Can I run Llama 3.1 70B on an RTX 5080 16GB?
No: Llama 3.1 70B needs about 55.2 GB at Q4_K_M with 32K tokens of context, 39.2 GB more than the RTX 5080 16GB holds, so it takes 4 of them (64 GB) in one tensor-parallel group. The smallest setup here that holds Llama 3.1 70B at Q4_K_M with 32K is 2× RTX 5090 (64 GB).
No Q4_K_M with 32K tokens of context on 4 cards
- Q4_K_M, 32K
- 55.2 GB
- RTX 5080
- 16 GB, 960 GB/s
- Short by
- 39.2 GB
- Tokens/s
- 35–50 tokens/s
At Q4_K_M with 32K tokens of context on 4 cards it writes about 35–50 tokens/s for one request on 4 RTX 5080 16GB cards.
Llama 3.1 70B on the RTX 5080 as the context fills
One request, FP16 KV cache, 0.5 GB plus 10% overhead; a minus sign is memory missing, and tight is less than 0.5 GB free.
| Context | Q4_K_M | Free | Tokens/s | Q8_0 | Free | Tokens/s |
|---|---|---|---|---|---|---|
| 4K | 45.6 GB | −29.6 GB | — | 78.7 GB | −62.7 GB | — |
| 8K | 47.0 GB | −31.0 GB | — | 80.0 GB | −64.0 GB | — |
| 16K | 49.7 GB | −33.7 GB | — | 82.8 GB | −66.8 GB | — |
| 32K | 55.2 GB | −39.2 GB | — | 88.3 GB | −72.3 GB | — |
| 64K | 66.2 GB | −50.2 GB | — | 99.3 GB | −83.3 GB | — |
| 128K (full) | 88.2 GB | −72.2 GB | — | 121 GB | −105 GB | — |
Llama 3.1 70B on more than one RTX 5080
| Cards | Best at 32K | Tokens/s | Q4_K_M longest | Q4_K_M tokens/s |
|---|---|---|---|---|
| 2× (32 GB) | Nothing fits | — | Does not fit | — |
| 4× (64 GB) | Q4_K_M | 35–50 | 51K | 35–50 |
Tensor parallel with the combined bandwidth, as in vLLM; llama.cpp splits layers by default and runs at about one card's speed.
Other options
- The next smaller setting, Q3_K_M, takes 46.8 GB at 32K, 30.8 GB over the RTX 5080; it does not fit with 0.5 GB to spare even with 1K tokens.
- Smallest setup for Q4_K_M at 32K: 2× RTX 5090 (64 GB).
Run Llama 3.1 70B on the RTX 5080 with llama-server
llama-server -hf bartowski/Meta-Llama-3.1-70B-Instruct-GGUF:Q4_K_M -c 52224Meta-Llama-3.1-70B-Instruct-Q4_K_M.gguf, 42.5 GB, from bartowski/
Questions
Can I run Llama 3.1 70B on an RTX 5080 16GB?
No: Llama 3.1 70B needs about 55.2 GB at Q4_K_M with 32K tokens of context, 39.2 GB more than the RTX 5080 16GB holds, so it takes 4 of them (64 GB) in one tensor-parallel group. The smallest setup here that holds Llama 3.1 70B at Q4_K_M with 32K is 2× RTX 5090 (64 GB).
How fast is Llama 3.1 70B on an RTX 5080 16GB?
At Q4_K_M with 32K tokens of context on 4 cards it writes about 35–50 tokens/s for one request on 4 RTX 5080 16GB cards.
What does a second RTX 5080 16GB change for Llama 3.1 70B?
Even two RTX 5080 16GB cards (32 GB) do not hold Llama 3.1 70B at 32K.
Try other settings in the VRAM calculator or the speed calculator. See also Llama 3.1 70B VRAM requirements, what LLMs an RTX 5080 16GB can run and every pair, or detect your own GPU. Model data checked .