This post will characterize importance of in memory compute in current state of infrastructure mainly for large, distributed AI training and Inference workloads especially when Moore law is marching to its end of life. Technology shrink past 2nm is getting costlier and harder due to CMOS lithography sophistication. AI workloads as per current generation require […]
Author: skapil
High Radix Data Center Switches Architecture and Feature Comparison for Scale-up and Scale-out Commercial Offerings
Current commercial offering uses high radix switches used in Rack scale computing used predominantly in AI servers in data center. Radix in case of a switch is the measure of number of external ports on a switch, high radix switches in general provide 128 to 256 external ports per switch which tends to reduce the […]
Security of interactions in AI Agentic workflows spawning MCP client and Server in addition to accessing external services /tools APIs from MCP server marketplace
This post is motivated by security aspect of agentic workflows growing rapidly and happen to access third party services/tools code from marketplaces to run in local machine. MCP (model context protocol) introduced by Anthropic and specified here https://github.com/modelcontextprotocol needs to be understood from security perspective. AI agents gain steam and can be used for variety […]
Understanding CPU and GPU Server Thermals, Evaluation Frameworks, Power/Thermal Stress Generation Through Virus Programs and common thermal Management algorithms
There is lot of cost imperatives in managing and evaluating thermals at system on chip level, server level and at the data center level. Here in this post, I intend to touch upon only for system on chip and server level. I intend to discuss power feed and thermal interactions, metrics used in thermal monitoring, […]
Characterizing Blockchain Server and its components like Data-structures and Hashes
This post discusses characterization of primitives needed for another important use case of block chain servers and highlighting need for new IPs to differentiate from commodity servers. As a motivation and relevance of this exercise, per finance.yahoo.com, finance sector is the largest adopter of blockchain infrastructure including servers, driven by demand in payments, asset tokenization, […]
Post Quantum Cryptography Ciphers for Encryption/Encapsulation and Decryption/Decapsulation for AI, HPC and Blockchain Servers
This post is in continuation of previous post on Sept-2,2025 related to Pre-quantum ciphers used for data in motion and rest. Post Quantum cryptography is taking shape in modern AI servers though it is still in early stages. There are two specific areas used for encryption and decryption like TLS (SSL) Protocols which use RSA, […]
Encryption/Decryption Ciphers Used for data at rest or data in motion for AI and HPC Servers
In this post I want to discuss specifically AES flavors of block ciphers and stream ciphers. These are widely used in block cipher mode of secure communication in IPSEC and TLS protocols, it is also used for file system and disk encryption besides that most importantly used with authenticated encryption in cloud services for accesses […]
Compression/ Decompression Primitives for accelerating AI and HPC workloads
This post discusses Compression and Decompression two important primitives which needs attention both for data at rest and data in motion especially from AI and HPC Servers which are heavily dependent on memory and storage capacity besides the speed of access to data sets used for training and inference use cases. It is common knowledge […]
AI Accelerators Vector/Matrix Processing Capability especially for LLM’s and SLM’s
Before diving into vector processing capabilities of accelerators, just want to discuss most important use case of it and best example of this would be transformer model used in LLM’s. Although most people by now know this backwards and forwards, but I still need to touch upon it. Most important vector extension is to facilitate […]
Recent AI SOC (System on Chip) Custom Solutions, Features and Architecture for tasks like Ranking/Recommendation Engines, LLM (Text and Multi-Modal) Training and Inference
In Recent past there have been lot of activity of custom chip solutions called accelerators created for Ranking/Recommendation Engines and Large Language Models (text and multi-modal) training and inference to be used in Hyper scalar data centers besides existing and established Nvidia and AMD GPU based solutions. Here I intend to discuss Meta Training Inference […]