Steve Carlson

Jim Keller, former Intel, AMD, and Tesla engineer, aims to challenge Nvidia's dominance in AI with Tenstorrent's energy-efficient chip design. Tenstorrent's upcoming second-generation AI chip promises superior energy and cost efficiency, presenting a viable alternative to Nvidia's offerings. The innovative design of Tenstorrent's chip features independent decision-making cores, potentially reshaping AI chip technology across diverse applications. #AI #Nvidia #semiconductor #semiconductors #chip #technology #innovation #research #computing Tenstorrent NVIDIA https://lnkd.in/g6a3vAu4

31

At Broadcom, we enable today's Gen AI and deliver breakthrough technologies that will power tomorrow’s AI. A cluster of one million XPUs requires over 10 million high bandwidth links. Our SerDes roadmap provides the right foundation for our high performance XPUs and switches to drive power efficiency and scale for these large clusters. Here is our sixth video of the “Enabling AI” video series. #enablingAI #connectedbybroadcom

679

6 Comments

"Scotty we need more power!" For those of use old enough to remember Star Trek TOS (and if you don't know what that is I'll wait for you to visit the wikipedias...) this phrase is burned indelibly into our brains. Invariably some nefarious force was holding the Enterprise against its will, or the ship was trying to outrun a baddie, or an incoming energy bolt threatened the integrity of the hull, and always the solution was "more power". The world we live in today is quite different. Not to say that the need for compute power has reduced, in fact quite the opposite given the enormous computational requirements of modern AI and HPC workloads. But that "power" must now be weighed against the power (energy) required to use it. In AI it's no longer about TOPS, it's about TOPS/W. And a laptop that dramatically outperformed its competitors but had only a 30 minute battery life would be considered fairly useless. So when the power (energy) requirements of datacenters start to be measured as a percentage of global energy consumption it's easy to see why managing and reducing power (energy) is a critical part of modern designs. Enter the Movellus Aeonic Power product family. By integrating the power supply for components of an SoC the power (energy) can be made more localized, smarter, more efficient, and the overall demand for energy can be reduced significantly. So rather than running all of your compute cores or AI acceleration clusters full tilt when only a few are needed, you can now modulate power at the core or block or cluster level to reduce the energy to precisely what is needed and no more, and this fine grained power control can yield immediate benefits. Advanced power savings like this were once attainable only by the largest and most advanced teams on the planet, but the field has been leveled. If you want to know more just send me a note, or come visit us at DAC 2024. Or just keep burning more energy than you have to, reducing your competitiveness and decreasing reliability, that's an option too... #AI #PowerDelivery #DAC2024

6

Don’t miss the Chiplets: Building the Future of SoCs virtual event on July 24 & 25! We’ll explore technical trends, market demand, and new application areas around chiplets concepts, challenges, technologies, and more with @EE Times and @Embedded.com. Hear from Cadence experts on the role of EDA companies in the chiplet transformation and how you can get a jumpstart in designing your own chiplets. https://ow.ly/OsEM50SzgPT

19

Efficiency is key to enable AI at the edge. In this video our VP of AI, Carlos Morales shows how efficiently and fast Ambiq Apollo5 (Cortex M55 on SPOT) can classify images. Reach out to hear more if you are curious! #edgeAI #CortexM55 #arm #efficiencymatters #subthreshold #SPOT #edgecompute #innovation #semiconductors

13

A very interesting blog by Giuseppe Piccolboni of Weebit Nano Ltd about the use of our #ReRAM for #neuromorphic applications

36

The market for embedded #NVM is expanding quickly, with various technologies competing for different applications. Want to hear more? Read this article I recently posted on embedded.com. #ReRAM #semiconductor

18

Great article with insights by Bernard Murphy at SemiWiki.com, "Arteris Frames Network-On-Chip Topologies in the Car." "... while there is plenty of complexity in evolving car architectures and the consequent impact on subsystem chip/chiplet designs, connected on-chip through both coherent and non-coherent networks, the intent behind these systems is quite clear. We just need to start thinking about total car system architecture rather than individual chip functions." #Arteris #NetworkOnChip #Automotive https://bit.ly/3UiKqls

25

Qualcomm has approached Intel Corporation for a potential acquisition, which could lead to one of the largest M&A deals in history, though it is still uncertain if an agreement will be reached due to possible antitrust scrutiny. Intel's shares rose following the report of Qualcomm's interest, despite the company facing significant challenges, including a 37% decline in stock value over the past year and ongoing restructuring efforts led by CEO Pat Gelsinger. Qualcomm is engaging with U.S. regulators in hopes that an all-American merger could alleviate regulatory concerns, while Intel is also exploring various options and has received a boost from recent announcements, including a deal with Amazon Web Services (AWS). #semiconductor #technology #Intel #Qualcomm #Amazon #chipmaker

11

1 Comment

Dive into an enlightening conversation with Cadence’s Vice President of Silicon Signoff, Sharad Mehrotra, on how the Tempus Timing Solution is revolutionizing large chip signoff. Check out the EDACafe interview here: https://ow.ly/rKEU50SrrgZ

8

Semiconductor aging is becoming an important factor for device reliability. This article from Brian Bailey provides a good explanation of the nature of the problem and what the semiconductor industry is doing about it.

