Modular Ghiplet Chiplets Redefine Flexible GPU Design
chetanarvindpatil 15h #Technology #Thread #Semiconductor #Manufacturing #Ghiplet The Semiconductor Ghiplet: 1/ - The Semiconductor Computing World First Moved Towards More-Than-Moore Via Chiplets. - Now, It Is Re-Inventing Another Path Towards More-Than-Moore. - This Time With Ghiplet = G(PU) + (C)hiplet --- chetanarvindpatil 15h Author 2/ - Ghiplet Represents A Shift Towards Modular GPU Architectures. - Where GPU Is Assembled From Multiple, Specialized GPU Blocks Rather Than Designed As A Single, Monolithic Die. - This Modular Approach Enables More Flexible And Scalable GPU Designs, Catering To Various Performance Requirements And Application Scenarios. - Mainly, The AI's Computer Needs. --- chetanarvindpatil 15h Author 3/ - By Leveraging Ghiplet Architecture, Companies Can Tailor GPUs More Precisely For Different Markets. - From High-End AI Computing Solutions To Low-Power, Energy-Efficient Mobile/Edge Computing Solutions. --- chetanarvindpatil 15h Author 4/ - Ghiplet Architectures Also Demand Developing And Deploying Advanced Interconnect Technologies To Ensure Efficient, High-Speed Communication Between Multiple Ghiplets. - Innovations In This Area, Including Silicon Photonics And Advanced Packaging Solutions Are Critical For Maintaining Low Latency And High Bandwidth, Essential For High-Performance GPU Applications. --- chetanarvindpatil 15h Author 5/ - Ghiplet Is Not A Formal Term Right Now. - But, Let This Be On Record That You Read About "Ghiplet" First Time Here. --- chetanarvindpatil 15h Author #chetanpatil - Chetan Arvind Patil - ChetanPatil.in chetanpatil.in #chetanpatil - Chetan Arvind Patil - Semiconductor and Beyond
Ledger Should Upgrade to Newer Secure Enclave
I would be really nice if Ledger would make a device with a newer Secure Enclave
Micron Poised for Record Highs Amid Memory Shortage
Can Micron do it again? $MU heading back towards record territory as memory shortages continue.
Google Grants TPU Compute Awards for Keras 3 Researchers
If you're a researcher in academia using Keras 3 (PhD student, postdoc, professor...) and you want to train on TPUs, you could receive compute awards from Google for your research. Google is running a new academic grant program, separate from...
China’s Node Surge Forces GFS to Target Allies
With China ramping up mature node at insane rates, where does that leave $GFS? Sell only to allies? UAE has some big aspirations here. Yes, I know, Mubadala, I was at AMD when we spun GF.
ThinkPad X1 Beats Gamer Laptops for Game Development
I have been using a Thinkpad X1 for almost a year, and it hasn't fallen apart yet, unlike the "gamer-class" laptops of the past few years that fall apart hardware-wise, and never even get to a working state software-wise. Though this...
COT Shift Signals Changing Custom ASIC Customer Landscape
Important commentary here on COT on the $CDNS call last night. Called out this trend last year. It will be important for $AVGO $MRVL $MTK customer dynamics in custom ASICs. https://t.co/lgWuwvj4oB
Silicon Oxide Memory Moves From Lab to Industry
Silicon Oxide Memory Breakthrough | Lecture 12: From Lab to Industry #science #graphene #pchardware https://t.co/78xHMIA1EU
Pixel 10A Debuts $499 Flat‑Back Design
Google has announced the Pixel 10A. Same $499 price. (Mostly) the same specs. Thinner-but-still-thick bezels. But now the back is *totally* flat, you see. How flat? See pics. And then read my story. https://www.bloomberg.com/news/articles/2026-02-18/google-announces-499-pixel-10a-just-before-expected-iphone-17e-reveal?srnd=homepage-americas
MI455X Set to Generate Revenue in H2 2026
I have very high confidence that MI455X will ship for revenue in 2H26. Token generation naturally will be after that. There’s always a ramp and the only question is how steep the ramp.
HyperAccel Unveils Bertha 500, Low‑latency AI Chip
South Korean AI chip startup HyperAccel @HyperAccel0112 is preparing to launch Bertha 500, an LLM inference accelerator aimed at cost-efficient, low-latency token generation. The company also plans an edge SoC version in partnership with Korean giant LG: https://t.co/TRgQrMNik2
Energy‑Per‑Prompt Redefines AI Query Power Cost
chetanarvindpatil #Technology #Thread #Semiconductor #Manufacturing #Energy The Semiconductor Energy Per Prompt: 1/ - Energy Per Prompt Defines How Much Electrical Energy Is Consumed To Deliver One Meaningful AI Response. - It Shifts Focus From Peak Power Or Total Datacenter Energy To The Cost Of...
