Quick Take: New Car SyndromeApril 9, 2009
John De Gelas over at Anandtech poses the “million dollar question” of upgrading memory and CPU versus replacing the entire system. So far, there is a 2:1 margin in favor of “throwing out the baby with the bath water” and replacing the entire server even when new CPUs are available. This curious “new car syndrome” that seems to have affected IT decision makers in the past will have to change in the “new” economy. Is Intel’s marketing strategy sound?
Some in the AMD fan-boy camp have accused Anand of “bending” to Intel by “always” presenting new Intel products in the glowing light of an post-coital cigarette. However, Intel has consistently “outed” more technology than AMD on many more fronts: hands-down, Intel gives Anand’s group much more to talk about! However, this relentless cycle of “product renovation” from Intel is an interesting formula of update and obsolescence that – like the “latest model” vehicle update – drags the user to a buying decision much earlier than the life-cycle warrants.
SOLORI is big on stable image platforms – Intel has SIPP for desktops and AMD’s cover desktop and server – and expanding an equipment line’s useful lifetime from 18 to 36 months with simple memory and processor upgrades seems an obvious choice. Perhaps the Intel-focus at AnandTech creates a reader bias towards disposable platforms than it would with more AMD followers. AMD Opteron users are likely more familiar with processor swaps than Intel users, given that a single-core AMD platform purchased in 2006 can be upgraded to 6-core Istanbul later this year with only another BIOS update – adding another 12-months to its life: that’s a 48-month product lifetime in tech!
An Eco-System standard is not about making purchasing decisions and guaranteed compatibility: it’s about making an investment in supplier and support chain but it’s also about extending the useful life of core technologies without heaping-on additional risks and expenses. The effortless progression of the AMD 2000 and 8000 series from single, to dual, to quad and now to six core systems is unmatched on the Intel side of the street. While John De Gelas quickly questions Intel’s specious ROI math, he then refers to a post from John Fruehe of AMD about this upgrade potential but then counters with his own anecdotal bias.
I use the term “anecdotal bias” because Mr. De Gelas is demonstrating a rationale formed through upgrade behavior carefully marketed by Intel: drop and swap. Look at the many CPU, socket and memory ping-pongs Intel has handed its user base is akin to the Apple iPod or iPhone market, and many users would play along. It would have been prescient of John to identify that Intel has entered a new market – one that AMD has proven benefits the consumer not the seller – and that Intel’s marketing will need to shift to accomodate a socket that has a 36-48 month lifespan?
What will a longer life cycle mean to Intel and how will that change the landscape of competition between the two chip makers? Intel’s initial salvo into the new 2P market (W5500 series) is a dithered array of performance variations across a $200-1,600 per socket palette. Oddly, Intel recognizes this shift in their 5500 product slick by pointing to “ROI refresh” as replacing existing servers. The “benchmark” number is 8-months ROI, based on ONE new N-EP-based server (X5570, $1,600/socket) replacing NINE 4-year-old, single-core Xeon systems.
SOLORI’s Quick Take: The subtext seems to be: you made a bad decision 4-years ago in purchasing a dieing technology – erase that mistake with N-EP. The fact is that the capture window for upgrades is anything note capable of taking a Nelalem-EP processor (limited to 144GB of RAM today). The REAL problem is that Intel’s server platforms and upgrade cycle is far shorter than 4-years making their ROI a nice work of marketing fiction. In fact, a 4-year old Xeon system would have been replaced every 18-months by the old rule. It will be interesting to see how the Intel marketing machine handles a platform that survives longer than 18-months.
SOLORI’s 2nd take: The data center is not a place for “gadget geeks” to “test drive” the latest tech: solid and reliable are better accolades for a data center than “constantly being updated.” With most companies’ IT budgets on morpheme drips, an Intel-prompted regen of the data center based on “the latest tech for tech’s sake” seems pricey compared to CPU replacements and memory updates available to the AMD camp. Time will tell, but it will be interesting to compare the budgets of AMD-based tech houses with Intel-based tech houses at the end 2009 to see who the real winners are…
Participate in John’s “data center upgrade” poll on his blog post at AnandTech.