16

1 Comment

Some thoughts about the current Nvidia AI networking wave. BroadCom has been the king of networking for the last 20+ years. Is Nvidia going to be able to topple them? Nvidia has their own niche "AI networks" which enables them to sell the chip/equipment with such a high margin they are able to simply out-spend their competitors? Kind of like BroadCom has done the last years in almost all other network segments? The competition is on , who will be able to build the best AI network chips?

7

3 Comments

Intel Corporation will now be spinning off it's Intel Foundry division into a wholly owned subsidiary and also consolidating all it's design operations into CCG (client computing group, or lpatop/PC division) and DCAI (datacenters and AI, the server group). Networking group NEX will remain for now, but it is too small now and I'm guessing it will be sold off. Qualcomm is already looking to acquire some of Intel's operations as per The Wall Street Journal; Qualcomm is most likely interested in the CCG piece. (https://lnkd.in/gYkgwxGq) Intel will likely break itself into three major pieces - Manufacturing/Foundry, CCG and DCAI will become separate companies. But let's talk about Intel Foundry : Intel consolidated all it's manufacturing and research into Intel Foundry which started reporting it's own P&L from Q1 of this year. But it wasn't a different legal entity which now Pat Gelsinger has announced will happen. But even then, without a much more massive US govt assistance, Intel can't pay to become competitive again. US govt is the only entity that can pay for, the R&D needed to catch to TSMC and match Samsung Semiconductor, pay for new fabs etc. However Intel's failure to get a customer (Broadcom most recently, and Qualcomm and NVIDIA long ago, dropped the 18A offering) created multiple problems - the US Commerce Department's conditons for funding were not being met. And since AMD never showed any interest, - any help to Intel Foundry, which has been a part of Intel (unlike help to TSMC and even Samsung Semiconductors - both of which have well established fab business and serve multiple customers), would be construe a help to wider Intel business which then competes with AMD, Qualcomm, Nvidia. The latest move to separate out the Foundry as a legal entity opens to path to solve multiple problems at month. In my speculation the following will happen before Q3'2025: (1) Intel Foundry will not only become a separate company (legal entity) as Pat Gelsinger has already announced, it will in fact drop the name "Intel" altogether and be names something like "American National Fab corporation" - National Semiconductor would have been a good name but it is already taken. (2) Intel (or whatever survives without Foundry) will have less than majority stake in the new entity. A consortium made of AMD, Qualcomm, NVIDIA, Texas Instruments, Analog Devices ... all the marquee names of US Semiconductor industry will also be "encouraged" by US Commerce department to aquire, between themselves, a majority stake. Even the likes of TSMC may be persuaded. (3) US Govt will subsequently open the floodgates of money to the foundry to make it a success. (4) The new owners will bring about management and technical ladder changes acutely needed. I hope this happens for the sake of the talented employees, the US and the world public.

8

🌟 𝑾𝒉𝒂𝒕 𝒊𝒔 𝒂 𝑺𝒖𝒏𝒅𝒂𝒚 𝒘𝒊𝒕𝒉𝒐𝒖𝒕 𝒂 𝒈𝒓𝒆𝒂𝒕 𝒓𝒆𝒂𝒅? Check out this insightful article on the challenges and costs of porting, the fundamental differences between CPUs and GPUs, hardware-software mismatch, and the immense demand for supercomputing power driven by AI and machine learning needs! 🌐 How NextSilicon effectively addresses these limitations and needs! Published by insideHPC 🔗 𝑹𝒆𝒂𝒅 𝑻𝒉𝒆 𝑨𝒓𝒕𝒊𝒄𝒍𝒆 𝑷𝒂𝒊𝒏 𝒐𝒇 𝑷𝒐𝒓𝒕𝒊𝒏𝒈: 𝑻𝒉𝒆 𝑳𝒊𝒎𝒊𝒕𝒂𝒕𝒊𝒐𝒏𝒔 𝒂𝒏𝒅 𝑪𝒐𝒔𝒕𝒔 𝒐𝒇 𝑮𝑷𝑼-𝑫𝒓𝒊𝒗𝒆𝒏 𝑯𝑷𝑪-𝑨𝑰 https://lnkd.in/dA4YUfXZ I’d love to hear your perspectives on navigating these challenges! #CPUs #GPUs #FPGAs #HPCHardware  #HPCSoftware  #MachineLearning #supercomputing #NextSilicon #Maverick #TechInsights #AI

11

In this bonus episode - "Free Gold - Understand the value you represent ... The Power of Impenetrable Confidence" in my Semiconductor Leadership Podcast Series. We are close to 1,500 subscribers in just under 3 months. I think we can do 2,000 by the summer. Please subscribe if it is of value to you: https://lnkd.in/diktXgVg In this exclusive bonus episode of the Semiconductor Leadership Podcast, we delve into the often underestimated power of confidence in the high-stakes world of BUSINESS. Discover how to have that impenetrable confidence to shape perceptions, influence decisions, and drive innovation in this fiercely competitive industry. We’ll explore a real-world scenario whereby someone experiments by handing $2,000 worth of gold for free to the random public and where confidence has been a game-changer where you can discuss strategies to build and project this essential quality, and analyze its impact on perceived value both within a company and in the market at large. Tune in to gain insights that could redefine how you lead and succeed in the semiconductor realm. Robert Quinn Ksenia V. Mia Chen

17

1 Comment