Memory Wall: A Bottleneck of the Data Center
The amount of generated data has exploded in recent years. And AI technology advances are also primarily based on data.
Computer systems are facing the limits of increasing bandwidth and memory capacity, a critical factor for handling data called a Memory Wall.
This could be a fundamental problem of Von Neumann’s architecture. MetisX will break the Memory Wall problem with MetisX’s Data Domain Specific Architecture.
This could be a fundamental problem of Von Neumann’s architecture. MetisX will break the Memory Wall problem with MetisX’s Data Domain Specific Architecture.
Data Domain Specific Architecture, Data-Centric Computing
Moore’s Law and Dennard’s scaling is already over. Prof. John Hennessy and David Patterson mentioned the solution in their book.
“Minor twists to existing cores may get us 10% improvements, but if we want order-of-magnitude
improvements while offering programmability, …… we need a drastic change in computer architecture
from general-purpose cores to domain-specific architecture (DSAs).”
MetisX tries to solve this with its own data domain-specific architecture specialized in large-scale data because it is no longer easy to keep up with data growth with traditional computer architecture.
Through our data domain-specific architecture, we want to contribute to shifting the hegemony of CPU-centric computer systems to data-centric.
MetisX tries to solve this with its own data domain-specific architecture specialized in large-scale data because it is no longer easy to keep up with data growth with traditional computer architecture.
Through our data domain-specific architecture, we want to contribute to shifting the hegemony of CPU-centric computer systems to data-centric.
CXL (Compute Express Link) and CXL Intelligent Memory
CXL is a next-generation data center interconnection standard led by Intel for CPU, memory, and accelerator.
MetisX develops smart, fast, and cost-efficient intelligent memory solutions based on CXL.
Our intelligent memory accelerates large-scale data processing applications that take a long time with CPUs, such as database analytics, AI data preprocessing, and NGS DNA analysis by tens of times the level of conventional data processing.
We expect these solutions to improve data center operating costs and time dramatically.
Our intelligent memory accelerates large-scale data processing applications that take a long time with CPUs, such as database analytics, AI data preprocessing, and NGS DNA analysis by tens of times the level of conventional data processing.
We expect these solutions to improve data center operating costs and time dramatically.
MetisX HW-SW Co-Optimized Architecture
COPYRIGHT ⓒ MetisX Co., Ltd. ALL RIGHTS RESERVED.