Cyberpunk 2077 is powered by REDengine 4, which has been specifically designed to deliver expansive, open worlds with exceptional visual complexity. This powerful engine has been tailored to the requirements of modern games and supports advanced technologies such as ray tracing and pathtracing, which depict the lighting, shadows and reflections in the dystopian world of Night City with impressive realism. The implementation of pathtracing, also known as full ray tracing, enables a complete simulation of light rays. This allows precise calculation of global lighting and physically correct lighting effects, which significantly increase the player’s immersion. However, this technology places immense demands on the hardware, as it requires an extraordinary amount of computing power from the GPU.
To optimize performance despite these computationally intensive effects, REDengine 4 supports innovative upscaling technologies such as DLSS (Deep Learning Super Sampling) and similar methods. These technologies reduce the actual render resolution and scale the image using intelligent algorithms in high quality, enabling a consistently convincing visual display with significantly reduced hardware load. In addition, the engine enables the integration of further optimizations, such as dynamic adjustments to the level of detail, to ensure smooth frame rates even in demanding scenes. This makes Cyberpunk 2077 a technical showcase game that combines both visual excellence and technological innovation.
Other REDengine 4 techniques include physically based rendering (PBR) for realistic material properties, dynamic weather and day-night cycles, and advanced AI for believable NPC interactions and traffic flows. In combination with an impressive variety of geometry and textures, Cyberpunk 2077 sets visual standards in modern games. The Cyberpunk 2077 test was conducted to compare performance and image quality under different conditions: native Ultra HD or WQHD resolution, DLSS 3 with frame generation and DLSS 4 with multi-frame generation (MFG). The latest features of the RTX Blackwell architecture were examined, especially in combination with ray tracing and path tracing.
Native vs. DLSS 3 with frame generation and DLSS 4 with multi-frame generation
n native Ultra HD resolution (UHD, 3840 x 2160 pixels), the GPU is heavily used, especially when ray tracing or patch tracing is activated, which ensures realistic light and shadow calculations. These technologies require immense computing power, which often leads to low frame rates. DLSS 3, which was widely used in the previous generation, offered a significant improvement here by upscaling the render resolution AI-supported and generating additional frames between regular rendered images. By using a neural network that analyzed motion vectors and optical flow, DLSS 3 was able to significantly increase the frame rate. This enabled a smoother display, but the image quality occasionally showed slight deviations from the native display in scenes with fast movements or complex details.
With the introduction of DLSS 4, this technology has been further developed and taken to a new level. The central innovation is the multi-frame generation (MFG), which can generate up to three additional frames per regularly rendered image. This is made possible by the improved fifth-generation tensor cores and optimized transformer models. DLSS 4 not only reduces latency and memory requirements, but also improves frame pacing, resulting in a smoother and more harmonious rendering. In addition, DLSS 4 integrates advanced technologies such as Ray Reconstruction, which selectively restores details in ray-traced scenes, providing picture quality that often exceeds native UHD display.
Individual metrics for WQHD and Ultra HD
Individual metrics such as percentile frame time, variances, latencies, power consumption and efficiency play a central role in evaluating the performance of graphics cards in demanding scenarios such as Cyberpunk 2077 with ray tracing. The percentile frame time provides information about the worst frame times and is an indicator of disruptive drops in performance. A low and consistent frame time indicates smooth gameplay. Variations in frame times, on the other hand, show fluctuations that can lead to micro-stutters, even if the average frame rate is high. Latencies are particularly relevant for reactivity in the game. Low system latencies, supported by technologies such as DLSS frame and multi-frame generation, ensure a direct gaming experience, even at high frame rates.
Power consumption increases significantly in scenarios with active ray tracing and high resolution, especially with native display. DLSS 3 and 4 can achieve efficiency gains here by reducing the GPU load. Efficiency, defined as watts per frame, is a critical comparative value as it shows how well a card converts performance in relation to energy. NVIDIA cards like the RTX 5090 benefit from optimized hardware for AI-based rendering technologies and show above-average efficiency. Together, all of these metrics provide a much more comprehensive picture of performance that goes beyond pure averages and allows for a more precise assessment of gaming and system stability.
