Component comparison
GeForce RTX 5070 vs GeForce RTX 4090
GeForce RTX 5070 vs GeForce RTX 4090 — detailed GPU comparison: performance, gaming FPS, specifications.
Advantages
A
GeForce RTX 5070
- + Release year (+2)
- + Die size (mm²) (+346 mm²)
B
GeForce RTX 4090
- + VRAM (GB) (+8 GB)
- + Memory clock (MHz) (+250 MHz)
- + Bandwidth (GB/s) (+362 GB/s)
- + Memory bus (bit) (+192 bit)
- + TGP / TDP (W) (+20 W)
- + Transistors (M) (+45,200M)
- + CUDA cores (+10240)
- + TMUs (+320)
- + ROPs (+96)
- + RT cores (+80)
- + Tensor cores (+320)
- + FP32 performance (TFLOPS) (+51.71 TFLOPS)
- + Pixel fill rate (GPixel/s) (+242.50 GPixel/s)
- + Texture fill rate (GTexel/s) (+807.70 GTexel/s)
- + Energy efficiency (+20.6)
- + PassMark (overall) (+7972)
- + 3DMark Fire Strike (+11242)
- + Normalized score (+5.4)
Detailed spec comparison
| Spec |
Component A
GeForce RTX 5070
|
Component B
GeForce RTX 4090
|
|---|---|---|
| Release year | 2025 (+2) | 2023 |
| VRAM (GB) | 8 GB | 16 GB (+8 GB) |
| Memory type | GDDR7 | GDDR6 |
| Memory clock (MHz) | 2000 MHz | 2250 MHz (+250 MHz) |
| Bandwidth (GB/s) | 672 GB/s | 1034 GB/s (+362 GB/s) |
| Memory bus (bit) | 192 bit | 384 bit (+192 bit) |
| TGP / TDP (W) | 100 W | 120 W (+20 W) |
| Lithography (nm) | 4 nm | 4 nm |
| Transistors (M) | 31,100M | 76,300M (+45,200M) |
| Die size (mm²) | 263 mm² (+346 mm²) | 609 mm² |
| Base clock (MHz) | — | — |
| Boost clock (MHz) | — | — |
| CUDA cores | 6,144 | 16,384 (+10240) |
| TMUs | 192 | 512 (+320) |
| ROPs | 80 | 176 (+96) |
| RT cores | 48 | 128 (+80) |
| Tensor cores | 192 | 512 (+320) |
| FP32 performance (TFLOPS) | 30.87 TFLOPS | 82.58 TFLOPS (+51.71 TFLOPS) |
| Pixel fill rate (GPixel/s) | 201.00 GPixel/s | 443.50 GPixel/s (+242.50 GPixel/s) |
| Texture fill rate (GTexel/s) | 482.30 GTexel/s | 1290.00 GTexel/s (+807.70 GTexel/s) |
| Energy efficiency | 21.3 | 41.9 (+20.6) |
| PassMark (overall) | 19,075 | 27,047 (+7972) |
| 3DMark Time Spy | — | 21,251 |
| 3DMark Fire Strike | 34,949 | 46,191 (+11242) |
| Geekbench OpenCL | — | — |
| Geekbench Vulkan | — | — |
| Normalized score | 69.6 | 75.0 (+5.4) |
In both
CUDA
DLSS
ReBAR
ray_tracing
FPS in games (1080p)
| Game | GeForce RTX 5070 | GeForce RTX 4090 | Delta |
|---|---|---|---|
| Battlefield 5 | 180 FPS | 170 FPS | +5.9% (A) |
| CS:GO | 240 FPS | 240 FPS | ≈ |
| Counter-Strike 2 | 300 FPS | 300 FPS | ≈ |
| Cyberpunk 2077 | 170 FPS | 147 FPS | +15.6% (A) |
| Far Cry 5 | 322 FPS | 173 FPS | +86.1% (A) |
| Fortnite | 300 FPS | 300 FPS | ≈ |
| Forza Horizon 4 | 270 FPS | 250 FPS | +8% (A) |
| Forza Horizon 5 | 329 FPS | 281 FPS | +17.1% (A) |
| GTA V | 140 FPS | 140 FPS | ≈ |
| Metro Exodus | 170 FPS | 156 FPS | +9% (A) |
| Overwatch | 240 FPS | 240 FPS | ≈ |
| PUBG | 140 FPS | 140 FPS | ≈ |
| Resident Evil 4 Remake | 210 FPS | 218 FPS | 3.8% (B) |
| The Witcher 3 | 436 FPS | 394 FPS | +10.7% (A) |
| Valorant | 400 FPS | 350 FPS | +14.3% (A) |
Specs from public sources (PassMark, Geekbench). FPS measured at 1080p medium settings.
Based on the comparison:
GeForce RTX 4090 wins 18 of the comparable criteria.
The choice depends on your use case.
🤔 How to choose
Universal advice for any use case.
- Reliability vs price: reliability (median 50) ↔ price (median 385.1).
- Weight vs CPU: mobility (median 48) ↔ overall performance (median 49.8).
- Upgradeability vs battery: upgradability (median 50) ↔ battery life (median 2.8).