Missed The Big CPU News This Month?

Sponsored Post: In case you missed the big news earlier this month, Intel introduced its 4th Gen Intel Xeon Scalable processors (formerly codenamed Sapphire Rapids) to a huge industry fanfare – groundbreaking datacenter silicon which promises to push the boundaries of performance for high performance computing (HPC), artificial intelligence (AI) and networking workloads.

You can still watch a livestream of the official launch event to hear Sandra Rivera, executive vice president and general manager of the Data Center and AI Group, and Lisa Spelman, corporate vice president and general manager of Intel Xeon Products explain just what’s so special about the latest Xeon CPUs.

The numbers speak for themselves – the most accelerators ever built into a datacenter processor; up to a 53 percent average performance gain for general purpose compute; total cost of ownership (TCO) improvements which range from 52 percent (databases) to 55 percent (AI real-time inferencing) and 66 percent (HPC) – the list goes on.

How do the 4th Gen Intel Xeon Scalable processors deliver all of this? A big dollop of the magic sauce is in the workload accelerators which can offload various algorithms and routines from the core CPU to run applications that much faster without having to rely on the core Xeon chip to do all the hard work. Have a look at the official Intel® Accelerator Engines web page for a more detailed explanation of exactly how the accelerators work and what they do.

It’s not just about superfast performance either. The 4th Gen Intel Xeon Scalable processors are the most sustainable chips the company has ever delivered, purposely designed with power/performance management features which can make optimal use of CPU resources to help meet customer sustainability targets. So much so that they offer a 2.9X average performance per watt efficiency improvement for real world workloads compared to earlier 3rd Generation Intel Xeon CPUs.

You can access more detailed performance benchmarks in the full product brief, which is available here. Intel estimates there are already more than 100 million Intel Xeon CPUs of one type or another currently in use, powering anything and everything from on-prem servers to “as a service” infrastructure platforms, networking equipment, wireless base stations and cloud platforms. With the 4th Gen Intel Xeon Scalable processors optimized to run applications like AI/ML, analytics, networking, security, storage and HPC, how long before there’s another 100m?

Sponsored by Intel.

Intel Legal Notices and Disclaimers

Sign up to our Newsletter

Featuring highlights, analysis, and stories from the week directly from us to your inbox with nothing in between.
Subscribe now


  1. On channel holding data and the channel foretells all, there are more like 1.9 billion Xeon components in use back to v2, old computers tied to their applications don’t go away easily and if it works don’t fix it. Ivy, Haswell and Broadwell are also fairly power efficient on application core utilization in relation the new massive multicores designed for hyperscale and cloud services.

    Intel needs to regain its ability to supply 120 M units plus of Xeon annually as it did between 2016 and 2021 exceeding that volume objective by at least x3, and where the majority of this overage walked out the back door unreported and laundered in sales package appeasing compliment syndicates and as tariff payments for country access. It’s the obvious reason looking back.

    Plowed by Intel into the dirt for all to see and guess who’s been digging it up for replacement through their stamp mill operations.

    Thank you Intel for funding the entire world.

    It’s time for Intel to stop funding the technical ascendancy of intra nation cartel and inter nation syndicate operations by goosing them with non-reported Xeon volume, and Core volume, that are other than revenue units. The tax man is waiting. Intel social welfare is theft.

    IDM 2.0 cannot thrive, and Intel cannot survive as Intel once did or always has on bribes of all kinds. It’s time for Intel to become the best practice, industrial and socially responsible, and legitimate governance institution Intel always proclaimed but never in fact truly was.

    Be damned industries inability to transmit as other than mimicking repeaters those misrepresented broadcast portrayals of nirvana in Intel land. It’s time to see outside and beyond the Intel invented realities because to do other leaves decision making stuck in the same old box.

    Mike Bruzzone, Camp Marketing

  2. Tim,

    I’m referring to slide #2 at the URL directly below which shows ebay offer data relied as a proxy for WW channel x86 and GPU volume also relied by AMD, Intel, Nvidia as a proxy for production beyond what might be estimated from the 10-Q and 10-K. Otherwise AMD, Intel, Nvidia rely on electronic trading site(s) data for inventory sales management administration treating every site like its own market segment.


    For estimating channel share and as a proxy for production, ebay replaces the Intel 2008 through 2014 disintegrated story problem cipher, replaces Barrett era 2005 through 2009 by quarter by percent product category tables that were then divided into total IDC / Gartner volume envelope, replaces the 2001 into 2004 underground by product supply statement, replaces the original MDR Intel annual relied for by product category production volume to estimate Intel revenue and margin quarters into future time, replaces Price Watch, replaces Store Board. Over time [most] everything returns to the channel for secondary resale?

    Channel inventory is a good proxy for actual AMD, Intel and Nvidia production volume?

    Here’s another brief offering the same slide updated at No. 1 in the set, with additional x86 and GPU share data all derived from ebay offer data that is highly screened to avoid duplication of components grade SKU weight.


    As some are aware who follow my work, I don’t believe AMD, Intel, Nvidia production, can accurately by accounted from financials dividing gross or net into total revenue and where that gross amount begins on $1K grade SKU average weighed price determined from channel in this example ebay x86 inventory weight on AMD and Intel $1K price, and for Nvidia on Total Revenue Total Cost assessment backing cost out until reaching TSMC cost : price to the customer.

    “85 M to 100 M total Scalable”? Cannel supply suggests much more.

    Intel has always been capable of producing more product then recognized and that’s good because Intel needs to make every unit produced a revenue unit. Now all anyone can hope is every unit of production is a demanded value.

    Mike Bruzzone, Camp Marketing

  3. I wonder how ARM will cope in complex calculations enviroments where x86 is king in drugs design, seismic analysis.
    I read that ARM needs at least 2X more transistors to run floating point workloads where xeon shines.

Leave a Reply

Your email address will not be published.


This site uses Akismet to reduce spam. Learn how your comment data is processed.