AMD Instinct MI300A vs NVIDIA RTX 3090 for local LLMs
0 reports on the AMD Instinct MI300A and 165 on the NVIDIA RTX 3090
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 5.7x higher on the AMD Instinct MI300A (5,300 GB/s on the AMD Instinct MI300A, 936 on the NVIDIA RTX 3090). The estimates below compare them on the models people run most.
AMD Instinct MI300A
- Memory
- 128GB unified
- Memory bandwidth
- 5,300 GB/s
- FP16 compute
- 980.6 TFLOPS
- Reports
- 0
NVIDIA RTX 3090
- Memory
- 24GB
- Memory bandwidth
- 936 GB/s
- FP16 compute
- 142 TFLOPS
- Reports
- 165
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 Instinct MI300A | NVIDIA RTX 3090 |
|---|---|---|
| Qwen3.8 27B | 54.0 t/sFP16 | 33.9 t/sQ4_K_M |
| Qwen3.8 125B · 6B active | 193.8 t/sQ6_K | 19.4 t/sQ2_K, experts in RAM |
| DeepSeek V4 Flash 284B · 13B active | 514.6 t/sQ2_K | doesn't fit at these settings |
| Qwen3.6 35B · 3B active | 194.3 t/sFP16 | 120.1 t/sQ4_K_S |
| Qwen3.6 27B | 54.0 t/sFP16 | 33.9 t/sQ4_K_M |
| Gemma 4 26B · 4B active | 145.8 t/sFP16 | 76.4 t/sQ5_K_S |
Change the context, RAM or model in the calculator: AMD Instinct MI300A · NVIDIA RTX 3090
Every report on each card, with its source: AMD Instinct MI300A · NVIDIA RTX 3090