mirror of
https://github.com/ikawrakow/ik_llama.cpp.git
synced 2026-02-21 13:44:10 +00:00
21 lines
682 B
Markdown
21 lines
682 B
Markdown
### 🔀 [#122](https://github.com/ikawrakow/ik_llama.cpp/pull/122) - Q6_0_R4
|
|
|
|
| **Author** | `ikawrakow` |
|
|
| :--- | :--- |
|
|
| **State** | ❌ **Closed** |
|
|
| **Created** | 2024-12-03 |
|
|
| **Updated** | 2024-12-03 |
|
|
|
|
---
|
|
|
|
#### Description
|
|
|
|
Follow up of #118, #119, #120, #121 for `Q6_0`.
|
|
|
|
Here is PP-512 for LLaMA-3.1-8B on `Zen4` (Risen-7950X), `ARM_NEON` (M2-Max) and `AVX2` (Ryzen-5975WX)
|
|
|
|
| Platform | Threads | Q6_0 | Q6_0_R4 | Speedup |
|
|
| ---: | ---: | ---: | ---: | ---: |
|
|
| ARM_NEON | 8 | 73.21 ± 1.10 | 94.96 ± 0.90 | 1.297 |
|
|
| Zen4 | 16 | 159.04 ± 0.58 | 257.25 ± 0.26 | 1.638 |
|
|
| AVX2 | 32 | 174.19 ± 0.58 | 231.53 ± 0.60 | 1.329 | |