Side-by-side GPU comparison: specs, memory, compute performance, and live cloud pricing.
Verdict
RTX 5080 has 4 GB more VRAM, making it better suited for large models and long context windows. For compute-bound workloads like training, RTX 5080 delivers 12.9× higher FP16 throughput. RTX 5080 supports a broader range of models (502 vs 486 from this catalog), giving more flexibility.
Specifications
| K80 | RTX 5080 | |
|---|---|---|
| VRAM | 12 GB | 16 GB |
| VRAM Type | GDDR5 | GDDR7 |
| Memory Bandwidth | 0.2 TB/s | 1.0 TB/s |
| FP16 Performance | 9 TFLOPS | 113 TFLOPS |
| Manufacturer | NVIDIA | NVIDIA |
| FP8 Support | No | Yes |
| FP4 Support | No | Yes |
Price / Performance
Based on cheapest single-GPU on-demand pricing. Lower $/TFLOP = better compute value; lower $/GB = better memory value.
| K80 | RTX 5080 | |
|---|---|---|
| $/hr (cheapest) | $0.18 | — |
| $/TFLOP (compute value) | $0.0206 | — |
| $/GB VRAM (memory value) | $0.0150 | — |
Cloud Pricing
Cheapest on-demand price per provider (single GPU).
K80
| Provider | On-demand | Spot | Rent |
|---|---|---|---|
| Microsoft Azure | $0.18/hr | $0.10/hr |
RTX 5080
No cloud pricing available.
Model Compatibility
Models from the catalog that fit on each GPU, grouped by required precision.
K80 (486 models)
RTX 5080 (502 models)
You might also compare…
Pricing data refreshed hourly · Last updated April 11, 2026 · Browse all comparisons