Introduction

Last week, at CES Las Vegas, NVIDIA officially introduced its highly anticipated GeForce RTX 50 series, marking a significant leap in GPU performance and efficiency for both desktop and laptop systems. The lineup, powered by the company's new Blackwell architecture, promises transformative advancements across gaming, content creation, and AI-driven tasks. With innovations like fourth-generation RT cores, advanced neural shaders, and new GDDR7 memory, NVIDIA is aiming to redefine the user experience for both gamers and professionals alike.



During NVIDIA's Editor's Day at CES, the company provided a detailed overview of the RTX 50 series, highlighting its architectural breakthroughs and their real-world applications. In addition to unveiling the hardware, NVIDIA showcased the capabilities of DLSS 4, the latest version of their deep learning super sampling technology. This update introduces multi-frame generation and transformer-based models, improving framerates and image quality. The presentation also explored the company's AI initiatives, demonstrating how NVIDIA's advancements extend beyond gaming to support a variety of applications in productivity, research, and content creation.

In this article we will go over the content presented by NVIDIA at their Editor's Day during CES, including performance enhancements offered by the new GPUs, the improved thermal designs, and the broader implications of DLSS 4. We'll also cover the upgrades tailored for creators, such as better generative AI performance and accelerated video editing workflows. Finally, NVIDIA offered insights into evaluating the gaming experience, emphasizing metrics like latency and image quality to give a more comprehensive view than just frames per second.

The Cards

At CES Las Vegas, NVIDIA announced four new desktop GPUs in the RTX 50 series: the RTX 5070, 5070 Ti, 5080, and the flagship 5090.

RTX 5090

The RTX 5090, in particular, stands out as an incredibly powerful option for gamers. Thanks to advancements in neural rendering and DLSS, the 5090 can handle demanding games smoothly, achieving amazing FPS rates (with upscaling and framegen). The GPU is equipped with a massive 32 GB of GDDR7 memory, the fastest memory available, running across a 512-bit memory interface. This provides an astounding 1.8 terabytes per second of memory bandwidth. All Blackwell GPUs support PCI-Express 5.0.

The number of GPU cores is 21,760, which is a +33% increase vs RTX 4090, which has 16,384 cores. The memory bus on RTX 4090 was 384-bit, so the 512-bit on RTX 5090 will result in additional performance improvements.


NVIDIA also highlighted the enhanced efficiency brought by DLSS 4. With DLSS on, gamers not only enjoy improved frame rates but also reduced latency, when considering absolute milliseconds. In direct comparisons with previous-generation models, the RTX 5090 shows 30-40% performance increase without DLSS 4, and more than +100% when multi-frame-gen is enabled. Do wait for our reviews though.


The thermal design of the RTX 5090 has been completely revamped to improve airflow and reduce noise. The GPU features a dual airflow design, which enhances cooling efficiency and reduces fan noise, even at high power levels. This new design is made possible by a smaller, more compact PCB, which allows for better airflow and greater overall system efficiency. As far as I know, the FE is the only card with such a small PCB and the custom designs from the various board partners use more traditional designs.