Nvidia’s new technique cuts LLM reasoning costs by 8x without losing accuracy

Nvidia researchers developed dynamic memory sparsification (DMS), a technique that compresses the KV cache in large language models by up to 8x while maintaining reasoning accuracy — and it can be ...

John Carmack proposes fiber-optic loops as high-speed AI cache

The thought experiment began with a number. Single-mode fiber optics can now transmit data at 256 terabits per second over 200 kilometers. Based on that capacity, ...

'Observational memory' cuts AI agent costs 10x and outscores RAG on long-context benchmarks

As AI agents move into production, teams are rethinking memory. Mastra’s open-source observational memory shows how stable ...

DRAM prices expected to double in Q1 as AI ambitions push memory fabs to their limit

TrendForce this week revised its estimates with analysts now predicting DRAM contract pricing will surge by 90–95 percent QoQ ...

AMD's unannounced dual-CCD 3D V-Cache monster chip appears in official AMD docs

The new AMD Ryzen 9 9950X3D2 has shown up in an EEC listing from Friday. When you purchase through links on our site, we may earn an affiliate commission. Here’s how it works. Despite the fact that ...
GitHub

CHIME: A Cache-Efficient and High-Performance Hybrid Index on Disaggregated Memory

We strongly recommend you to run CHIME using the r650 instances on CloudLab as the code has been thoroughly tested there. We haven't done any test in other hardware environment. If you want to ...
IEEE

ICGMM: CXL-enabled Memory Expansion with Intelligent Caching Using Gaussian Mixture Model

Abstract: Compute Express Link (CXL) emerges as a solution for wide gap between computational speed and data communication rates among host and multiple devices. It fosters a unified and coherent ...
IEEE

Caching Optimization in User-Centric Networks: A Stochastic Geometry Perspective

Abstract: User-centric network (UCN) is regarded as a promising technology to provide users with high network capacity through a cluster of cooperative base stations (BSs). However, to support many-to ...