Despite their popularity among hardcore gamers, SLI and Crossfire setups have gone the way of the dodo, with most PC builders arming their systems with a single battle-ready graphics card. However, there are still some niche cases where you can get better results with multiple graphics cards. And I don’t just mean server PCs or home labs either.

If you’ve got a second GPU gathering dust, here are some ways you can put it to good use by pairing it with your everyday PC.

4 Split LLM processing workloads

And train large language models

If you want maximum privacy and customization in your text or image generators, you can try hosting AI tools like Ollama on your local hardware. However, LLM-heavy workloads can require a lot of firepower on your end – to the point where certain high-end models may be held back by your graphics card.

But if you’ve got a separate GPU in your computing arsenal, you can slot it into your PC and leverage its computing prowess in your AI tasks. In fact, AI experts and machine learning enthusiasts can even turn their normal PC into an LLM-training powerhouse by distributing the processing load across separate GPUs.

3 Speed up the Cycles renderer in Blender

Ideal for artists knee-deep in rendering tasks

Thanks to its open-source origins, versatile nature, and massive community, Blender has quickly become one of the best utilities for 3D modeling, rigging, and animation. However, rendering your picturesque Blender models can take a toll on your graphics card, especially if you use something as demanding as a Cycles renderer.

Thankfully, Blender supports multi-GPU rendering, and enabling it is as simple as selecting your preferred GPUs as the Cycles Render Devices within the Systems tab of the Blender Preferences menu.

2 Circumvent the lack of 32-bit PhysX support

This one’s for Blackwell GPU owners

Between the terrible pricing and lackluster generational increase in performance, calling the RTX 5000 series' launch rocky would be an understatement. But the icing on this thermally challenged, smoky cake was Team Green getting deprecating support for 32-bit CUDA and PhysX. Sure, only a handful of titles may rely on the older version of CUDA and PhysX, but you’re out of luck if you want to enjoy the OG Mirror’s Edge, Borderlands 2, or Batman: Arkham Asylum at max graphical fidelity on your shiny new RTX 5000 card.

Or at least, that would be the case if not for the ingenuity of the PC community. Redditor jerubendo used a dual-GPU build, where an RTX 3050 performs PhysX calculations while an RTX 5090 tackles the rest of the graphical heavy-lifting. All in all, it’s a really neat way to use your spare graphics card to bypass the lack of support for 32-bit PhysX on the newest Team Green cards.

1 Gain extra FPS via Lossless Scaling

A legit way to download more FPS

Let’s face it: frame generation isn’t the same as raw performance. But if you’re on an outdated graphics card and want to enjoy story-driven titles, turn-based RPGs, and other slow-paced games at solid graphical fidelity, frame generation is definitely worth checking out. Thanks to Lossless Scaling, you don’t have to wait for Nvidia to add frame generation support to your favorite titles either.

The best part? You can offload Lossless Scaling’s frame generation workloads to a secondary GPU while your primary graphics card renders the actual game. Having recently tested Lossless Scaling with my GTX 1080 + RTX 3080 Ti setup, I can confirm that the app makes dual GPUs a viable option for gamers who don’t mind using fake frames to level the playing field in poorly-optimized titles.

There are still a couple of caveats to dual-GPU setups

All that said, you should be aware of certain issues with adding a second graphics card to your setup. Modern cabinets (and even motherboards) aren’t exactly built for dual-GPU systems, so you might have some issues accommodating two graphics cards inside the PC. Likewise, you’ll need a fairly beefy power supply that can run both GPUs at their full potential.

Then there’s the issue that you’ll need GPUs with similar specs (preferably on the high-end side) for most of the setups I’ve listed in the article. For instance, running Lossless Scaling or splitting the rendering load in Blender makes a lot of sense when you want to pair an RTX 4080 with, say, a 4070 or a 3080. But if you’re expecting to run Cyberpunk 2077 at 4K Ultra settings with 120 FPS on an RTX 4070 + GTX 1050 Ti setup, you’ll be sorely disappointed when the latter fails to keep up with its (relatively) powerful sibling.