AMD64 Family 23 Model 104 Stepping 1, AuthenticAMD LLM benchmark
On the AMD64 Family 23 Model 104 Stepping 1, AuthenticAMD, measured single-user chat decode speeds are: ~1.2B models at 23.9 tok/s (Q8_0, ollama; Measured, 1 result, source run). Short-chat workload, local runtimes, dense models only; quantization, backend and settings vary between submissions, so check each row in the tables below.
Fastest single result in these submissions (any model size)
23.9 decode tok/s
llama3.2 (1.2B) via ollama (Q8_0). Workload: chat-short. Reported hardware: AMD64 Family 23 Model 104 Stepping 1, AuthenticAMD (6c) + 15GB. see full run
llama3.2
| Workload | Model size | Reported hardware | Backend | Quant | decode tok/s | prefill tok/s | TTFT | Run |
|---|---|---|---|---|---|---|---|---|
| chat-short | 1.2B | AMD64 Family 23 Model 104 Stepping 1, AuthenticAMD (6c) + 15GB | ollama@0.34.0 | Q8_0 | 23.88tok/s | 39.41tok/s | 3,172ms | r_b1rn5gi6b3b |
| chat-long | 1.2B | AMD64 Family 23 Model 104 Stepping 1, AuthenticAMD (6c) + 15GB | ollama@0.34.0 | Q8_0 | 14.69tok/s | 93.91tok/s | 33,563ms | r_b1rn5gi6b3b |
| concurrent-decode | 1.2B | AMD64 Family 23 Model 104 Stepping 1, AuthenticAMD (6c) + 15GB | ollama@0.34.0 | Q8_0 | 19.96tok/s | no data | no data | r_b1rn5gi6b3b |
| agent-trace | 1.2B | AMD64 Family 23 Model 104 Stepping 1, AuthenticAMD (6c) + 15GB | ollama@0.34.0 | Q8_0 | 16.80tok/s | 203.2tok/s | 7,977ms | r_b1rn5gi6b3b |