Temperature curves and clock
The temperature curve of a GPU varies considerably depending on the load profile. In a constant stress test, the GPU remains at full load throughout, which leads to stable heat development, as energy consumption and cooling remain constant. Gaming scenarios, on the other hand, are characterized by changing loads. Technologies such as power gating and separate power rails reduce the activity of unused units, resulting in fluctuating energy consumption and temperature fluctuations. Accelerated Frequency Switching adjusts clock frequency and voltage within microseconds, which optimizes energy consumption but generates irregular heat.
After a few minutes in the temperature target, however, the behavior of the GPU changes fundamentally. Dynamic clock and voltage adjustments now occur with different parameters as cooling and heat dissipation have stabilized. While constant loads offer thermal stability, dynamic scenarios show the GPU’s flexibility to adapt to variable requirements in an energy-efficient manner, as long as they do not end in the thermal limit as with the FE.
Memory temperatures remain constant even during dynamic load changes thanks to the efficient cooling design. A separate power supply via separate power rails ensures a stable power supply, regardless of GPU fluctuations. The central placement of the circuit board in combination with large heat pipes and an optimized cooling fin array ensures even heat dissipation and prevents thermal hotspots in the memory area. While the GPU core exhibits significant temperature fluctuations, the memory remains stable thanks to the constant power supply and optimized air circulation. Even under high load, the memory temperature reaches a maximum of 75 °C, which ensures acceptable performance and a long service life for the memory chips.
The clock rates under full gaming load are at a maximum of 2820 MHz and only reach almost the 2900 MHz mark under minimal load. Our OC special (see link below) shows how to achieve even more. This card here still managed around 3.1 GHz and just under 3000 MHz for memory, as long as you set the fans to 100%. The cooling of the FE leaves hardly any overclocking reserves.
Thermography during load operation
Thermography with the Optris PI 640 enables precise temperature measurements and provides detailed insights into the heat distribution. With a resolution of 640 x 480 pixels and a thermal sensitivity of less than 75 mK, the camera makes even the smallest temperature differences visible. The correct emissivity setting, based on comparative measurements with K-type sensors, ensures accurate results. This allows hotspots and areas with efficient heat dissipation to be reliably identified.
The cooling design ensures reasonably even heat dissipation, which is reflected in the thermography by homogeneous temperature zones. The GPU and voltage converters still exhibit well-controlled temperature peaks, while the memory temperatures remain stable. Thanks to a frame rate of up to 32 Hz, the camera can also capture dynamic temperature changes. The measurements were carried out in three scenarios: in idle mode (see above), in gaming mode and under full load, each after 30 minutes of running time.
Noise emissions (“volume”) and fan speeds
The analysis of the fan curves and audio measurements in an acoustic measurement chamber provides information about the noise development of the GeForce RTX 5070 FE. However, due to the smaller cooling surface compared to some custom models, the fan speeds rise very sharply under load, which is reflected in increased noise development. At almost 2900 rpm, the Founders Edition is already at its limit here.
The noise levels correlate with the temperature curves, which indicates that the fan curves are specifically tuned to achieve a balance between cooling and noise.
However, the cooler is already quite tightly dimensioned, so that the card revs up over time as if there were no tomorrow. At over 41 dB(A) under load, the volume is already above the limit of a still pleasant range. There is quite a lot of noise, especially as the low-frequency component at around 300 Hz is quite unpleasant and can be clearly heard (audio sample below).
The coil beeping phenomenon can no longer be heard under load, but it is there. The coil noises are caused by mechanical vibrations in the induction coils of the VRMs, which are caused by the rapid changes in the current flow. The intensity of the coil beeping thus depends strongly on the load situation of the GPU. Under high loads, especially in applications with very high frame rates (FPS), the current consumption of the GPU increases significantly. In such scenarios, the voltage converters operate at maximum efficiency, which keeps the switching frequencies constantly high and increases the mechanical vibrations in the coils.
Of course, you can also listen to this now and “enjoy” the slightly deeper sound of the card:
- 1 - Introduction and details of the GB205-300
- 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 5 problems, power consumption in theory and practice
- 11 - Load peaks native vs. DLSS4, PSU recommendation
- 12 - Cooler, temperatures, thermography, noise
- 13 - Summary and conclusion
41 Antworten
Kommentar
Lade neue Kommentare
Veteran
Mitglied
Urgestein
Veteran
Veteran
Veteran
Veteran
Veteran
Urgestein
Neuling
Veteran
Urgestein
Urgestein
Urgestein
Urgestein
Urgestein
Urgestein
Urgestein
Urgestein
Alle Kommentare lesen unter igor´sLAB Community →