Heatsink test with external K-type sensor
Of course, the question now arises as to whether the modules with their foam “thermal pads” will remain stable at all over a longer period of time in the stress test. So I loaded the most aggressive XMP profile with 6400 Mbps, tCL 32 and 1.4 V and started Karhu to force the modules to their thermal limit.
Since the temperature sensor of the modules, which can be read out in software tools such as HWinfo, is located in the SPD chip and this is actually cooled by the real thermal pad of the PMIC area, I did not want to rely on this measured value. It is possible that the temperature at the PMIC and SPD chip is quite comfortable, but the ICs a few centimeters next to it are actually on the verge of boiling. To investigate this, I use a K-type temperature sensor with the Elmorlabs KTH thermometer. The sensor is placed directly between one of the memory chips and the foam layer. I deliberately use one of the theoretically hottest ICs, which is on the side with the thicker foam layer, is surrounded by other ICs and also has the other module opposite it. The passive air flow from the neighboring radiator is also sealed off. Burn, baby, burn!
After a good hour in Karhu, the modules have reached their maximum temperature and 63 °C can be read out on the SPD via HWinfo. The external type K sensor measured 67 °C. And that with the XMP profile with 1.4 V VDD/VDDQ and the most unfavorable conditions possible. Now you could perhaps add roughly 10 °C for the potential sauna environment in a housing with a heat pump a la RTX 5090, but even then we are still well away from the maximum temperature of 85 or 95 °C specified for DDR5 ICs according to JEDEC.
Yes, the cooling solution for these Viper Venom modules from Patriot could hardly be more suboptimal. Perhaps no heat sink at all would even be better in the end. But it must also be soberly acknowledged that stable XMP operation is guaranteed, as intended by the factory. No errors were detected in the stress test and with the measured temperatures we are far away from possible performance regressions, let alone hardware degradation. In this respect, there are significantly worse RAM kits on the market.
52 Antworten
Kommentar
Lade neue Kommentare
Urgestein
Urgestein
Urgestein
Veteran
Urgestein
Urgestein
Urgestein
Urgestein
Urgestein
Urgestein
Veteran
Veteran
Urgestein
Urgestein
Veteran
Urgestein
Mitglied
Urgestein
Mitglied
Alle Kommentare lesen unter igor´sLAB Community →