GPU Comparison

Model numbers are marketing, not measurement — a new mid-range card is not automatically faster than an older high-end one. The comparison above works from a normalised performance score for rasterised gaming, the same figure the Bottleneck and FPS calculators use.

VRAM is a cliff, not a slope

Running short of video memory does not gradually reduce your frame rate; it causes sudden, jarring stutter as data is shuttled over the PCIe bus. That makes VRAM capacity qualitatively different from raw speed. 8 GB is increasingly tight at 1440p with high textures, 12 GB is a reasonable floor for 4K, and 16 GB or more gives comfortable headroom — and matters again if you run local AI models.

Ray tracing changes the ranking

Two cards that match in traditional rendering can be far apart once ray tracing is enabled, where NVIDIA generally holds an advantage. If you intend to play with ray tracing on, check the Ray Tracing Performance tool rather than relying on rasterised figures.

Upscaling is part of the buying decision now

DLSS, FSR and XeSS all reconstruct a higher-resolution image from a cheaper one, and their quality differs. Frame generation adds more frames but also latency, so it helps most when your base frame rate is already reasonable. These features increasingly decide which card feels faster in practice.

Check it physically fits

High-end cards are large, heavy and hungry. Before buying, confirm length and slot clearance with the GPU Size Checker and power requirements with the PSU Calculator — the right connector matters as much as the wattage.