Add B300 GPU preset and GLM-5-NVFP4 on 8xB300 config

Add NVIDIA B300 (Blackwell Ultra) to hardware presets: same GB202 die as
B200 (2.25 PFLOPS BF16 dense) but with HBM3e 12-Hi stacks (288 GB,
12 TB/s — 50% more capacity and bandwidth than B200).

Add nvidia/GLM-5-NVFP4 HuggingFace config.json and a matching simulation
config for 8xB300: FP4 weights (~372 GB) leave ~1.9 TB for KV cache,
yielding 82k blocks per instance (3.8x more than the BF16-on-B200 setup).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

src/hardware_presets.rs

@@ -20,6 +20,7 @@ pub const AVAILABLE: &[&str] = &[
     "a100-80gb",
     "a100-40gb",
     "b200",
+    "b300",
     "2xh100",
     "4xh100",
     "8xh100",
@@ -32,6 +33,9 @@ pub const AVAILABLE: &[&str] = &[
     "2xb200",
     "4xb200",
     "8xb200",
+    "2xb300",
+    "4xb300",
+    "8xb300",
 ];
 
 /// Resolve a hardware preset by name.
@@ -48,6 +52,7 @@ pub fn resolve(name: &str) -> Option<HardwareConfig> {
         "a10080gb" | "a100" => Some(make_config(count, &A100_80GB)),
         "a10040gb" => Some(make_config(count, &A100_40GB)),
         "b200" => Some(make_config(count, &B200)),
+        "b300" => Some(make_config(count, &B300)),
         _ => None,
     }
 }
@@ -121,6 +126,13 @@ const B200: GpuBase = GpuBase {
     pcie_gen: 6,
 };
 
+const B300: GpuBase = GpuBase {
+    flops: 2.25e15,   // 2250 TFLOPS BF16 dense (same GB202 die as B200)
+    mem_bw: 12.0e12,  // 12 TB/s HBM3e 12-Hi (50% more than B200 8-Hi)
+    hbm: 288.0e9,     // 288 GB HBM3e 12-Hi
+    pcie_gen: 6,
+};
+
 /// Build a [`HardwareConfig`] from a base GPU spec × TP count.
 ///
 /// Compute, HBM bandwidth, and HBM capacity scale linearly with `n`.