2 billion transistors and a die area of 750 mm². It contains 192 SM (Streaming Multiprocessor) blocks, totaling 24,576 shaders, 192 RT cores, and 768 Tensor Cores. The RTX 5080 and 5070 Ti are expected to use the GB203 chip.
A key upgrade in the RTX 5000 series is the adoption of GDDR7 memory. The RTX 5090 boasts 32GB of GDDR7 VRAM, a 512-bit memory bus, and 1792 GB/s bandwidth. GDDR7 memory significantly enhances data transfer speeds, contributing to the cards' overall performance boost. The series also supports PCI Express Gen 5.0 for fast data transfer speeds and DisplayPort 2.1 for higher bandwidth, refresh rate, and resolution support.
Models and Specifications
The RTX 5000 series includes several models catering to different performance tiers and budgets:
- RTX 5090: The flagship model designed for 8K gaming and AI-heavy workloads.
- RTX 5080: A high-end option suitable for serious 4K gamers.
- RTX 5070 Ti: A mid-to-high-tier card offering excellent price-to-performance.
- RTX 5070: An excellent choice for 1440p gaming and entry-level 4K.
- RTX 5060 Ti: A solid option for budget-conscious gamers.
Specific specifications for the RTX 5090 include 18,432 CUDA cores, a base clock of 2.2 GHz, a boost clock of 2.9 GHz, 32GB of GDDR7 memory, a 512-bit memory bus, and a TDP of 480W. The RTX 5080 features 10,752 CUDA cores and 16 GB of GDDR7 memory.
Performance and Expectations
Early performance benchmarks indicate significant performance gains compared to the RTX 4000 series. The RTX 5090 shows substantial improvements in gaming scenarios, with Cyberpunk 2077 (RT Ultra) seeing a jump from 57 FPS on the 4090 to 98 FPS on the 5090. Call of Duty: Modern Warfare III runs at 195 FPS on the 5090, compared to 118 FPS on the 4090, and Flight Simulator 2024 hits 87 FPS on the 5090 versus 54 FPS on the 4090. Nvidia claims the RTX 5070 can achieve RTX 4090 performance through AI and DLSS 4.
Nvidia's Deep Learning Super Sampling (DLSS) technology continues to evolve, with DLSS 4 introducing Multi Frame Generation. This new version generates two frames instead of one, further boosting frame rates and visual fidelity. The RTX 5000 series also demonstrates improvements in power efficiency. Blackwell chips intelligently distribute the load and throttle down during less demanding tasks, keeping temperatures under control.
Pricing and Availability
The RTX 5090 is priced at $1,999. The RTX 5080 has a MSRP of $999, while the RTX 5070 and 5070 Ti are priced at $549 and $749, respectively. The RTX 5090 and 5080 were released first on January 30, 2025, followed by the RTX 5070 Ti on February 20, 2025, and the RTX 5070 on March 5, 2025.
