DLSS 5 Mod for Ampere GPUs Delivers Poor Performance, Tanks Frame Rates

DLSS 5 Mod Enables Older Hardware, With Steep Performance Penalties

A community-driven mod has successfully enabled NVIDIA's latest Deep Learning Super Sampling (DLSS) 5 technology to run on older, but still capable, NVIDIA Ampere (RTX 30-series) GPUs. While the prospect of bringing next-generation upscaling features to hardware that predates the RTX 40-series is enticing, the reality is proving to be a significant performance drain. Early reports indicate that most games see frame rates plummet to single digits, rendering them unplayable. Even with high-end Ampere cards, achieving playable frame rates requires drastically reducing in-game settings, with some scenarios topping out around 30-40 FPS.

DLSS 5, which leverages AI and dedicated hardware (Tensor Cores) to upscale lower-resolution images to higher resolutions, is a key feature for modern gaming, offering improved visual fidelity without a commensurate performance hit. Typically, DLSS is exclusive to NVIDIA's Ada Lovelace architecture (RTX 40-series GPUs) due to its reliance on newer, more powerful Tensor Cores and Optical Flow Accelerators. This mod, however, circumvents these hardware restrictions by patching the necessary DLL files. The goal is to trick the game into believing a compatible GPU is present, allowing the DLSS 5 algorithms to execute on the older hardware.

The technical feat of making DLSS 5 operate on Ampere GPUs is noteworthy. It implies that the core algorithms, while optimized for newer hardware, can still be executed, albeit inefficiently, on the previous generation. This suggests that the foundational AI models and processing pathways might have some degree of backward compatibility. However, the performance degradation is a stark reminder of the hardware advancements NVIDIA has made. The newer Tensor Cores in RTX 40-series cards are significantly more powerful and efficient, capable of handling the complex computations required by DLSS 5 with far greater speed.

Consider the difference between running a highly optimized, cutting-edge AI model on a dedicated, state-of-the-art AI accelerator versus running it on a general-purpose processor that wasn't designed for such intensive parallel computation. The latter can often perform the task, but it will be orders of magnitude slower and generate substantially more heat. This is precisely what appears to be happening with DLSS 5 on Ampere GPUs. The older Tensor Cores are being asked to perform computations they were not designed for at this scale, leading to the abysmal frame rates observed.

Performance Analysis: A Steep Price for New Tech

The performance figures emerging from early adopters are concerning. In most modern titles, even with the DLSS 5 mod applied and in-game settings minimized, frame rates are reportedly falling into the single digits. This level of performance is far below what is considered playable for any type of gaming, from fast-paced shooters to slower-paced strategy games. For context, a smooth gaming experience is typically considered to be around 60 FPS, with 30 FPS being a minimum acceptable threshold for many.

Even in less demanding scenarios or with specific game titles that might be more amenable to the mod, high-end Ampere GPUs like the RTX 3090 or RTX 3080 are struggling to break past 30-40 FPS. This is a dramatic drop from their native performance without DLSS, and significantly lower than what native DLSS 3 or DLSS 2 would provide on compatible hardware. The mod essentially trades visual quality (by forcing lower internal rendering resolutions) for upscaling, but the underlying processing bottleneck on the Ampere hardware prevents a meaningful performance uplift.

The surprising detail here is not that performance is poor, but the sheer extent of the degradation. One might expect a performance hit when forcing newer technology onto older hardware, but dropping to single digits suggests a fundamental mismatch in processing capabilities. It highlights how much of DLSS 5's efficiency and effectiveness is tied to the specialized hardware introduced with the Ada Lovelace architecture.