SerpApi.Org
There was a time, not too long ago in the grand scheme of things, when one GPU just wasn't enough. If you wanted the fastest machine possible, you bought two identical Nvidia GPUs, connected them with an SLI bridge, and hoped your favorite game could take advantage of both.
That's no longer an option for consumer GPUs, and nobody tries to build a multi-GPU setup for gaming alone. But Nvidia didn't kill SLI in one swift blow.
SLI was every enthusiast's dream come true
Although our wallets weren't too pleased
For a certain kind of PC enthusiast, one with deep pockets, one flagship GPU was only the beginning when SLI was a thing. People built entirely new systems around two matching graphics cards, buying an SLI-certified motherboard, a humongous PSU, and a case with enough room to keep both cards from suffocating.
Some builders even went as far as three-way or four-way SLI, not because it was even remotely sensible, but because they wanted the fastest and most excessive gaming PC money could buy.
The appeal was easy to understand, and although I couldn't afford it myself at the time, I certainly dreamed of such a setup, too. When SLI worked properly, two powerful GPUs could push resolutions and graphics settings that a single card struggled to handle, while also producing the kind of benchmark results enthusiasts loved showing off. An SLI bridge became a visible badge of a no-expense-spared, no-holds-barred build.
The experience never matched the benchmark charts
Does double the hardware equal twice the performance?
The problem was that adding a second GPU didn't automatically double your frame rate. SLI depended heavily on individual game support and Nvidia's driver profiles, so the results varied wildly. One game might scale beautifully and make the enormous investment feel worthwhile, while another barely used the second card at all. Some games even ran worse or developed visual glitches until SLI was disabled.
Average frame rates didn't tell the whole story, either. Multi-GPU setups could suffer from uneven frame delivery and microstutter, making a high benchmark result feel less smooth than you'd expect based on the fps. VRAM also wasn't simply added together, since both GPUs generally needed their own copies of the same assets.
Nvidia removed SLI one piece at a time
Each generation made multi-GPU support more exclusive
Nvidia didn't just wake up one morning and announce that SLI was dead and gone. It slowly narrowed the feature until hardly anyone had any real use for it (in gaming, that is).
Three- and four-way configurations fell out of favor during the Pascal era, and the RTX 20 series replaced the familiar SLI connector with NVLink on only a handful of expensive GPUs. By the RTX 30 series, the RTX 3090 was the last consumer model with the necessary connector, turning what once had been an enthusiast feature into something super niche.
The software side disappeared alongside the hardware. Nvidia stopped creating new SLI driver profiles for RTX 20-series and older GPUs, instead leaving new multi-GPU support up to individual game devs. The RTX 40-series then arrived without NVLink, and Nvidia kept it off the RTX 50-series, too. Even the RTX 5090 doesn't have it.
Native multi-GPU support was the escape hatch nobody used
DirectX 12 kept multi-GPU gaming technically possible
SLI disappearing didn't make multi-GPU gaming impossible overnight, because DX12 and Vulkan gave developers the tools to divide work between multiple graphics cards. In theory, that gave game studios far more control over how each GPU was used.
In practice, though? This meant a lot of work, including painful resource management, synchronization, and testing across a tiny number of dual-GPU PCs. It just wasn't practical. While a few games added multi-GPU support, most developers understandably decided that the extra effort just wasn't worth it for such a small audience.
Multi-GPU tech survived, just not for games
No one's dual-wielding RTX 5090s
SLI didn't kill off the idea of using multiple GPUs, though. These setups are still common in professional rendering, scientific computing, and AI, where a workload can often be divided into large chunks and spread across several graphics cards. In those situations, all that extra hardware, power draw, and software development can be justified.
Gaming is a completely different story. A game has to produce each frame quickly and consistently while responding to user input, and coordinating several GPUs only adds more complexity. These days, both Nvidia and game devs lean into software instead of hardware, with features like DLSS.
Dual GPUs for gaming aren't coming back
At a time when most people are having a hard time justifying the expense of buying just one graphics card, can you imagine if we were suddenly allowed to buy two? Gaming on ultra would slowly become optimized for that dual GPU setup again, and we'd all be in shambles at the idea of needing two RTX 5080s or RTX 5090s to max out the settings at 4K.
Thankfully, that's not going to happen; I don't think we'll see dual GPUs in gaming again. But you can still use an old GPU with Lossless Scaling; it'll just have nothing to do with SLI.
