Desktop GPU · NVIDIA
NVIDIA GeForce RTX 5090 specifications and launch price
Specifications transcribed from NVIDIA's official product page, read on . Current retail prices and stock are not collected; retailer links are searches, not matched listings.
NVIDIA GeForce RTX 5090 specifications
| Memory | 32 GB · GDDR7 |
|---|---|
| Shader units | 21,760 |
| FP32 (theoretical) | 104.8 TFLOPS (derived: 2 × 21,760 shader units × 2.41 GHz) |
| Launch price | $1,999 (starting at, announced Jan 6, 2025)Official announcement |
| Value at launch price | 52.4 GFLOPS per $ (derived FP32) |
| Base clock | 2.01 GHz |
| Boost clock | 2.41 GHz |
| Launch | Unknown |
| Source | Official specs Read |
| Find retailers | Search NeweggSearch Amazon |
NVIDIA desktop GPUs side by side
| GPU | Memory | FP32 (theoretical) | Launch price | Value at launch price | Launch |
|---|---|---|---|---|---|
| NVIDIA GeForce RTX 4060 | 8 GB · GDDR6 | 15.1 TFLOPS (derived: 2 × 3,072 shader units × 2.46 GHz) | $299 (starting at, announced May 18, 2023) | 50.5 GFLOPS per $ (derived FP32) | Unknown |
| NVIDIA GeForce RTX 4060 Ti | Size not recorded (see official page) · GDDR6 | 22.1 TFLOPS (derived: 2 × 4,352 shader units × 2.54 GHz) | Not recorded | Not available | Unknown |
| NVIDIA GeForce RTX 5050 | 8 GB · GDDR6 | 13.1 TFLOPS (derived: 2 × 2,560 shader units × 2.57 GHz) | $249 (starting at, announced Jun 24, 2025) | 52.6 GFLOPS per $ (derived FP32) | Unknown |
| NVIDIA GeForce RTX 5060 | 8 GB · GDDR7 | 19.2 TFLOPS (derived: 2 × 3,840 shader units × 2.5 GHz) | $299 (starting at, announced Apr 15, 2025) | 64.2 GFLOPS per $ (derived FP32) | Unknown |
| NVIDIA GeForce RTX 5060 Ti | Size not recorded (see official page) · GDDR7 | 23.6 TFLOPS (derived: 2 × 4,608 shader units × 2.57 GHz) | Not recorded | Not available | Unknown |
| NVIDIA GeForce RTX 5070 | 12 GB · GDDR7 | 30.8 TFLOPS (derived: 2 × 6,144 shader units × 2.51 GHz) | $549 (starting at, announced Jan 6, 2025) | 56.1 GFLOPS per $ (derived FP32) | Unknown |
| NVIDIA GeForce RTX 5070 Ti | 16 GB · GDDR7 | 43.9 TFLOPS (derived: 2 × 8,960 shader units × 2.45 GHz) | $749 (starting at, announced Jan 6, 2025) | 58.6 GFLOPS per $ (derived FP32) | Unknown |
| NVIDIA GeForce RTX 5080 | 16 GB · GDDR7 | 56.3 TFLOPS (derived: 2 × 10,752 shader units × 2.62 GHz) | $999 (starting at, announced Jan 6, 2025) | 56.3 GFLOPS per $ (derived FP32) | Unknown |
| NVIDIA GeForce RTX 5090 | 32 GB · GDDR7 | 104.8 TFLOPS (derived: 2 × 21,760 shader units × 2.41 GHz) | $1,999 (starting at, announced Jan 6, 2025) | 52.4 GFLOPS per $ (derived FP32) | Unknown |
Launch dates keep the precision the cited source states (day, month or quarter); "Unknown" means we have not yet verified a launch date from a cited source. Intel's quarters are its stated "Launch Date", which Intel defines as the date the product was first introduced; retail availability can come later. Equal core counts or clocks do not mean equal performance. FP32 is a theoretical peak, not a measured result. NVIDIA figures are derived here as 2 × shader units × boost clock, rounded down. Launch prices are the maker's announced US launch price (NVIDIA "starting at", AMD's stated "SEP") on the announcement date shown, not today's retail price; "Not recorded" means we have not recorded a dated launch announcement of one price for that exact product (the RTX 5060 Ti launched as 8 GB and 16 GB models at different prices; an undated list price, such as Intel's "Recommended Customer Price", is not used). Value at launch price is the FP32 figure per launch dollar, so it inherits the same not-comparable-across-vendors caveat. Check the exact model, board and seller on the retailer's site. See the methodology.