Remove Pre 7.5 support from CUDA 13

Author: robballantyne

.github/workflows/build-cuda.yml

@@ -79,7 +79,7 @@ jobs:
           - cuda_version: '13.0.1'
             cuda_version_short: '13.0'
             cuda_tag: '13.0.1-cudnn-devel-ubuntu22.04'
-            architectures: '61-virtual;70-virtual;75-virtual;80-virtual;86-virtual;89-virtual;90-virtual;100-virtual;120-virtual'
+            architectures: '75-virtual;80-virtual;86-virtual;89-virtual;90-virtual;100-virtual;120-virtual'
 
     steps:
       - name: Checkout
@@ -244,7 +244,7 @@ jobs:
             - CUDA 12.6 - Architectures: 6.1, 7.0, 7.5, 8.0, 8.6, 8.9, 9.0
             - CUDA 12.8 - Architectures: 6.1, 7.0, 7.5, 8.0, 8.6, 8.9, 9.0, 10.0, 12.0
             - CUDA 12.9 - Architectures: 6.1, 7.0, 7.5, 8.0, 8.6, 8.9, 9.0, 10.0, 12.0
-            - CUDA 13.0 - Architectures: 6.1, 7.0, 7.5, 8.0, 8.6, 8.9, 9.0, 10.0, 12.0
+            - CUDA 13.0 - Architectures: 7.5, 8.0, 8.6, 8.9, 9.0, 10.0, 12.0
             
             ## Architecture Reference
             - 6.1: Titan XP, Tesla P40, GTX 10xx

README.md

@@ -34,7 +34,7 @@ The official llama.cpp repository does not provide pre-built CUDA binaries. This
 | 10.0 | B200 | ❌ | ✅ |
 | 12.0 | RTX Pro series, RTX 5000 series | ❌ | ✅ |
 
-**Note:** Blackwell architectures (10.0 & 12.0) require CUDA 12.8 or later.
+**Note:** Compute Capability < 7.5 support ends with CUDA 12.9 
 
 ## Usage