The Nvidia GeForce RTX 5090: A Next-Gen Leap Fueled by AI
Nvidia's latest flagship, the RTX 5090, ushers in a new era of PC gaming, but its performance gains are less straightforward than previous generations. While raw power shows a significant improvement over the RTX 4090 in benchmarks, real-world gaming performance often reveals a more modest uplift, especially without DLSS 4's Multi-Frame Generation. This AI-powered feature, however, is where the RTX 5090 truly shines, delivering impressive frame rate increases and a glimpse into the future of gaming.
The upgrade's value hinges on your gaming setup and preferences. For those with 4K displays and high refresh rates (240Hz+), the AI-generated frames offer a substantial performance boost. However, for users with lower-resolution monitors, the upgrade may not be cost-effective.
Nvidia GeForce RTX 5090 – Image Gallery
5 Images
RTX 5090 – Specifications and Innovations
Built on Nvidia's Blackwell architecture, the RTX 5090 boasts a significant increase in CUDA cores (21,760 vs. 16,384 on the RTX 4090), resulting in a substantial raw performance boost. This is complemented by increased Tensor Cores (680 vs. 512) and RT Cores (170 vs. 128), further enhancing AI and ray tracing capabilities. The 5th-generation Tensor Cores support FP4 operations, reducing VRAM dependency for AI workloads. The card also utilizes 32GB of GDDR7 VRAM, offering a speed and efficiency upgrade over GDDR6X.
Despite its power, the RTX 5090's 575W power draw is a substantial increase over the RTX 4090, impacting power efficiency. Nvidia's focus here is clearly on raw performance. The shift to a Transformer Neural Network (TNN) for DLSS improves image quality and reduces artifacts.
Multi-Frame Generation, an evolution of DLSS 3's Frame Generation, allows the generation of multiple frames from a single rendered image, drastically boosting frame rates. However, similar to its predecessor, it's most effective when paired with DLSS upscaling and a solid base frame rate (around 60fps).
Purchasing Information
The Nvidia GeForce RTX 5090 launched January 30th, starting at $1,999 for the Founders Edition. Third-party models will likely command higher prices.
Founders Edition Analysis
Despite its 575W power consumption, the RTX 5090 Founders Edition surprisingly features a smaller, dual-slot design. Nvidia achieved this through a redesigned PCB layout and a highly effective cooling solution. Temperatures peaked around 86°C during testing, remaining within safe operating limits. The dual-fan configuration draws air from the bottom and expels it through the top, eliminating the need for rear exhaust vents.
The design maintains a similar aesthetic to previous generations, featuring a silver 'X' design and illuminated GeForce RTX logo. The power connector is a new, angled 12V-2x6 design, supposedly more efficient and secure than the previous generation's 12VHPWR connector. An included adapter converts four 8-pin PCIe power connectors to the new standard.
This compact design makes it suitable for smaller PC builds, unlike its predecessors. However, third-party versions are expected to be larger.
DLSS 4: Addressing the "Fake Frames" Concern
While the RTX 5090's raw performance uplift is notable, its true next-gen advantage lies in DLSS 4's Multi-Frame Generation. This feature, powered by a new AI Management Processor (AMP) core, significantly improves frame generation efficiency and reduces latency. The AMP allows for more efficient task allocation across the GPU.
The new frame generation model is 40% faster and requires 30% less memory than the previous generation. It generates three AI frames per rendered frame, effectively tripling frame rates in supported games. The AMP's Flip Metering algorithm mitigates latency issues. Multi-Frame Generation requires a good base frame rate to avoid significant latency.
At launch, DLSS 4 support will encompass a wide range of games, though testing was limited to beta versions of Cyberpunk 2077 and Star Wars Outlaws. Results were impressive, with substantial frame rate increases at 4K with ray tracing enabled. While minor artifacts were observed, they were largely insignificant.
RTX 5090 – Performance Benchmarks
The RTX 5090's performance varies across games. 3DMark benchmarks show a significant generational leap over the RTX 4090 (up to 42% faster). However, in many games, CPU bottlenecks limit the performance gains, particularly at 4K. The upgrade may not be substantial for users already owning high-end GPUs.
Testing was conducted with a Ryzen 7 9800X3D CPU and the latest drivers. DLSS 4 was disabled for comparative benchmarks. Results varied; some games showed modest improvements (10-25%), while others exhibited near parity with the RTX 4090. The RTX 3090 served as a baseline, highlighting the significant generational leap since 2022.
Specific game benchmarks are detailed in the original article, showcasing both impressive results and instances of CPU bottlenecking. One outlier was Assassin's Creed Mirage, which experienced performance issues potentially attributable to driver bugs.
Nvidia GeForce RTX 5090 – Benchmark Charts
14 Images
Conclusion
The RTX 5090 is undeniably powerful, currently holding the title of the fastest consumer graphics card. However, its performance gains over the RTX 4090 are less dramatic than expected in many current games due to CPU bottlenecks. Its true potential lies in DLSS 4's Multi-Frame Generation, offering a significant frame rate boost for high-end setups. The card is a strong investment for those prioritizing cutting-edge technology and the future of AI-enhanced gaming, but for others, the RTX 4090 remains a compelling choice.