NVIDIA GeForce RTX 5090 Founders Edition Review – The 600-Watt Powerhouse in Gaming and Lab Tests

1 - Details zur Blackwell GB202 GPU 2 - DLSS4 einfach und im Detail erklärt 3 - Neurale Shader als echte Game-Changer? 4 - Pathtracing: Grundlagen und Verbesserungen mit Benchmarks 5 - Testsystem und Equipment 6 - Gaming: Full-HD 1920x1080 Pixels (Rasterization Only) 7 - Gaming: WQHD 2560x1440 Pixels (Rasterization Only) 8 - Gaming: Ultra-HD 3840x2160 Pixels (Rasterization Only) 9 - Gaming: WQHD 2560x1440 Pixels, Supersampling, RT & FG 10 - Gaming: Ultra-HD 3840x2160 Pixels, Supersampling, RT & FG 11 - DLSS4 und MFG: Cyberpunk 2077 im Detail 12 - DLSS4 und MFG: Alan Wake 2 im Detail 13 - PCIe 5 Probleme, Leistungsaufnahme in Theorie und Praxis 14 - Lastspitzen nativ vs. DLSS4, Netzteilempfehlung 15 - Kühler, Temperaturen, Thermografie, Geräuschentwicklung 16 - Zusammenfassung und Fazit

Detailed view of gaming in Ultra HD: Native vs. DLSS and MFG X4

In Cyberpunk 2077, the graphics card requires peak performance of up to 600 watts at native resolution in UHD and extreme settings. These values are caused in particular by the enormous demands on the GPU’s computing power when high resolutions and maximum details are used without the support of AI-based scaling or efficiency technologies. Such peak loads require a reliable power supply, as they not only demand the power reserves of the 12V2X6 design, but also place high demands on the stability of the power supply units.

The next two graphs offer a high-resolution analysis of a single one of the previously considered 20 ms intervals, this time resolved in 10 µs steps. This level of detail allows a precise insight into the behavior of the power supply during short-term load changes, which in practice are triggered by sudden GPU requirements, for example during render spikes or frame changes. The first graph shows the power consumption during this extremely short period. It reveals the subtle fluctuations and peaks that are typical for the operation of modern high-performance graphics cards. Such peaks can be many times higher than the average value in the short term and place high demands on the response speed and stability of the power supply unit. We measure up to 900 watts here and get a little fright.

The second graph shows the current flowing through the power supply cables at these 10 µs intervals. The abrupt changes that occur with highly dynamic loads become particularly clear here. These fine details are crucial to understanding how short-term load changes are compensated for by the power supply architecture and what impact they can have on overall system stability. These measurements also clarify the meaning of the ATX 3.1 standard, in particular the requirement for 200% power reserve at load peaks in the range of 1 millisecond. As the graphics card has very high performance requirements in peak load situations, it is essential that the power supply unit not only remains stable in the short term, but also has sufficient reserves to absorb such fluctuations without causing voltage dips.

By using DLSS 4 in conjunction with MFG x4 (Multi-Frame Generation), this energy requirement can be significantly reduced. DLSS 4 uses AI algorithms to scale the image quality efficiently, which means that the GPU has to use less raw power to simulate native UHD quality. At the same time, MFG x4 significantly increases the frame rate by generating additional frames based on previous images. In practice, this leads to a lower average power consumption while simultaneously improving the gaming experience through smoother image playback. The available measurement data illustrates this clearly. The first image shows the power consumption in intervals of 20 milliseconds, a critical period for power supply units, as supervisor chips in power supply units have to recognize and process peak loads in this time frame.

The second graph shows the currents flowing through the power supply cables. This data underlines the importance of even and stable power distribution in order to protect the hardware components and minimize efficiency losses.

The reduction in power consumption through DLSS and MFG impressively demonstrates how modern technologies can not only improve energy efficiency, but also reduce the load on the power supply components. This not only contributes to the longevity of the hardware, but also reduces operating costs and the ecological footprint of the system. But again, I have the high-resolution measurements with the 10µS intervals. First the load peaks in the power consumption…

… and now the currents:

Load behavior in the Torture Test

Furmark is often used as an extreme load test for graphics cards, as it pushes the GPU to an atypically constant and maximum load that far exceeds what occurs in most real-world applications or games. The software uses intensive computation of graphics processes that fully utilizes both the GPU’s shader and memory controllers. This creates a thermal and electrical load that serves as a worst-case scenario to test the stability and performance of the graphics card and power supply. In this test, the power consumption of the card can even exceed the specified TDP of 575 watts and reach peaks of up to 600 watts. This is because the TDP of a graphics card describes the average thermal power dissipation under normal operating conditions, while Furmark generates a maximum and sustained load that pushes the limits of power consumption. Possible short-term load peaks are also included in these values.

The test is particularly relevant for checking the power supply components of the system, such as the power supply unit and the voltage converters, for their load capacity and efficiency. Since Furmark generates an almost unrealistically high load, it is not considered representative for everyday usage scenarios, but it is extremely useful for uncovering possible weak points in the hardware or cooling. The fact that the card reaches the 600 watt mark also shows how important it is that the power supply unit is sufficiently dimensioned, as it must not only cover the nominal TDP, but also offer sufficient reserves for such peak loads. Furmark thus serves as a kind of stress test that ensures that the entire system remains stable even in extreme situations.

The high-resolution measurements during a Furmark test provide crucial insights into the behavior of the power supply and the power consumption of the graphics card in extreme situations. Furmark generates a constant thermal and electrical load through its continuous maximum utilization of the GPU, which can be analyzed with the highest precision in these measurements. Data recorded at intervals of just a few microseconds show how the power consumption and currents vary over time. Particularly noticeable are the short-term load peaks that become visible in the measurements and can significantly exceed the average power consumption. Such peaks are caused by rapid changes in the utilization of individual GPU components, such as shader arrays or memory controllers, which Furmark specifically stresses.

These measurements are particularly relevant in the context of the ATX 3.1 standard, which requires power supply units to be able to compensate for short-term peaks of up to 200% of the nominal load for a duration of up to 1 millisecond. The high-resolution data clearly shows that such peaks in Furmark are not just theoretical considerations, but actually occur and can sometimes even come critically close to the limits of the power supply designs.

Summary of the load peaks and a power supply recommendation

A power supply unit with a rated output of 1000 watts that meets the requirements of the ATX 3.1 standard is a suitable choice for reliably covering the power consumption values and load scenarios described. The maximum peak loads of the graphics card, which can reach up to 600 watts in extreme situations such as Furmark or very demanding games, make a high power reserve necessary. Together with the load of the rest of the system, such as the CPU, RAM and other components, this results in a requirement that can reach up to 900 watts at peak times.

A power supply unit with 1000 watts not only offers sufficient scope in such scenarios, but also enables short-term load peaks to be absorbed, which are covered by the ATX 3.1 standard. This standard requires the ability to cope with short-term power requirements of up to 200% of the nominal load for a duration of up to one millisecond. The selected power supply can therefore also handle peak loads of up to 2000 watts without stability problems or voltage dips.

The dimensioning to 1000 watts also ensures that the power supply works in an efficient load range in typical operating scenarios, ideally between 50 and 70 % of the nominal power. This optimizes both the energy efficiency and the longevity of the power supply unit. 80 PLUS Platinum or Titanium certification also ensures high efficiency and low heat generation. Another advantage of this choice is that it is future-proof. By supporting modern standards such as the 12V2X6 power design, the power supply is equipped for current and future high-performance graphics cards. It offers sufficient capacity for future hardware upgrades without the need for replacement and ensures long-term stability and reliability even under demanding conditions.