This is a really cool way Nvidia is handling the memory wall.
This is a AI post classified by AGK Intelligence as AI infra & evals (a launch), kept by the AI Radar because it carries real work, not commentary.
$NVDA $SNDK This is a really cool way Nvidia is handling the memory wall. HBM is the scarcest, priciest thing in an AI factory, so Nvidia is squeezing way more work out of every byte: + Shrink it: 4-bit NVFP4 cuts the KV cache up to 50% vs FP8, so about 2x more context stays on the GPU + Stop recomputing: Dynamo reuses context and moves it from GPU to CPU to storage + Park it: when an agent waits on a tool call, the Vera CPU holds its context in up to 1.5TB of memory + Share it: CMX lets the whole pod reuse context from flash Vera Rubin does up to 30x the throughput per megawatt of GB30
Posted by Ren (54.8k followers) 20 h ago · 38 likes · 9.5k views · view the original post on X. Kept by the AI Radar as AI infra & evals.
More AI work like this
- Enterprise AI is only as useful as the data and business context behind it. — @databricks
- DecodeWall: another benchmark I'm building, focused on concurrent calls. — @ivanfioravanti
- KindlingAI's GLM 5.3 Flash for one Spark recipe is here! 700k total KV on DGX OS (up to… — @mmastrac
- A small model coaching a giant? Aiona checks the held-out tie and SMF’s memory-card… — @MichaelGannotti
- Ranking inference providers six ways — @AdamHoltererer
- Harness adapters training is now live! — @yb2698
- NVIDIA is giving Kimi K3.1 Pro away for free. — @VaibhavSisinty
- The world’s first co-packaged optics Ethernet switch is in full production. — @NVIDIAAIInfra
Every post is read and classified by AGK Intelligence: what it is, which market it belongs to, and whether the link is a real tool. 65.6k posts from 5k X accounts over the last 14 days, 10.4k tools, 19 markets. New posts are collected several times an hour and the full ranking is recomputed every hour — last update 2026-11-06 16:05 UTC. Full method.