NVIDIA provides details about GeForce RTX4090 Performance Figures, and New Features

NVIDIA shared first concrete performance numbers for the GeForceRTX 4090 GPU in a behind-closed door presentation with the media. due to launch on October 12th. While the public presentation was vague and used only vague reference points, actual FPS numbers were found in this slide. These numbers were based upon 4K resolution tests with DLSSSuper Resolution set to Performance Mode. The processor used was an Intel i9 12900K with 32GB RAM, running on Windows 11 64-bit.

All of the below-listed tests were performed in games that have DLSS 3 support. This highlights the importance and value of this new technology for achieving maximum performance. NVIDIA then divided these results into two brackets: today’s games and tomorrow’s games.

Today’s Games

Tomorrow’s Games

  • Cyberpunk 2077 (Max RT – Overdrive Mode) – 90 FPS
  • NVIDIA Racer (FullRTX) – 80 FPS
  • Justice (FullRT) – 81 FPS
  • Portal RTX – 117 FPS

The GeForce RTX 4090’s performance improvements are more noticeable in games that use advanced ray tracing techniques. For example, games that don’t have RT enabled such as Microsoft Flight Simulator, Warhammer 40,000 and Darktide get only 2X performance boosts. However, the UE5 demo, F1 22 and Cyberpunk2077 offer close to 4X increases.

CD Projekt RED’s game will soon get the Overdrive Mode update, which adds more complicated RT calculations. The GeForce RTX 4090 runs 4X faster when DLSS 3 is enabled. Racer RTX is about 4.5X faster, Justice is close to 5X, Portal RTX is about 6X.

NVIDIA also discussed hardware enhancements that were specifically designed to optimize ray-tracing on the Ada Lovelace architecture. The reason is that the most complex titles are more difficult to ray trace than Battlefield V’s 39 RT operations per pixel. Cyberpunk 2077 Overdrive Mode requires 635 RT operations per pixels, for instance.

Shader Execution Reordering is the first and most important addition. Jonah Alben, NVIDIA’s Senior Vice President for GPU Engineering, explained that rays can’t run the same program in a ray-traced game. This means that they will have to be idle while the first one finishes.

SER is a new stage in the ray tracing pipeline that allows rays to run on the same program to be grouped together. This increases efficiency.

NVIDIA says SER allows significant performance improvements in Cyberpunk2077: Overdrive Mode (+44%) Portal RTX (29%) and Racer RTX (20%)

Displaced Micro-Meshes, a new innovation in the GeForce RTX 4090 as well as the Ada Lovelace architecture overall, is also introduced. It is again targeted at optimizing ray-tracing for geometry. The new RT core of the third generation is capable of processing an optimized BVH. As a result, BVH build performance and storage requirements are improved.

Simplygon and Adobe tools support displaced micro-meshes.

Opacity Micro-Maps also make it easier to help RT cores understand how rays should affect irregular objects. This is achieved by opacity masks, which have predetermined opacity states such as translucent, opaque, and unknown. OMMs are a way to save time and improve performance.

We’ve also sent NVIDIA a few questions about DLSS 3 SER, DMM, OMM, and DMM. Once we hear back, we’ll report.

Products mentioned in this post

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *