AMD Gorgon Halo 192GB vs AMD Instinct MI300A for local LLMs
0 reports on the AMD Gorgon Halo 192GB and 0 on the AMD Instinct MI300A
Which is faster for local LLMs?
Nobody has reported the same model at the same quant on both cards yet, so there's no measured head-to-head. Memory bandwidth, which caps how fast a card can write, is 19.4x higher on the AMD Instinct MI300A (273 GB/s on the AMD Gorgon Halo 192GB, 5,300 on the AMD Instinct MI300A). The estimates below compare them on the models people run most.
- Memory
- 192GB unified
- Memory bandwidth
- 273 GB/s
- FP16 compute
- 61.4 TFLOPS
- Reports
- 0
- Memory
- 128GB unified
- Memory bandwidth
- 5,300 GB/s
- FP16 compute
- 980.6 TFLOPS
- Reports
- 0
Estimated by the calculator
The quant the calculator recommends for each card and its estimated speed at a 32,768-token context with 32 GB of system RAM, from memory bandwidth and the model's shape.
| Model | AMD Gorgon Halo 192GB | AMD Instinct MI300A |
|---|---|---|
| Qwen3.8 27B | 2.8 t/sFP16 | 54.0 t/sFP16 |
| Qwen3.8 125B · 6B active | 9.5 t/sQ8_0 | 193.8 t/sQ6_K |
| DeepSeek V4 Flash 284B · 13B active | 12.5 t/sQ4_K_M | 514.6 t/sQ2_K |
| Qwen3.6 35B · 3B active | 10.0 t/sFP16 | 194.3 t/sFP16 |
| Qwen3.6 27B | 2.8 t/sFP16 | 54.0 t/sFP16 |
| Gemma 4 26B · 4B active | 7.5 t/sFP16 | 145.8 t/sFP16 |
Change the context, RAM or model in the calculator: AMD Gorgon Halo 192GB · AMD Instinct MI300A
Every report on each card, with its source: AMD Gorgon Halo 192GB · AMD Instinct MI300A