HPC Customer Engages Sondrel for High End Chip Design
November 25, 2024 | SondrelEstimated reading time: 1 minute
Sondrel, a leading provider of ultra-complex custom chips, has announced that it has started front end, RTL design and verification work on a high-performance computing (HPC) chip project for a major new customer.
Ollie Jones, Sondrel’s CEO, said, “HPC designs are a key area for Sondrel because they require large, ultra-complex custom chips on advanced nodes, which are our speciality. We have developed the skills, tools and advanced design methodologies to be able to create billion-transistor designs at leading nodes. For example, we have one customer for such designs that has contracted us for the past eight years to work on each generation of its advanced node chips.”
He explained that HPC designs require multicore processors running at maximum clock frequencies and utilising the very latest, coherent Network on Chip (NoC) technology, advanced memory and high bandwidth IO interfaces, so that the chip can deliver the highest possible performance. The coherent NoC enables data to move between processors, memory and IO whilst enabling processors to reliably share and maintain data that they have available in their caches. As a result, the need to access off-chip memory may be significantly reduced, which can introduce latencies that hinder performance.
“HPC is in huge demand for next generation applications that demand tremendous computing power such as AI, scientific modelling, data centres and internet infrastructures,” concluded Jones. “In every case, the key is understanding how to move data around in the optimal way to maximise performance and that is an area where we have world-class expertise and in-house technologies such as our Advanced Modelling Process. This is part of our suite of in-house tools and flows, which have been perfected over many years, and enables us to analyse exactly what is going on inside a chip design to ensure that the data flow is balanced properly and that the processors are not stalled waiting for data. It runs through every stage of the chip design to ‘prove’ that the design meets its specification and performance requirements, which is why we are the partner of choice for these projects as customers can see at every phase that it is on specification and on track.”
Suggested Items
Compal, ZutaCore Collaborate to Showcase Groundbreaking Waterless Two-Phase Liquid Cooling Server Solutions at SC24
11/25/2024 | Compal Electronics Inc.Compal Electronics, a global leader in server innovation, has partnered with ZutaCore®, a leading provider of waterless direct-to-chip two-phase liquid cooling (2P DLC) solutions, to introduce a series of groundbreaking server solutions.
Keysight, Instrumentix Partner to Launch Complete Trade Monitoring Solution for Financial Markets
11/21/2024 | Keysight TechnologiesKeysight Technologies, Inc. expanded its financial capital markets portfolio through a partnership with Instrumentix to introduce a cutting-edge trade solution.
Gartner Forecasts MENA IT Spending to Grow 7.4% in 2025
11/20/2024 | Gartner, Inc.IT spending in the Middle East and North Africa (MENA) region is projected to total $230.7 billion in 2025, an increase of 7.4% from 2024, according to the latest forecast by Gartner, Inc.
ASMPT: Highly Flexible Die and Flip-chip Bonder for Co-packaged Optics Production
11/20/2024 | ASMPTThe high-precision AMICRA NANO die and flip-chip bonder has been specially developed for the production of co-packaged optics where which optical and electronic components are integrated in a common housing. With its exceptional process stability and a placement accuracy of ±0.2 μm @ 3 σ, this innovative bonding system is ideally equipped for the communication technology of the future.
New Ultrafast Memory Boosts Intel Data Center Chips
11/19/2024 | IntelWhile Intel’s primary product focus is on the processors, or brains, that make computers work, system memory (that’s DRAM) is a critical component for performance. This is especially true in servers, where the multiplication of processing cores has outpaced the rise in memory bandwidth (in other words, the memory bandwidth available per core has fallen).