- 102
- 20 553
Memory Fabric Forum
Приєднався 16 жов 2023
The Memory Fabric Forum is a MemVerge initiative designed to accelerate CXL education, solution development, and co-marketing. The initiative is open to like-minded organizations and individuals. The videos on this channel our crowd-sourced from users and vendors like you. Contact frank.berry@memverge.com about publishing your CXL-related videos.
Cloud scale Deployment with CXL Memory
Ahmad Danesh, Samir Rajadnya, Astera Labs, Microsoft - Cloud services provided by companies like Microsoft need to address performance bottlenecks to enable applications to perform more complex computations on larger datasets running in virtualization environments. These applications drive the need for significantly higher memory capacity for data processing and require data centers to increase memory capacity while improving memory utilization to lower TCO. This presentation will explain the challenges of deploying memory in cloud-scale data centers and discuss how CXL memory overcomes these challenges by delivering the necessary RAS, security, and management capabilities.
Переглядів: 104
Відео
CXL 2 0 Use Case Using Both DDR4 & DDR5 on the Same Server to Allow Memory & Bandwidth Scalin
Переглядів 2122 місяці тому
Geof Findley, Montage - Montage Technology, with over 20 years of leadership in memory interface products, is at the forefront of CXL (Compute Express Link) technology, offering solutions that integrate both DDR4 and DDR5 memory into data center environments. CXL 2.0 enables high-bandwidth, low-latency interconnects between CPUs and memory expanders, allowing for scalable memory expansion and p...
Enhance AI and Database Performance Experience Microns CZ122 at OCP!
Переглядів 772 місяці тому
Vishnu Balraj, Micron - Join us to discover how Microns CZ122, integrated into a RAG pipeline, can revolutionize your AI inference by significantly reducing latency and response times. Experience the power of combining CZ122 with Famfs (Fabric Attached Memory File System) to supercharge your database performance, delivering higher sustained operations per second and improved p99 latency. Learn ...
Enhancing CXL Memory RAS through a Forgotten Coding Theory
Переглядів 582 місяці тому
Tong Zhang, ScaleFlux - As DRAM technology advances towards 10nm and beyond, fault tolerance becomes increasingly crucial for data centers. DRAM fault tolerance heavily relies on error correction coding (ECC), with Reed-Solomon (RS) code being widely adopted. This presentation will introduce a novel RS decoding solution that significantly enhances DRAM fault tolerance compared to the state-of-t...
Expanding CXL Software Ecosystem through HMSDK on Linux
Переглядів 1262 місяці тому
Honggyu Kim, SK hynix - With the advent of CXL memory, some systems are becoming more heterogeneous having memory systems with different latency and bandwidth characteristics. They are usually handled as different NUMA nodes in separate memory tiers and CXL memory is used as slow tiers because of its protocol overhead compared to DIMM attached local DRAM. In this kind of systems, we need to be ...
Exploring CXL Memory Disaggregation Use Cases and System Benefits
Переглядів 1302 місяці тому
Jungmin Choi,Myoungseo (Matthew) Kim, SK hynix - In this session, we dive into powerful use cases for DCD (Dynamic Capacity Device) based memory pooling and sharing specified in CXL 3.1 that can reduce costs from a datacenter operations perspective. Each host connected to the memory pool can manage resources flexibly through dynamic memory allocation. We also address use cases for system benefi...
Mature at Scale Memory Fabrics for all Performance and Price Points
Переглядів 3032 місяці тому
Rob Davis, NVIDIA - As data-centric applications shift to AI, selecting the right memory fabric becomes essential for achieving optimal performance and cost-efficiency. This talk provides a comprehensive overview of mature in market memory fabric technologies, focusing on Ethernet, InfiniBand, and NVLink, each catering to different performance and price requirements. It will start by examining ...
Memory Fabric Technology Landscape
Переглядів 2542 місяці тому
Charles Fan, MemVerge - The size of AI models and deployment of AI infrastructure has exploded, exposing gaps in technology and a land rush to fill the gaps. Dr. Charles Fan will identify key memory fabric technologies needed for the growth of AI to continue and make predictions about how the memory fabric technology landscape will change.
Future of AI Networks UAL and Ultra Ethernet
Переглядів 2892 місяці тому
J Metz,Kurtis Bowman, AMD - AI Networks are pushing architectural limits in data centers. The question of data movement between the various pieces of the puzzle have brought many interested groups together to address the problems. Two organizations recently have been created to solve the problems of “scale up and “scale out fabrics - UALink and Ultra Ethernet Consortium, respectively. This talk...
