Snagged 32GB RAM at a decent price.
My llm machine is now chugging through some representative benchmarks that the model wasn’t trained on.
Dual channel makes quite the difference if you have to offload experts to RAM.
Next is sniping a 7900 XTX. That should give me Qwen3.8 at current Qwen3.6 speeds with comparable precision.
HERMETICVM
npub1rf…ck3t8
2026-10-28 20:09:09 UTC
Author Public Key
npub1rfw075gc6pc693w5v568xw4mnu7umlzpkfxmqye0cgxm7qw8tauqfck3t8Published at
2026-10-28 20:09:09 UTCEvent JSON
{
"id": "258bcadad3254b9ef86be23f21f0fa8e940693dee816335f3865cde038106de9",
"pubkey": "1a5cff5118d071a2c5d46534733abb9f3dcdfc41b24db0132fc20dbf01c75f78",
"created_at": 1793218149,
"kind": 1,
"tags": [],
"content": "Snagged 32GB RAM at a decent price. \n\nMy llm machine is now chugging through some representative benchmarks that the model wasn’t trained on. \n\nDual channel makes quite the difference if you have to offload experts to RAM.\n\nNext is sniping a 7900 XTX. That should give me Qwen3.8 at current Qwen3.6 speeds with comparable precision.",
"sig": "40cfb82e742ae4d96c2edb13a7cb00ad8595bea47732810cd06a9a4aa48911bb12933f5657ab9161a850f5153e1381c25b94c194b83823c2cbc6a5f2a0646bc4"
}