GeeCoolGuest
GeeCoolThe 13790F is the China-exclusive 16-core/24-thread part at 5.2GHz with 33MB of L3 (3MB over the 13700F's 30MB), VNNI INT8 and no iGPU. For AI, the cache's landing point: a bit more hit room for embedding vocabularies and hot small-model weights, squeezing a hair of latency off single-stream inference — mostly below the threshold of perception.
Its real value is the price anchor: exclusive models usually carry channel pricing, so take it when the same config is cheaper than the 13700F. Don't pay extra for 3MB of cache — AI data work runs on memory and cores; cache is garnish, not a deciding factor.