PG506-242
NVIDIA
24GB VRAM · 933 GB/s bandwidth · 10.3 FP16 TFLOPS · 165W TDP
The PG506-242 has 24GB of VRAM with 933 GB/s memory bandwidth and 10.3 TFLOPS FP16 compute. At Q4 quantization, it can comfortably run Gemma 3 4B (289 tok/s), Qwen 2.5 7B (152 tok/s), Llama 3.1 8B (144 tok/s). Models larger than ~41B parameters won't fit even at Q4. Electricity cost is approximately $18/month at 165W TDP.
What LLMs can you run?
| Model | Params | Q4 Weight | Fit | Decode |
|---|---|---|---|---|
| Gemma 3 4B | 4.0B | 2 GB | comfortable | 289 tok/s |
| Qwen 2.5 7B | 7.6B | 4 GB | comfortable | 152 tok/s |
| Llama 3.1 8B | 8.0B | 4 GB | comfortable | 144 tok/s |
| Mistral Small 24B | 24.0B | 12 GB | comfortable | 48 tok/s |
| Gemma 3 27B | 27.4B | 14 GB | comfortable | 42 tok/s |
| Qwen 2.5 Coder 32B | 32.5B | 16 GB | comfortable | 35 tok/s |
| Llama 3.3 70B | 70.6B | 35 GB | won't fit | |
| Qwen 2.5 72B | 72.7B | 36 GB | won't fit | |
| Llama 3.1 405B | 405B | 202 GB | won't fit | |
| DeepSeek R1 671B | 671B | 336 GB | won't fit |
Similar GPUs
| GPU | VRAM | BW | TFLOPS | TDP |
|---|---|---|---|---|
| A30 PCIe | 24GB | 933 GB/s | 10.3 | 165W |
| PG506-207 | 24GB | 933 GB/s | 10.3 | 165W |
| PG506-217 | 24GB | 933 GB/s | 10.3 | 165W |
| PG506-232 | 24GB | 933 GB/s | 10.3 | 165W |
| GeForce RTX 3090 | 24GB | 936 GB/s | 35.6 | 350W |
Compare with another GPU
Select another GPU to compare specs and model performance side by side.