Perf: Add cmp-tune insane profile, update Proxmox LXC privileged container docs, and update benchmarks to 28.8 tok/s with PCIe Gen2 and unlocked 610.43.03 driver

scripts/start-server.sh

@@ -24,7 +24,7 @@ exec /opt/llama.cpp/build-nccl/bin/llama-server \
   --split-mode tensor \
   --flash-attn on \
   --batch-size 1024 \
-  --ubatch-size 256 \
+  --ubatch-size 512 \
   --jinja \
   --threads 12 \
   --cache-type-k q4_0 \

tuning/cmp-tune.conf

@@ -0,0 +1,13 @@
+# /etc/cmp-tune.conf - Profile-driven tuning for CMP 50HX / CMP 90HX
+
+[efficient]
+power_w = 170
+core_offset = 200
+mem_offset = 1000
+clock_max = 1950
+
+[insane]
+power_w = 220
+core_offset = 250
+mem_offset = 1200
+clock_max = 2050