In QHD (2560 x 1440 pixels), frame rates can be significantly increased by using DLSS (Deep Learning Super Sampling) and FSR (FidelityFX Super Resolution), even when sophisticated technologies such as ray tracing are activated. Both processes use intelligent upscaling methods to maintain high image quality while reducing render resolution, thereby reducing the load on the GPU. DLSS uses AI-powered supersampling to convert low-resolution images into high-resolution representations through neural networks. It also offers frame generation for supported games, where additional frames are generated to increase the perceived fluidity of the game. FSR pursues a similar goal, but is based on open standards and works on both AMD and Nvidia cards. Both technologies significantly reduce the render load and thus also make ray tracing in QHD playable more smoothly.
The tests were carried out without DLSS4 and MFG, but the simple frame generation was used for NVIDIA and AMD cards where available. Nvidia graphics cards, including the RTX 5090, also show strong performance with patch tracing enabled, which enables particularly realistic lighting simulations. This allows Nvidia cards to continue to deliver outstanding performance in scenarios with complex lighting effects, while AMD cards rely on classic optimizations. The choice between DLSS and FSR depends on the GPU used and the support provided by the respective game.
Summary and overview of all games in WQHD with super sampling, ray tracing and frame generation
Determining an average value from the cumulative FPS values of 11 games also serves to provide a general assessment of a system’s gaming performance in Ultra HD. The frame rates of the individual games are added together and divided by the number of games to calculate a simple average value. However, this average value is only meaningful to a limited extent, as it does not take into account the specific characteristics and requirements of the individual games. Games vary greatly in their load on the CPU, GPU and memory, which can lead to different performance results. For example, a particularly CPU-intensive game can influence the average value downwards, while a less demanding game could artificially increase the values. As a result, this approach only provides a rough guide and is not suitable for detailed analysis of hardware performance in different scenarios.
Normalizing FPS values across different games serves to objectively and comparably represent the performance of a hardware. As different games place varying demands on the CPU and GPU, raw FPS values are not always directly comparable. For normalization, the FPS values of all tested games are brought to a uniform scale and only then calculated as an average value. This approach allows for a balanced assessment of hardware performance that is not distorted by the specific characteristics of individual titles. I use this metric for both the average FPS and the P1 Low, i.e. the percentile with the slowest rendered frames. The GeForce RTX 5080 takes over 10 percent off the RTX 4080 Super.
The P1 Low is a measured value that is used in benchmarks to evaluate the frame rate (FPS) and describes the lowest value that the top 99% of the measured FPS do not fall below. In contrast to the average value, which summarizes the general performance of a system, the P1 Low focuses on the analysis of performance drops, which can be particularly disruptive during gaming. The calculation is performed by sorting the FPS readings according to their size. The highest value of the lower 1% is then specified as P1 Low. This isolates serious frame drops that occur rarely but can significantly impair the perceived fluidity and stability of a game. The P1 Low value is therefore an important measure for evaluating the stability of the frame rate in addition to the average performance. Problems such as micro-stutters or irregular frame distributions, which could remain undetected with a purely average view, are made visible by this key figure. In current tests, the GeForce RTX 5080 outperforms the RTX 4080 Super in this area by more than 8 %, which indicates significantly better stability and lower frame drops.
The power consumption of a system in QHD with activated technologies such as DLSS 3 or FSR is largely determined by the GPU load, especially when these technologies are combined with ray tracing. Intelligent upscaling and frame generation reduce the number of pixels actually calculated, which can significantly reduce the GPU load and thus the energy requirement compared to the native render resolution with ray tracing. This generally leads to a noticeable reduction in power consumption, although the effect varies depending on the game and implementation. Nvidia graphics cards benefit in particular, as they are able to process patch tracing with high efficiency.
However, activating ray tracing increases the hardware requirements considerably, as the simulation of realistic light and shadow effects requires a large number of additional GPU computing units. Despite this additional load, the upscaling optimizations result in an overall lower power consumption compared to pure raster graphics rendering in WQHD. The use of frame generation, if supported by Nvidia or AMD cards, only has a minimal impact on energy consumption. As the calculation of additional frames takes place on specially optimized hardware units, the GPU is not extensively loaded, which underlines the efficiency of this function. The following measured values for GPU and CPU power consumption illustrate in which scenarios which component is operated under maximum load and how the various technologies affect the energy efficiency of the system.
The efficiency is then the counter calculation, so to speak, where you compare the amount of watts used (average over the respective game) and the average FPS of the game in question:
Individual metrics and details
Of course, as always, there are also the individual metrics for all 11 games tested:
- 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 →