Differences between the rendering methods
The comparison of the frame time curve and the actual power consumption over the entire benchmark process is particularly revealing, as it enables an in-depth analysis of the performance under different rendering methods. This visualization reveals not only the average system load, but also dynamic fluctuations and peaks in real time. At native resolution, the frame-time curve typically shows completely different fluctuations and higher power consumption, as the GPU carries the entire load of the rendering tasks. The effect of the FG or MFG, on the other hand, is reflected in a more irregular power consumption, which can quickly reach the limits of the hardware at peak times. In extreme cases, such peaks can also lead to unstable frame rates and micro-stutters. Multi-frame generation (MFG) can lead to short-term high load peaks due to the complex calculation of several frames from a single rendered image, especially in dynamic scenes with complex ray tracing, which I will examine in detail later on.
Frame generation (FG) reduces the load on the GPU in theory, of course, as additional frames are generated without the GPU having to render all pixels traditionally. The frame-time curve becomes smoother. Nevertheless, load peaks can occur, for example in the event of sudden changes in the scene or intensive ray tracing calculations. Multi-frame generation (MFG) with DLSS 4 shows an even smoother frame time curve, as several frames are generated from a single rendered frame. The number of load peaks changes and the power consumption remains somewhat more constant over the entire benchmark run. This method not only optimizes the GPU load, but also improves efficiency as fewer compute cycles are required for the same or better visual quality.
The parallel display of the two curves shows how the different methods affect frame rate stability and energy efficiency. It shows how DLSS technologies help to minimize load peaks by smoothing the GPU load while enabling consistent power consumption. This is particularly important for energy-efficient systems or scenarios where thermal limits need to be considered:
This type of analysis not only enables an objective evaluation of performance, but also provides valuable insights for the optimization of hardware and software in real game scenarios.
Interim conclusion
The RTX 5080 showed solid but limited frame rates in tests both with and without frame generation in native resolution with ray tracing enabled. DLSS 3 was able to significantly increase performance, but its frame generation did not achieve the stability and image quality made possible by the further development in DLSS 4. With the new iteration, which integrates multi-frame generation (MFG), DLSS 4 not only delivers the highest frame rate, but also the most consistent image quality. This is particularly evident in combination with pathtracing, which benefits greatly from the enhanced ray tracing capabilities of the RTX Blackwell GPUs. The results impressively show how crucial AI-supported technologies are for the representation of complex graphical scenarios, especially in graphically demanding titles such as Cyberpunk 2077.
For an in-depth analysis of the technical aspects and performance of Cyberpunk 2077, Fritz Hunter’s follow-up is an excellent addition. This follow-up contains high-resolution video comparisons and precise depictions of the different graphics modes, including DLSS and pathtracing implementations. It also includes detailed research on frametimes and framepacing, as well as a comprehensive latency analysis that shows how different settings and technologies affect in-game response times. These comprehensive insights are particularly valuable for enthusiasts who value both maximum image quality and optimized performance.
- 1 - Introduction and details of the Blackwell GB203-400-A1 GPU
- 2 - Test system and equipment
- 3 - Gaming: Full-HD 1920x1080 Pixels (Rasterization Only)
- 4 - Gaming: WQHD 2560x1440 Pixels (Rasterization Only)
- 5 - Gaming: Ultra-HD 3840x2160 Pixels (Rasterization Only)
- 6 - Gaming: WQHD 2560x1440 Pixels, Supersampling, RT & FG
- 7 - Gaming: Ultra-HD 3840x2160 Pixels, Supersampling, RT & FG
- 8 - DLSS4 and MFG: Cyberpunk 2077 in detail
- 9 - DLSS4 and MFG: Alan Wake 2 in detail
- 10 - PCIe Gen5 problems, power consumption and standards
- 11 - Load peaks and power supply recommendation
- 12 - Cooler, temperatures, thermography, noise development
- 13 - Summary and conclusion
181 Antworten
Kommentar
Lade neue Kommentare
Mitglied
Urgestein
Veteran
Urgestein
Veteran
Mitglied
Veteran
Veteran
Veteran
Veteran
Veteran
Urgestein
Mitglied
Veteran
Veteran
Urgestein
Urgestein
Urgestein
Urgestein
Alle Kommentare lesen unter igor´sLAB Community →