[Hardware] Upscaling can't save the world's slowest 'modern' GPU — FSR doubles performance on GT 1030, but titles still barely playable

[protaa]

 

After testing FSR and XeSS upscaling on a GTX 660, YouTuber GCS Hardware has moved on to the GT 1030 (the GDDR5 variant) to see how the slowest DX12-capable GPU copes with upscaling. According to a video they published on the matter, the GT 1030 technically can take advantage of FSR upscaling, but it still isn't enough to boost FPS to playable territory, except on GTA V: Enhanced Edition.

The YouTuber expanded their testing suite with more modern titles, featuring: Kingdom Come Deliverance II, Space Marine II, Alan Wake 2, Cyberpunk 2077, CS 2, and GTA V Enhanced.

The GT 1030 delivered a measly 8 FPS average at native 1080p resolution in Kingdom Come Deliverance II. Enabling FSR 3.1 performance mode significantly boosted performance to 15 FPS average, and moving to a 720p render resolution with the same settings yielded an even greater 21 FPS average. Despite the gains, performance was still well below playable FPS.

Space Marine II was even worse, averaging just 6 FPS on the GT 1030 at 1080p native resolution with low settings. FSR performance mode upscaling more than doubled performance to 13 FPS, but the game was still far from playable.

Alan Wake 2 was the worst-performing game of the bunch, performing in seconds per frame rather than frames per second at native resolution. Native 1080p performance was below 0 FPS average, and FSR ultra performance mode upscaling boosts performance to 3 FPS average.

Cyberpunk 2077 at 1080p native low settings achieved 12 FPS average on the GT 1030. FSR 2.1 ultra performance mode boosted FPS to 21 FPS average, and FSR 3 boosted FPS to 18 FPS average. XeSS had the opposite effect, achieving just 8 FPS average in performance mode.

CS 2, boasted similar issues on the GT 1030 as on the GTX 660, which the YouTuber reviewed two weeks ago. Native performance was decent; 100 FPS average at low settings, but FSR upscaling did not provide any noteworthy performance improvements beyond native rendering.

GTA V Enhanced ran surprisingly well despite overflowing the GT 1030 GDDR5's 2GB memory capacity. The game ran at 30-40 FPS average while consuming 3 GB of memory (video memory overflows to system RAM when full). FSR 3 performance mode boosted performance to 45 FPS, and FSR 1 performance mode boosted performance to 50 FPS average.

Similar to GCS Hardware's GTX 660 testing, Nvidia's Pascal-based GT 1030 is not powerful enough to play the modern AAA games at playable FPS with upscaling, except for older titles featuring upscaling support.

Still, the GT 1030's performance is a significant improvement from the GTX 660. The Pascal-based GPU was technically capable of performing significantly better with upscaling, in most of the aforementioned titles. The GTX 660, by contrast, did not benefit from upscaling at all in any game GCS Hardware threw at it.

Nvidia's newer Pascal architecture contributes to the GT 1030's ability to take advantage of FSR upscaling better than the GTX 660. FSR is designed to use FP16 calculations to process the necessary data to upscale an image from a lower resolution to a higher resolution with as few artifacts as possible.

Pascal was the first Nvidia architecture to natively support FP16 compute. By contrast, the GTX 660 was based on the Kepler architecture, which only features FP32 and FP64 compute support. FSR has an FP32 fallback mode for these types of GPUs, but using FP32 compute reduces performance.

 

https://www.tomshardware.com/pc-components/gpus/upscaling-cant-save-the-worlds-slowest-modern-gpu-fsr-doubles-performance-on-gt-1030-but-titles-still-barely-playable