pp-head: 2x200700 f16 KV (ctx 401400, parallel 2), split 0.40,0.60 = max 3090 capacity

scripts/frank-pp-head.env

@@ -25,16 +25,16 @@ LISTEN_PORT=7770
 RPC_SERVERS=192.168.0.235:50052
 
 # Model split fractions, one per device in order [RPC0, iGPU].
-# 1,1 = 50/50: the 3090 gets the first ~32 layers, the Strix the rest.
-# Leave empty for the default proportional-to-free-memory split; note the
-# Strix reports ~96 GiB (GTT ceiling) vs the 3090's ~24 GB, so the default
-# lands around 20% of layers on the 3090 and 80% on the Strix.
-TENSOR_SPLIT=1,1
+# 0.40,0.60 = 40% of layers on the 3090 (~26 layers) - the maximum that
+# fits its 24 GB card with 2x200700 f16 KV (see the KV math in
+# docs/Pipeline Parallelism/Qwen3.8 PP Deployment.md). The 3090 is 2-3x
+# faster per layer, so shifting weight to it is also a speed win.
+TENSOR_SPLIT=0.40,0.60
 
-# Context pool. 131072 single slot keeps the KV small (~4.5 GB total at
-# q8_0) on both sides; raise for high-context tests (see docs).
-CTX_SIZE=131072
-PARALLEL=1
+# Context pool. 401400 total = 2 parallel slots x 200700 per slot
+# (n_ctx_seq = n_ctx / n_seq_max, llama-context.cpp:293).
+CTX_SIZE=401400
+PARALLEL=2
 
 # Speculative decoding. Matches production frank-default-large: MTP (bundled
 # head, n_max 4 per the published tuning curve) with the ngram-mod fallback.
@@ -50,9 +50,11 @@ SPEC_NGRAM_MOD_N_MIN=48
 SPEC_NGRAM_MOD_N_MAX=64
 SPEC_NGRAM_MOD_N_MATCH=24
 
-# KV on GPU, quantized. Keep K == V so Flash Attention stays enabled.
-CACHE_TYPE_K=q8_0
-CACHE_TYPE_V=q8_0
+# KV on GPU, unquantized (f16). 16 full-attention layers x 2 (K,V) x 4 heads
+# x 256 dim x 2 bytes = 64 KiB/token -> 401400 tokens = 24.5 GiB total.
+# Keep K == V so Flash Attention stays enabled.
+CACHE_TYPE_K=f16
+CACHE_TYPE_V=f16
 
 # Batch sizing: production-verified values (2026-08-14).
 BATCH_SIZE=8192