NVIDIA and AMD’s GPU naming conventions, like CPUs, follow a real, decodable logic — genuinely useful to understand before comparing cards across generations.
NVIDIA GeForce RTX 4070 Ti Super
| | | | |
| | | | +-- Super = a mid-generation refresh variant
| | | +----- Ti = a higher-performance variant of the base card
| | +-------- 070 = performance tier within the generation
| +------------ 4 = generation (RTX 40-series)
+---------------- RTX = ray tracing capable product line
AMD Radeon RX 7900 XTX
| | |
| | +-- XTX = highest performance variant in this tier
| | (XT is a step below, plain is the base card)
| +------- 900 = performance tier within the generation
+---------- 7 = generation (RX 7000 series)
Why Generation Matters More Than the Tier Number Alone
A current-generation 70-series NVIDIA card frequently outperforms the previous generation’s 80-series card, thanks to genuine architectural improvements (not just raw core count) between generations — comparing purely by tier number across different generations is a common, genuinely misleading mistake when shopping.
VRAM: The Number Worth Checking Independently
Unlike CPU naming, GPU model numbers don’t reliably encode VRAM capacity — the same model number occasionally ships in different VRAM configurations, so checking the actual specified VRAM amount directly, rather than assuming it from the model name alone, is worth the extra minute, especially for AI/local-LLM use cases where VRAM capacity specifically is often the binding constraint.