AI Wearables Turn Pets Into Smart Companions
AI wearables for pets. What a world we live in! (And it's cool, I had a demo recently).
Ex-Altera CEO Leads Vsora to Challenge Nvidia, AMD
Former Altera CEO Sandra Rivera has joined French AI chip startup @vsora_dsp as chair of the board. I had a chat with her about raising Vsora's profile, raising funding, and competing with Nvidia, AMD and the entire landscape of startups...
Massive Compute Unlocks New Data Manipulation for AI Centers
Massive compute capabilities enable a whole new way of manipulating and using data, and a potential bonanza for AI data centers. https://t.co/c1vSecTfww #AIdatacenters #abstractions #virtualtwins #digitaltwins #AgenticAI #semiEDA
Vivo X Fold 5 Ranks Second Best Foldable Camera
Randomly jumped back on the Vivo X Fold 5 and this is still the second best foldable camera around behind only the X Fold 3 Pro.
Quantum Breakthrough Makes Once‑impossible Problems Solvable
I’ve been waiting for a moment like this. Google’s new quantum chip solved a 150-year-old problem in 2 hours. 13,000× faster than one of the world’s best supercomputers. That’s not incremental. That’s a different category. Their Willow chip ran an algorithm that models how atoms...
Memory Limits Will Bind Semiconductors Through 2030
Hard to really say past 2027 for certainty at this point, but its very easy to lay out the scenarios that keep memory constrained to 2030. Really not just memory, the depth of the constraints are so deep across...
AI Boom Sparks DRAM Crunch Across All Companies
NEW ODD LOTS - THE DRAM EPISODE The AI boom has created voracious demand for memory, meaning even consumer-facing companies are now having to compete for the chips @TheStalwart & I speak with @rwang07, analyst at SemiAnalysis, about the big DRAM crunch https://t.co/HYFcg7nK8Z
Shared Knowledge Multiplies Value in Industrial Collaboration
What truly defines great collaboration? 👉 Sharing knowledge. That’s exactly what we live in our cooperation between Indu-Sol GmbH and GROLLMUS . 🔌 Indu-Sol contributes deep expertise in industrial communication and fieldbus systems like PROFINET and PROFIBUS. 🧠 At Grollmus,...
US Controls Slow China, Ignite Homegrown Chip Drive
NYT: "While Washington’s export controls have slowed China’s chip development, they have added fuel to Beijing’s decade-long push to make strategic technologies like semiconductors and A.I. entirely at home." https://t.co/OcHRXob06N
Introducing Our 7th‑Gen Robotic Hand, Human‑Level Precision
I've been waiting 3 years to show you this We just launched our 3rd-gen humanoid, but we’re already on our 7th-gen hand Our team has quietly worked for years to approach parity with a human hand Excited to share a sneak peek of...
OneWeb Boost
.@EutelsatGroup: OneWeb rev up 59.7% in 6M to Dec. 31; 600 aircraft have our connectivity hardware; hosted payloads, ground segment seen as fresh revenue sources; cancelling @Thales_Alenia_S GEO sat saves EUR 100M in capex, costs 60M in impairment charges.https://t.co/f8cngyWfes https://t.co/iWpscm5aIi
Start Building AI Now—No $10k Machine Needed
Alex bought a $10,000 Mac to run @openclaw with powerful local models. He isn’t alone. Know a guy with several NVIDIA H200’s in his house. Now that I am building I get it. Having your own local AI changes your whole approach...
Google Backs Apple Vision Pro, Hinting at AI Partnership
It's great to see the Apple Vision Pro being invested in by Google. This might have something to do with the Google and Apple AI deal, and if it does, it's a great sign. I hope to see Google Genie...
Bluetooth Pacemakers Could Be Tracked via War‑Driving
For the Nancy Guthrie case, an idea and maybe a crazy one but she had a pacemaker which often implantable devices use bluetooth such as Medtronic's. Couldn't you war-drive (drones even better) with a high gain antenna with amplifiers -...
Nvidia Chips Generate $8‑10 Economic Ripple per Dollar
"When I look out over the next two, three years, a ripple effect for every dollar spent on $NVDA chips, there's an $8-10 multiplier across software and infrastructure" 🎙️ @DivesTech joins Dan on the latest Okay, Computer. https://t.co/kyBad40Wln