Intel Reorg gives Gelsinger control of the data center

0



If something doesn’t work, change it. Sometimes you glue things together to create some kind of synergy and then pull them apart to get a necessary focus. The pendulum swings back and forth and back and forth, creating and destroying value (think of building conglomerates and tearing them apart) or careers (not everyone survives change by definition).

In other words, with the reorganization to be expected, with which we still call Intel’s Data Center Group, when it has been called the Data Platforms Group under General Manager Navin Shenoy for several years. During the reorganization, the relatively new Chief Executive Officer and previously not only Intel’s first Chief Technology Officer (2001) and also first General Manager (2005) of its Digital Enterprise Group was the first implementation of the company’s data center group.

Gelsinger, who learned from Intel’s own founders, is not lazy – he was one of the designers of the 80386 processor and the architect of the 80486 processor that brought Intel to the data center back in 1985 – and probably still is Best person to direct whatever Intel Data Center Group wants to call. Especially after running storage giant EMC and server virtualization giant VMware since leaving Intel in 2009. But Gelsinger cannot lead the data center group because he is the CEO of Intel. But he can do the next best thing, namely get rid of the data center group and create new “groups” that fulfill their functions, which are actually departments, and whose heads are directly subordinate to him. And that’s exactly what he did. And the reorganization, which will go into effect on July 6, shows how Intel plans to organize itself to tackle the many competitive threats it has in the data center these days.

First of all, when you have the lion’s share of the data center computing power, everyone is firing on you. And so Gelsinger wants to break the threat to his hegemony in the computer sector into justifiable areas. At the same time, of course, some pieces of Intel can be glued together to form a larger group of groups that, in theory at least, can create some synergy as well as a more cohesive story for Intel to tell partners and customers.

To do this, the Data Platforms Group – again what most people still refer to as the Data Center Group – is split into two parts. Sandra Rivera is General Manager of the new Datacenter and AI Group, which will include the Xeon processor lines as well as the Arria, Stratix and Agilex FPGAs. Rivera, who formerly led Intel’s Network Platforms Group (the part of Intel that was primarily concerned with expanding into the telecommunications and service provider arena), will also lead Intel’s AI strategy, which presumably means it Accelerators and the neuromorphic Loihi processors also take control of the AI ​​from Habana Gaudi and Goya. (Intel did not specifically address that part.) In her most immediate previous role, Rivera was Intel’s Chief People Officer, but make no mistake. Gelsinger clearly wants people with technical skills to run the Intel groups, and Rivera has an electrical engineering degree from Penn State University. (Go lions!)

With AMD on the rise, and Arm server chips from Ampere Computing, as well as cloud builders and hyperscalers for home use, it’s never been more difficult to run the core data center processing business at Intel. And there has never been a better environment in which to form new leaders for future leadership. Gelsinger knows this well because that’s what Andy Grove saw in him in the 1980s and 1990s, when Gelsinger rose the ranks of Intel and taught him how to run Intel. Rivera worked in sales at Dialogic, a telecommunications equipment company based in AT & T’s old turf in New Jersey, and then served as general manager of the computer telephony division at Catalyst Telecom, a telecommunications value added business. Interestingly, Intel bought Dialogic for $ 780 million at the height of the dotcom boom, but that was a few years after Rivera joined Catalyst. The point is, Rivera has a different experience than many other Intel executives and has been with Intel for more than two decades.

The other half of the former Data Platforms Group is now the Network and Edge Group, and Gelsinger managed to get away with Nick McKeown, one of the founders of Barefoot Networks (which Intel acquired for its programmable Ethernet switching business two years ago) to convince a full-time job at the chip manufacturer and becomes General Manager for the Network and Edge Group. (Which Intel mysteriously abbreviates NEX, which we will never do here The next platform.) McKeown has been a professor of electrical engineering and computer science at Stanford University since 1995 and has been an innovator in software-defined networking, most recently as co-founder of Nicira (which was acquired by VMware for $ 1.26 billion in 2012). After spending a little time thinking, McKeown founded Barefoot Networks in 2013 to use the P4 programming language to create programmable switch ASICs.

Intel is expanding its X and building it upe The GPU business is set to rival both Nvidia and AMD, and to that end, a new Accelerated Computing Systems and Graphics Group (abbreviated to AXG by Intel for some reason) under Raja Koduri, the chief architect and general manager of Intel’s Cores, Founded & Visual Computing & Edge Computing Solutions since 2017. (That title didn’t make much sense as we suspect a lot of Koduri’s work wasn’t like that either. But Intel has been trying to put out engineering and manufacturing fires for years.) Koduri has Broke his teeth at the graphics chip manufacturer S3 in the 1990s and joined AMD in 2001 as Chief Technology Officer for the Graphics Products Group. He then went to Apple for four years, from 2009 to 2013, to lead graphics chip development and then returned to Apple. AMD is working on AMD Radeon GPUs after acquiring ATI Technologies.

Eventually – and this will fuel speculation that Intel will buy VMware – Gelsinger created a new Software and Advanced Technology Group and brought Greg Lavender, former chief technology officer at VMware, to Intel to do so. Lavender has also been named Intel’s Chief Technology Officer, performing a dual role – just as Gelsinger did at Intel many years ago. Lavender was responsible for the web infrastructure software stack at Sun Microsystems after the company he worked for, Innosoft, was acquired by Sun in March 2000. Lavender became responsible for the Solaris operating system in 2008 when Solaris opened and manufactured on X86 iron and remained until Oracle acquired Sun in 2010. Lavender has been a professor of computer science at the University of Texas since 1994. Incidentally, as part of this CTO job, Lavender will lead Intel Labs as well as drive Intel’s entire software agenda – and it will be interesting to see what that could be.

These are the four pillars of Intel’s data center strategy, and people are all reporting directly to Gelsinger and the strategy they are going to be implementing will no doubt come from above. Three out of four are die-hard, seasoned tech geeks who have been into this IT game for a long time; you get the chance to become an experienced technology freak. Which is of course refreshing to look at. And frankly, the core data center business is the easiest of the challenges Intel faces in the data center. It has been much less successful with discrete graphics, GPU accelerated computing, switching, and specialized AI and neuromorphic chips. We can discuss how well Intel has done with network interfaces and SmartNICs, and what its DPU strategy looks like going forward. Mixed results have also been achieved in the traditional high-performance computing segment.

Presumably Intel’s financial data is being compiled to reflect these new groups, and we can get a glimpse of how it is going to go ahead and maybe four quarters back.



Leave A Reply

Your email address will not be published.