AMD Instinct MI350X 288GBBudget pick
C40.6 tok/s decode
~$8,000 MSRP
Can it run?
Runs with offload (needs ~6.6 GB host RAM)
Using Q4_K_M in Ollama
Fit status
Runs with offload (needs ~6.6 GB host RAM)
Decode
45.6 tok/s
TTFT
4245 ms
Safe context
15K
Memory
201.2 GB / 192.0 GB
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Agentic Coding | F | Too heavy | 46.7 tok/s | 6033 ms | 26K |
| Chat | C | Runs with offload | 46.7 tok/s | 2262 ms | 8K |
| Coding | C | Runs with offload (needs ~6.6 GB host RAM) | 45.6 tok/s | 4245 ms | 15K |
| RAG | F | Too heavy | 46.7 tok/s | 7541 ms | 26K |
| Reasoning | C | Runs with offload (needs ~6.6 GB host RAM) | 45.6 tok/s | 5017 ms | 15K |
How K EXAONE 236B A23B (236B params) fits at each quantization level on B100 192GB (192.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 92.0 GB | Low | D39 |
Q3_K_S | 3 | 115.6 GB | Low | C42 |
NVFP4 | 4 | 132.2 GB | Medium | C44 |
Q4_K_MBest for your GPU | 4 | 144.0 GB | Medium | C44 |
Q5_K_M | 5 | 169.9 GB | High | C44 |
Q6_K | 6 | 193.5 GB | High | F0 |
Q8_0 | 8 | 252.5 GB | Very High | F0 |
F16 | 16 | 483.8 GB | Maximum | F0 |
Upgrade options
~$8,000 MSRP