HP Itanium growth is a misleading number
HP Itanium servers running HP-UX are direct replacements for HP PA-RISC servers running HP-UX. To make specific HP-UX processor generation comparisons to the entire instruction set architecture of a competitor is bogused. It is an uptake rate, not a growth rate. It would be like Sun comparing the quarter over quarter UltraSPARC T1 growth rate to the total HP-UX growth rate.
Second, Matt talks about "Integrity revenue" growing faster than UltraSPARC T1 revenue. In absolute dollars, this is probably true, given Integrity is a high-end server processor and UltraSPARC T1 is a low-end server processor. I don't need Matt to tell me HP's Integrity servers are overpriced. And again, this is replacement revenue, so it is not really growth unless HP is selling more total HP-UX systems.
Matt, I want to see the combined Integrity/HP-9000 numbers. They are the same servers, even the same chipsets, just different processors. With HP-UX, they are the same OS.
Customers need to realize the operating systems which matter going forward are OSs which run on sub-$100K systems. Why? Because a four-socket Tigerton system in November of 2007 is now as powerful as a 16-socket IBM POWER4 Regatta in 2001. And with VMware, it is just as virtualizable. Larger systems are becoming niche players, the new mainframes: capable, and necessary, but low volume and very expensive.
If HP wants HP-UX and Itanium to survive, it should significantly broaden its Integrity blade server offerings. If Itanium is supposed to be the scalable CPU, why does HP only offer two-socket Integrity blades? HP has to make a four-socket, 64 GB memory Integrity blade in the same form factor as their four-socket Xeon and Opteron blades. And then HP should create a double-wide eight-socket Integrity blade. If customers could stuff four 8-socket HP-UX servers into a single C-Class chassis, HP might have something to offer.
If HP does not want to do this, then it is relegating Itanium and HP-UX to a replacement market, destined to shrink in units at the inverse rate of Moore's law (what Sun's Papadopoulos calls "Blue Shift"). In this case, HP has to figure out its post-Itanium enterprise play. Will it be scalable x86 using Xeon, Opteron, or perhaps both? What will the OS be? Windows? VMware ESX? Linux? Solaris? There are challenges across the board here.
Let's be real. The former will not happen. The latter is a forgone conclusion. Tukwila will come, with a common socket it will be able to share with Xeon. Poulson will never see the light of day.
Several years ago, some people speculated on Intel's "Plan B", project "Yamhill". That speculation, repeatedly denied by Intel, was true. I bet HP has an Itanium "Plan B", with chipset engineers making sure Xeons with CSI will work on the Superdome follow-on, and with VMS and NSK engineers working feverishly on x86 ports. But are they porting HP-UX to x86, or is HP-UX a dead OS walking?