We conducted a series of tests on an Asus TUF RTX 5090, to check its load balancing, using various benchmarks and a game (Cyberpunk). With Powenetics PMD conducted, we didn’t find any load balancing issues, just huge power consumption readings, which weren’t inline with what the Powenetics reports. Please note that Powenetics uses current shunt resistors immune to EMI and in line with the power lines, so they are way more accurate and affordable than current clamps.
On the other hand, the current clamps I used hooked on a scope, can offer much higher data poll rates, which are essential to check load balancing. After all, in this test, I don’t care much about accurate power readings, but how amperage is distributed on the 12V-2×6 cable’s gauges.
Some claim that PMDs like Powenetics can affect and improve load balancing, so in the next series of tests that I will conduct, I will remove Powenetics and have the cable directly connected to the power supply. I hope I had time to run this test sooner, but this is a pretty tough period for me, in terms of work (and not only).
The topic I want to develop in this article is the following. During the tests I conducted for the needs of this video, I noticed that the GPU’s header and especially the 12V-2×6 cable near the GPU’s end got way hotter at a lighter benchmark (V-Ray), compared to heavier benchmarks like Furmark and 3D Mark Speed Way. This troubled me till I figured a possible explanation.
The GPU’s header and lower-end of the 12V-2×6 cable are right beside/below the GPU’s heatsink, so they come in contact with the airflow. This means the GPU’s fans assist in their cooling, so in heavy benchmarks where the GPU’s power consumption is 600W sustained and the thermal loads are high, fan speed is also high, assisting in the 12V-2×6 header and cable cooling. On the other hand, when the GPU’s power consumption is lower than 600W but still high enough (say 450W), the fans don’t spin at high speeds, hence the 12V-2×6 header and cable are not cooled optimally, leading to the higher temperatures I noticed in V-ray.
Please note that the tests were conducted at a low ambient temperature, 21°C. However, inside a chassis with an RTX 5090 monster and a strong CPU, the ambient temperature can easily exceed 40°C, so the temperatures I present can be 20°C higher. To make matters worse, the graph above shows the cable’s 12V-2×6 header (remember that in the previous temperature graph, I had the temperature probe a little above the connector on the 12V gauges). Look how high the temperature went in 3D Mark Speed Way, at 21°C ambient after just 30 minutes. If the connector on the 12V-2×6 cables reaches 64°C at just 21°C ambient, at over 40°C ambient it can go close to 90°C with the limit for this connector being set at 105°C. And remember from my previous tests, that 3DMark Speed Way and Furmark are not the worst case scenarios! Because the GPU’s fans help remove some of the heat in this area.
More research is required, and I need to find the time to do it. I will be even more thorough in the next set of test sessions!
Good stuff, keep up the good work!