Earlier this month, NVIDIA released it GeForce RTX Super series graphics cards that are designed to go up against the Radeon RX 5700-series "Navi" graphics cards. Since then, AMD has launched their Radeon RX 5700 and RX 5700 XT, which we thoroughly reviewed. NVIDIA designed the RTX 2060 Super to beat the RX 5700 without stepping on the RTX 2070's toes, while the RTX 2070 Super uses the larger 13.6 billion-transistor "TU104" silicon to compete with the RX 5700 XT.
NVIDIA carved the original RTX 2060 out of the "TU106" silicon by disabling a quarter of its memory subsystem and enabling just 1,920 out of the 2,304 CUDA cores available on the silicon. The original RTX 2070 maxes out this chip. The new RTX 2060 Super has the full 256-bit wide GDDR6 memory bus with which it talks to 8 GB of memory. NVIDIA also upped the CUDA core count to 2,176, just a TPC short of the RTX 2070. The RTX 2070 Super is carved out of the larger "TU104" silicon by enabling 2,560 out of the chip's 3,072 CUDA cores, while leaving its memory subsystem untouched.
During our reviews of the original RTX 2060, we noticed that almost all PCBs had pads for 8 memory chips, which leads us to believe that the RTX 2060 Super was part of NVIDIA's plans all along. It wouldn't surprise us to find out that most custom-design RTX 2060 Super graphics cards are based on the existing RTX 2060 PCBs with all memory pads populated and only the very premium overclocked models using the larger PCB meant for the RTX 2070. With their custom-design boards, NVIDIA allows their partners to innovate not just with cooling, but giving their cards a small degree of factory overclocking and increases to the board power limit.
In this review, we take a look at the Galax GeForce RTX 2080 Super EX, which is actually the first graphics card from Galax we review in a decade. Unlike other vendors, Galax isn't asking for a price premium over the RTX 2060 Super Founders Edition, yet includes features like idle fan stop, a better cooler, and an overclock out of the box.