Upscaling and frame generation are critical features today, especially since game optimization has seemingly only gotten worse. If you’re using a GPU that doesn’t natively support either, you’re in luck: There is a third-party option that works on any GPU.
Upscaling and frame gen are the future
AI-driven upscaling, like DLSS, and frame generation have become ubiquitous on all newer GPUs from all the major manufacturers, and there is no sign it is going away.
The improvements over the last several generations have taken it from the territory of an interesting experiment to something that I sorely miss when it isn’t around. I’ve become particularly attached to frame generation in cinematic, single-player titles like Cyberpunk 2077, since they’re exceptionally demanding and don’t require perfect responses like some online competitive games do.
Unfortunately, if you’re using an older GPU (including the exceptionally long-lived 10-series GPUs from NVIDIA) or integrated graphics, you’ve mostly been left out in the non-AI enhanced cold.
That is where a program called Lossless Scaling comes in. Lossless Scaling enables you to use an “old-fashioned” GPU to do the same kind of upscaling and frame generation that you get with NVIDIA’s RTX series GPUs, just a bit less efficiently.
It is available on their website or on Steam for $7.
How does Lossless Scaling work without special hardware?
Upscaling and frame gen have largely become synonymous with NVIDIA in recent years, mostly due to the incredible success of their DLSS technology. However, other companies have competing technologies that are also capable, like AMD’s FSR and Intel’s XeSS.
Whatever they’re called, all of those technologies leverage specialized hardware on your graphics card that is optimized for the kind of low precision math (FP8 or INT8 for example) that is useful for AI workloads. They then render the game at a lower resolution and use artificial intelligence to upscale to whatever resolution you want. In general, the lower the native resolution, the harder it is to upscale while maintaining the same visual quality using AI upscaling.
What is NVIDIA DLSS, and How Will It Make Ray Tracing Faster?
At NVIDIA’s CES 2019 presentation, the company showed off a new technology called DLSS.
However, that specialized hardware isn’t strictly necessary. Even GPUs without dedicated hardware are capable of doing the math required for AI technologies, they’re just worse at it.
You can use a second GPU
The GPU you’re using with Lossless Scaling might be a bit worse with AI workloads than an RTX 5080, but Lossless Scaling does offer a way to counter it: a second GPU.
If you have a spare PCIe slot in your PC, along with a second GPU and a large enough power supply, you can dedicate one GPU to rendering the game and one GPU to handling the upscaling and the frame generation. Splitting the burden between two GPUs theoretically allows you to get better performance than if you were just using a single GPU.
Lossless Scaling’s dual GPU option requires a bit of setup, but it doesn’t require identical GPUs. That means if you can pick up a used GTX 1060 for a good (low) price on the used market, you can use it in tandem with your 1070 Ti, or your RX 580.
Multiple settings for multiple kinds of games
Lossless Scaling has another feature that NVIDIA and AMD do not: style-specific upscaling. It includes multiple scaling techniques that are optimized for different purposes. They are:
- LS1
- AMD FidelityFX Super Resolution
- NVIDIA Image Scaling
- Integer Scaling
- Nearest Neighbor
- xBR
- Anime4K
- Sharp Bilinear
- Bicubic CAS
For general purpose modern gaming, the Lossless Scaling’s developer recommends LS1 or FSR. If you’re playing a pixel-art game, you should use Integer Scaling or xBR instead.
While the games I play don’t benefit from the Integer Scaling, XBR, or Anime4K scaling models, it is nice that it offers something for everyone.
Lossless Scaling has some limitations
Lossless Scaling, and upscaling in general, really shine when you’re “GPU-bound.” In other words, when you graphics card is the thing that is slowing you down. In those situations, using upscaling technology bumps down the resolution that you’re rendering the game at and then uses a different approach to make it look like it is your ideal resolution, which can be less demanding on your GPU.
That makes your game run more smoothly. If, however, you’re in a situation where your CPU is the limiting factor, then scaling won’t help you. You can render at whatever low resolution you want, but your performance will still be poor. Adding a second GPU in this scenario to try to “help Lossless Scaling out more” won’t improve the situation either, since ultimately, the GPU wasn’t the problem in the first place.
CPU or GPU Bottleneck? How to Tell (and Which Is Worse)
How to find a PC’s weak point so you can properly optimize your system.
Frame Generation is similarly limited. If you’re limited by your CPU rather than your GPU, you’ll still feel lag because your CPU is struggling to keep up, even if it looks like you’re getting 60 FPS. It is a distinct and very frustrating form of lag to deal with, and you’re better off trying to tweak settings that reduce the load on the CPU if you’re stuck in this situation.
Lossless Scaling itself is only available on Windows, but if you’re willing to do a little bit of tinkering, you can get Lossless Scaling’s frame generation algorithm running on Linux. I’ve even seen it work on a Steam Deck with surprisingly good results.
