Embedded FPGA (eFPGA) can reduced memory
bandwidth requirements by more than 10x and allows efficient
execution of future operators and activation functions
MOUNTAIN
VIEW, Calif., July 2, 2024
/PRNewswire/ -- Flex Logix® Technologies, Inc., the leading
supplier of embedded FPGA (eFPGA) IP and reconfigurable DSP/SDR/AI
solutions, today announced additional applications for embedded
FPGA to improve the value proposition for AI Accelerators.
First, memory bandwidth – the scarcest resource in AI
Accelerators, especially in the cloud, where model
weights/parameters exceed 100 billion and HBM memory is expensive
and scarce. Techniques for saving memory bandwidth are evolving
faster than hardware.
"Embedded FPGA (eFPGA) can enable innovations in sub-INT4 data
and weight representations (e.g. ternary, 2 bit, 3 bit, mixed or
mat-mul free) to be converted on the fly by eFPGA into existing
TPUs," said Cheng Wang, Flex Logix CTO & SVP Software +
Architecture. "This can also be mixed with innovations in sparsity
that can further reduce the memory bandwidth requirements.
Aggregate memory bandwidth reduction can be up to 16x."
Second, higher performance. AI models are rapidly evolving. With
most TPUs, new operators and activation functions must be handled
by a much slower processor. eFPGA can be used to run the new
operators and activation functions at much higher performance.
Flex Logix is already using these concepts in its own InferX AI
optimized for edge vision AI models and DSP.
Get more information on EFLX eFPGA and a block diagram of weight
memory reduction at www.flex-logix.com.
About Flex Logix
Flex Logix is a reconfigurable computing company providing
leading edge eFPGA, DSP/SDR and AI Inference solutions for
semiconductor and systems companies. Flex Logix eFPGA enables
volume FPGA users to integrate the FPGA into their companion SoC,
resulting in a 5-10x reduction in the cost and power of the FPGA
and increasing compute density which is critical for
communications, networking, data centers, microcontrollers and
others. Its scalable DSP/SDR/AI is the most efficient,
providing much higher inference throughput per square millimeter
and per watt. Flex Logix supports process nodes from 180nm to 7nm,
with 5nm, 3nm and 18A in development. Flex Logix is
headquartered in Mountain View,
California and has an office in Austin, Texas. For more information,
visit https://flex-logix.com.
For general information on InferX and EFLX product lines, visit
our website at this link.
Copyright 2024. All rights reserved. Flex Logix and EFLX are
registered trademarks and INFERX is a trademark of Flex Logix,
Inc.
View original content to download
multimedia:https://www.prnewswire.com/news-releases/flex-logix-boosts-ai-accelerator-performance-and-long-term-efficiency-302187642.html
SOURCE Flex Logix Technologies