NVIDIA RTX 5070 - 12GB usable
Specifications
| Brand | NVIDIA |
|---|---|
| Model | RTX 5070 |
| Usable VRAM | 12GB |
| Architecture | Blackwell |
| CUDA / Stream Processors | 6,144 |
| Memory Bandwidth | 672 GB/s |
| TDP | 250W |
| FP32 TFLOPS | 31 |
Current Offers
Used from $700New from $770
Prices last updated:
GPUDojo is reader-supported. When you buy through links on our site, we may earn an affiliate commission.
Price History
- eBay$700near high
- Amazon$770current
- Newegg$840at high
For AI / LLM Use
Entry-level for local AI. Handles 7B-8B models well.
What Models Can It Run?
- 14B Q4_K_M, 7B full precision
- 7B Q6_K, 14B Q3_K (tight)
- 7B Q4_K_M only
Estimated Performance
Generation: ~50 tokens/sec
Prefill: ~554 tokens/sec
Recommended Quantisations
- Q4_K_M for 14B (tight fit)
- Q6_K or Q8 for 7B models
Pros & Cons
Pros
- Blackwell architecture: broad CUDA software support
- Consumer card: easy to install, display output
Cons
- 12GB usable VRAM: may need quantization for 30B+ models
- Moderate memory bandwidth: not the fastest for inference
Community Verdict
No community reviews yet for the RTX 5070. Know a good review? Let us know.