Unified Memory Framework unified API for diverse memory technologies
Переглядів 1262 місяці тому
Sergei Vinogradov, Igor Chorazewicz - Intel - SW adoption is required to leverage emergent memory technologies fully. A middleware layer between applications and system level (OS and drivers) plays a significant role in achieving that. In this talk, we will present Unified Memory Framework (UMF) - an open-source project that provides consistent mechanisms for SW developers to work with memory h...
Marvell Structera Overcoming Computings Memory Challenges with Optimized CXL Devices
Переглядів 1192 місяці тому
Khurram Malik - AI and cloud applications often require far more memory capacity and bandwidth than standard servers can provide, leading to lower than expected performance, longer time-to-completion and an inefficient use of resources. The new Marvell Structera product line of CXL devices enables data center operators to circumvent these problems by adding more memory and-or computing capacity...
VMware Memory Vision for Real World Applications
Переглядів 1042 місяці тому
sudhir balasubramanian, Arvind Jagannath, VMware by Broadcom - VMware has been on an evolving journey on memory innovations mainly first with persistent memory, then with memory tiering, and is now extending that with CXL. CXL provides an opportunity for VMware (by Broadcom) to further improve on performance, and provide further customer benefits such as TCO reduction, server consolidation, and...
SMART Modular: CXL Memory Expansion Advantages
Переглядів 3084 місяці тому
Torry Steed of SMART Modular provides an overview of the company's CXL Add-In Cards (AICs) and memory modules, and configuration guidelines for adding CXL memory to a server.
Intel Compute Express Link™ Enablement
Переглядів 954 місяці тому
Anil Godbole of Intel provides and update on the company's support for CXL.
Memory Machine™ Software for CXL
Переглядів 2484 місяці тому
MemVerge Director of Product Management Steve Scargall provides and overview of Memory Machine X, the company's software that can deliver intelligent tiering for CXL memory inside a server, and memory sharing for fabric-attached CXL memory. Increasing GPU utilization 6:53 Accelerating RAG pipeline 9:43 Boosting performance of transaction processing 11:50 Speeding analytic processing with GISMO ...
MSI: S2301 CXL Memory Expansion Server
Переглядів 1354 місяці тому
MSI: S2301 CXL Memory Expansion Server
Accelerating AI & ML with CXL-Attached Memory
Переглядів 4134 місяці тому
Accelerating AI & ML with CXL-Attached Memory
CXL 2.0 Use Case: Using Both DDR4 & DDR5 on the Same Server
Переглядів 1484 місяці тому
CXL 2.0 Use Case: Using Both DDR4 & DDR5 on the Same Server
CXL Switch for Scalable & Composable Memory Pooling/Sharing
Переглядів 2464 місяці тому
CXL Switch for Scalable & Composable Memory Pooling/Sharing
Enabling New Memory Applications Using CXL IP
Переглядів 1024 місяці тому
Enabling New Memory Applications Using CXL IP
Hyperscale Composable Memory Systems with Dynamically Adjusting Compressed Tier
Переглядів 544 місяці тому
Hyperscale Composable Memory Systems with Dynamically Adjusting Compressed Tier
Liquid-Markets-Solutions: Introduction to UberNIC
Переглядів 2224 місяці тому
Liquid-Markets-Solutions: Introduction to UberNIC
You Don't Know Jack: CXL Fabric Orchestration and Management
Переглядів 1064 місяці тому
You Don't Know Jack: CXL Fabric Orchestration and Management
NVMe Over CXL: How CXL Lets Us Do Controller Memory Buffers the Right Way
Переглядів 4334 місяці тому
NVMe Over CXL: How CXL Lets Us Do Controller Memory Buffers the Right Way
FleX: Bringing CXL to the Motherboard
Переглядів 1544 місяці тому
FleX: Bringing CXL to the Motherboard
CXL Native Memory: Do We Really Need DDR?
Переглядів 3044 місяці тому
CXL Native Memory: Do We Really Need DDR?
VMware Memory Tiering: Customer-Ready Today
Переглядів 1624 місяці тому
VMware Memory Tiering: Customer-Ready Today
NVIDIA Networking for HPC, AI, and Accelerated IO
Переглядів 1,9 тис.4 місяці тому
NVIDIA Networking for HPC, AI, and Accelerated IO
At-scale Systems: Interconnecting Massively Parallel xPUs
Переглядів 2854 місяці тому
At-scale Systems: Interconnecting Massively Parallel xPUs
Excellent presentation! It's always the auxiliary hardware that's the long-term dark horse. Also, where breakthrough innovation happens, away from hype.
The content is excellent, but it feels like the presenter is simply reading text written by someone else. :)
Wonderful to learn about what Montage has proven with their MXC. The value proposition for easily expanding bandwidth and capacity is compelling. Well done.
Will all the videos be uploaded?
They should all be up by tomorrow!
very nice presentation!!
🎉
👏 'promosm'