‘Defective batch of AMD Radeon RX 7900 cards suffers from high GPU temperatures’
The high GPU temperatures experienced by some Radeon RX 7900 reference cards may be caused by a defective vapor chamber in one or more video card batches. An OEM reports this to Igor’s Lab. The temperature problems came to light last week.
Sources from a vapor chamber manufacturer state that the temperature problems are caused by a defect in the cooler, writes Igor Wallossek of Igor’s Lab. At least one batch of reference cards is said to have left the factory with too little liquid in that vapor chamber, resulting in high hotspot temperatures. That would also explain why the temperatures differ between horizontal and vertical mounting of the video card, Wallossek writes. The condensate in the cooler cannot rise if the GPU is mounted horizontally.
The temperature issues with certain Radeon RX 7900 XT and XTX reference cards came to light in the past week. In some cases, the video cards reached high hotspot temperatures of up to 110 degrees Celsius, which leads to thermal throttling.
The vapor chamber in the AMD Radeon RX 7900. Source: der8auer, via YouTube
Professional overclocker Roman Hartung, also known as der8auer, investigated the problem last week. Hartung tested four different Radeon RX 7900 video cards in vertical and horizontal positions. The overclocker then also came to the conclusion that the high temperatures were caused by a problem with the vapor chamber. Hartung now has opened the vapor chamber, but no design errors found. This makes Igor Wallossek’s statement plausible, although it has not yet been officially confirmed.
AMD already acknowledged last week that certain RX 7900 reference cards suffer from high hotspot temperatures. The company advised users a statement to Tom’s Hardware to contact its customer service. AMD has not yet addressed the cause of the problem and has not yet responded to Hartung and Wallossek’s findings. Not all RX 7900 reference cards seem to suffer from high hotspot temperatures. It is not known exactly how large-scale the problem is.