Recovering 4.1 GiB of RAM on NVIDIA DGX Spark: KV pool jumps from 262k to 938k tokens
marian_nmt · x · 2026-10-02
Marcin Junczys-Dowmunt's full write-up: DGX Spark ships 128 GB unified memory but Linux exposes less. Installing NVIDIA's official 64 KiB kernel (+2.098 GiB, keeping the 4 KiB kernel as GRUB fallback) plus reclaiming the 2046 MiB headless display reservation yields 4.096 GiB total, growing a GLM-5.3-Flash KV pool from 262,144 to 937,984 tokens (3.58x). Includes install commands, before/after MemTotal numbers, and the emjotde/spark-ram-reclaim repo.
Related event: Reclaiming 4.1 GiB of Hidden Memory on NVIDIA DGX Spark(7 posts)→
More from Infra
- LithosAI launches LithosBox: millisecond snapshot-and-fork sandboxes for AI agents — JiaZhihao · 2026-10-03
- Wish list: a Qwen4 27B with 100B+ Engram offloaded to RAM and NVMe for local users — casper_hansen_ · 2026-10-03
- State of Local AI 2026: one gaming GPU now matches the world's best model from Feb 2026 — Scobleizer · 2026-10-03
- Leak: DeepSeek V4.1 Pro ~2T params, trained sparse from scratch with no instabilities — teortaxesTex · 2026-10-03
- Stas Bekman demos his cluster-audit AI skill running on an AWS cluster — StasBekman · 2026-10-03
- Running 180B Qwen3.8-Flash-Next at 40-50 t/s on 64GB RAM + 16GB VRAM with Strata — danamir_ · 2026-10-03