Is Memory to Blame?
Tom’s Hardware Germany analyzed its infrared images of the GTX 2080 Ti reference card to investigate rumors that Micron‘s GDDR6 packages are overheating, thus causing the errors. Thermal measurements indicate that the M6 and M7 GDDR6 modules run well over 90C under some conditions. The modules are located directly over internal power supply tracks embedded in the PCB. These tracks run between the PWM nodes and the GPU socket. The memory modules could run hotter than the 96.1C measurement shown in our image due to the high currents flowing through the power tracks embedded in the PCB and heat migration from the VRMs. Also, note the close proximity of the hotspot.
We don’t know many of the technical details surrounding Micron‘s GDDR6 packages, but we do know they have a maximum safe operating temperature of 95C that we have obviously exceeded in our German lab. In either case, the symptoms of the failures also seem to support the theory that overheated memory is to blame. Most readers have complained of failures after the cards had been used for some period of time, and in some cases, the cards even work correctly after a cooling-off period. The problems also appear to be more prevalent in partner cards that use cheaper cooling solutions.
https://www.tomshardware.com/news/rtx-2080-ti-gpu-defects-launch,37995.html
唔好超張卡太勁住啦,除非上custom loop