Intel Arc B580 12GBBudget pickC103.4 tok/s decode
~$249 MSRP
Can it run?
Very compromised (needs ~0.3 GB host RAM)
Using Q5_K_M in Ollama
Fit status
Very compromised (needs ~0.3 GB host RAM)
Decode
23.2 tok/s
TTFT
8341 ms
Safe context
14K
Memory
4.6 GB / 4.0 GB
Offload
10%
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | D | Very compromised (needs ~0.3 GB host RAM) | 22.6 tok/s | 12436 ms | 27K |
| Chat | D | Very compromised (needs ~0.3 GB host RAM) | 23.2 tok/s | 4550 ms | 7K |
| Coding | D | Very compromised (needs ~0.3 GB host RAM) | 23.2 tok/s | 8341 ms | 14K |
| RAG | D | Very compromised (needs ~0.3 GB host RAM) | 22.6 tok/s | 15544 ms | 27K |
| Reasoning | D | Very compromised (needs ~0.3 GB host RAM) | 23.2 tok/s | 9858 ms | 14K |
How Llama 3.2 3B Instruct (3B params) fits at each quantization level on Intel Arc A370M 4GB (4.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 1.2 GB | Low | D36 |
Q3_K_S | 3 | 1.5 GB | Low | D38 |
NVFP4 | 4 | 1.7 GB | Medium | D39 |
Q4_K_MBest for your GPU | 4 | 1.8 GB | Medium | D39 |
Q5_K_M | 5 | 2.2 GB | High | C41 |
Q6_K | 6 | 2.5 GB | High | C43 |
Q8_0 | 8 | 3.2 GB | Very High | C45 |
F16 | 16 | 6.1 GB | Maximum | F0 |
huggingface-cli download hf-bartowski--llama-3-2-3b-instruct-ggufUpgrade options
Intel Arc B580 12GBBudget pick~$249 MSRP
~$269 MSRP
RTX 2060 6GBBiggest leap~$349 MSRP
Intel Arc A580 8GBIntel upgrade