Update ollama.md

Author: lwindolf

Cheat Sheets/DevOps Linux/ollama.md

@@ -4,7 +4,34 @@
     ollama pull <model>
     ollama run <model>
 
-## Simple usage with curl
+## Checking HW
+
+Find out the amount of VRAM
+
+    lspci -v | grep -A 20 VGA
+
+In the output watch for `Memory at`:
+
+    0000:00:02.0 VGA compatible controller: Intel Corporation TigerLake-LP GT2 [Iris Xe Graphics] (rev 01) (prog-if 00 [VGA controller])
+    	[...]
+    	Memory at 6078000000 (64-bit, non-prefetchable) [size=16M]
+    	Memory at 4000000000 (64-bit, prefetchable) [size=256M]         <<<--- 256 MB VRAM not really good :-)
+    	[...]
+
+Check for AVX 512 support
+
+    cat /proc/cpuinfo | grep avx512
+
+General performance to decide which models to run locally
+
+| GPU         | RAM   | VRAM        | Models / Quantization               |
+|-------------|-------|-------------|-------------------------------------|
+| no          | 16GB  | <8 GB       | smallest 3B models only, 4-bit GGUF |
+| yes         | 32GB  | 8 GB        | up to 7B models, q2-3               |
+| yes         | 128GB | 16 GB       | up to 13B models, GGUF/EXL2         |
+| yes         | 128GB | 24 GB       | up to 30B models                    |
+
+## Simple API usage with curl
 
 Get available model names with