March 26th, 2009
Nehalem servers could spark a $27bn feeding frenzy
I spent most of Wednesday listening to Dell executives and their customers talk about Dell’s new IT systems, especially its new Nehalem based servers.
In the press room before the event, I was trying to read Dell’s press release, and joking with colleagues that it was full of all the same phrases that I had read ten, even twenty years ago. I’m sure you know what I mean. Here is a sample:
Dell today debuted a breakthrough portfolio of products and services specifically designed to help businesses become more efficient throughout the IT life cycle and maximize their most valuable resources: people, time and money. The combination of innovative software and services with next-generation storage and server products, offers customers technology solutions specifically designed to cut costs and increase efficiency.
And it continues in the same way for pages and pages. It could be describing any IT gear produced at anytime over the past 20 plus years. There’s not much of a story there. But, actually there is, if you look beyond the press release.
Here is the story: Dell is introducing IT systems based on Intel’s Nehalem server microprocessor, and it’s a killer system. The microprocessor is a quad core design built on Intel’s second generation 45 nm technology, which means the transistors are super tiny and so are the connections. That means electrons travel shorter distances, that makes it super fast, and it means it needs less power.
Then when you add the advanced architecture, the quad-cores, the built in support for virtualization technologies, integrated memory controller, and all the other features and tweaks from Dell (and fellow server builders), you have a super fast IT package.
And it’s supercomputer fast — at least of the 1992 variety. A Nehalem server blade will give you the power of the fastest supercomputer in the world in 1992: 93 gigaflops, costing $130 million.
On Monday, Dell and IBM, Sun, HP, Cisco, etc, will release pricing on their Nehalem blade servers, and you’ll probably be able to buy 50,000, or so, for about same amount of money as one 1992 supercomputer.
And there is more, much more…
Andy Grove, former Intel CEO and Chairman, instilled within Intel the notion of “eat your children” and that’s exactly what Nehalem does. Intel is super-confident that its Nehalem servers will not only out-perform by a huge margin all of its competition, but also its own family of single core Xeon microprocessors.
Intel and Dell executives said that a single Nehalem server can replace nine single core servers, and up to 18 servers if you turn on its virtualization technologies. While at the same time reducing power consumption by 20 per cent.
Nehalem customers will be able to recoup their costs within 8 months.
That means that for the next four years and four months of a typical five year depreciation cycle, Nehalem servers are pure cash machines.
OK, it’s not just Intel’s Nehalem technologies, it is also a bunch of additional technologies from companies such as VMware, Symantec, Dell, etc, that help boost the overall performance and utility of Nehalem servers. The point that is missing in the bland press releases about Dell’s (and you’ll see it in HP, IBM, Sun too…) Nehalem servers, is that the magnitude of the performance gain will finally allow Dell, and others, to take on the RISC server market head-on.
Deja vu Itanium . . .
The X86 based server market represents about $28 billion, RISC servers represent about $27 billion, according to Intel’s Kirk Skaugen, VP and GM of Intel’s Server Platform Group. Nehalem performance is potentially so great, and the lifetime ownership costs are so low, that the Intel architecture can finally, and decidedly, take on the RISC market with a vastly superior product.
If Intel, Dell, and all the other Nehalem server vendors are right in their claims then we will see a massive rebuilding of data centers everywhere. And it not only RISC, but also many single core X86 servers that will be ripe for replacement. The market opportunity is in the tens of billions of dollars.
The cost savings from an 18 to 1 server replacement opportunity will be more than any data center manager can stand. Or at least more than their boss can stand. It will result in a tsunami of IT systems replacement that we’ve never ever before witnessed.
For Intel it will be have been a long journey. Slaying the RISC microprocessor server market was what Itanium was designed to do when it was launched nearly 8 years ago.
Tom Foremski reports on the business and culture of Silicon Valley at the intersection of technology and media. He also writes at Silicon Valley Watcher. See his full profile and disclosure of his industry affiliations.
Subscribe to Tom Foremski: IMHO via Email alerts or RSS.
Follow on Twitter:
