Thermocouple Calibration

The first step is to address the basics by checking the thermocouples for any temperature deviations, which provides an initial assessment of accuracy. Type-K thermocouples from certain vendors are notorious for their temperature differences, so we made sure to test for that before recording any actual measurements. While this article is not a scientific study, efforts were made to minimize variables. Conducting an equipment test before an experiment is essential to ensure reliable and accurate results. Properly functioning equipment reduces the risk of collecting flawed data. By confirming that all thermocouples are operating correctly, we can uphold the integrity of the experiment and ensure that the data gathered is both valid and reproducible.



Here, we are examining whether each thermocouple reaches boiling (100°C) at approximately the same temperature. The largest deviation observed is between 1°C and 1.5°C from the target, making Probe #4 (used for heat spreader) the most significant outlier.

Logging Setup

*Probe #3 and #2 are the same Thermal Couple



Due to the wide variety of DRAM heatspreaders available from different memory vendors, the decision was made to focus this article around DRAM temperatures rather than a specific brand by using third party heatspreaders. In many ways, these results represent a best-case scenario: the heat spreader is thick, allowing for effective thermal soak, and provides full front and back PCB coverage to maximize heat dissipation. Using the data collected here, further articles can be created to explore specific questions about the strengths and design limitations of individual memory vendors' heatspreaders.

Thermal Probe Location Test

We start out by trying to replicate Igor's Lab test. Unfortunately due to that article using an infrared camera, the results cannot be matched. Igors' article results showed 57 °C for the SPD Sensor and 65°C for the DRAM package, While we measured 43.5°C and 43.7°C respectively using thermal couples. In this instance thermal couples and the infrared camera data from Igors article cannot be directly compared. With that known, a baseline for this article has been set.

With Heat Spreader

With the heat spreader attached and probes connected, we can now compare the SPD sensor reading to the actual DRAM temperature. A small discrepancy in temperature arises because the SPD sensor is attached to the PCB and getting the reading from that specific spot. Due to the heat spreader's full coverage and high thermal dissipation, the temperature difference between the PCB and the DRAM ICs remains minimal in this test.

DRAM Frequency vs Temperature

With a baseline established and confirmation that the SPD sensor measures the PCB temperature only, we can now shift our attention to DRAM IC temperatures and what factors are most influential. The first variable to examine is placing a load on the memory at a frequency and fixed voltage. As the frequency increases, a corresponding rise in DRAM temperature is observed.

DRAM Voltage vs Temperature

The next test aims to evaluate the impact of voltage alone on memory temperatures. Using 2800 MHz (DDR5-5600) as the set frequency. The chart shows a temperature difference of 3.3°C between 1.25 V and 1.50 V at the 30-minute mark, increasing to 4.3°C by the end of the 1-hour test. This is fairly minor on its own, but will compound with a memory frequency.