MIPS’ Post

The extended version of our unzipFPGA work has been accepted by ACM Transactions on Design Automation of Electronic Systems ! unzipFPGA enhances CNN hardware accelerators with an on-the-fly mechanism of generating weights, drastically reducing the off-chip memory bandwidth requirements during inference. Key features to look out for include: 1) detailed description of the on-the-fly weights generation process, 2) a new automated hardware-aware methodology for selecting the per-layer compression ratios to achieve a balanced performance-accuracy trade-off, and 3) a comparison with optimised embedded GPU implementations with a geo. mean 2.3x higher energy efficiency across diverse CNNs. unzipFPGA was co-led by Javier F. and in close collab with Nicholas Lane. Many thanks to both! Preprint: https://lnkd.in/ebgSdMRy #deeplearning #ai #accelerators

#Technology #Thread #Semiconductor #Manufacturing #DVFS The Semiconductor DVFS AI: 1/ - Dynamic Voltage and Frequency Scaling (DVFS) Has Been The Core Feature Of XPUs. - Primary Helps In Balancing Power And Performance Setting. ---- 2/ - DVFS Works Better If The XPU Can Capture The Features Of Application. - And, Then, Can Accurately Translate It Into System Requirements. ---- 3/ - For This, Several Solutions Have Been Proposed. - A Quick Google Scholar Search Shows More Than 30K Papers. - Including The Two I Was Part Of. ---- 4/ - Now, With The Increase In AI Focused Silicon Architecture And Applications. - The Focus Is Shifting Towards How AI Can Optimize DVFS. - More So, When Chiplet Will Take Over, And The Goal Will Be To Manage DVFS Across Multiple Silicon Areas. ---- 5/ - Intel Is One Company That Has Started Working On AI Powered DVFS. - So Far, There Seems To Be No Significant Difference Compared To What Was Already Proposed By Several Research Papers. - Still, It Would Be Interesting To See How New Age AI Chips Benefit From DVFS AI. ---- #chetanpatil - Chetan Arvind Patil - www.ChetanPatil.in

Arteris at EE Times | Electronic Engineering Times 2023 virtual event. Catch Frank Schirrmeister's presentation to see how NoC IP is the essential connectivity fabric for the acceleration of AI/ML silicon and systems. Join us on the hot #AI topic - registration is still open for this virtual conference. https://bit.ly/3SakrMo #LLMs #artificialintelligence #machinelearning #chatGPT #semiconductor #SystemOnChip #NetworkOnChip

We are always innovating... HPC and AI. Discover how our ultra-high-density networking powers next-gen AI networks: https://lnkd.in/eeH2qYuP

Unlock the future of ML/AI with Alif MCU: Join one of our events and dive into limitless possibilities!