Why did multi-GPU gaming fall out of favor, and could SLI make a comeback?

Multi-GPU setups rarely delivered twice the performance. Each game needed specific support from both the developers and the graphics drivers, and that support was often incomplete or inconsistent. Since only a small number of players used two GPUs, developers had little incentive to spend time optimizing for it. In most cases, buying one faster card was simpler and more reliable.

Having two physical PCIe slots doesn’t necessarily mean a motherboard can support multi-GPU performance properly. Consumer platforms may not provide enough PCIe lanes or bandwidth, especially when both slots are being used heavily. Proper configurations often need workstation- or server-style lane layouts, which are uncommon on mainstream systems.

Modern graphics APIs can technically support multiple GPUs through features such as explicit multi-adapter support in DirectX 12. The catch is that the game itself has to implement and maintain that support. A few specialized titles have experimented with it, but the development cost is high and the audience is too small for it to become common.

The biggest issue was frame pacing. Even when the average frame rate looked good, the two cards often rendered frames unevenly, causing noticeable stutter and jitter. Keeping both cards’ frame buffers synchronized also required a lot of bandwidth, and that becomes even more difficult as modern games use more VRAM.

The technology hasn’t completely disappeared, since some multi-GPU options still exist for certain professional or compute workloads. Gaming is where it stopped making much sense: inconsistent scaling, extra power and heat, synchronization problems, and limited game support usually outweigh the benefits.