Hi. Patrick O’Rourke here – I am also a Product Manager over in the Windows Server Division. I’m writing because Cactus league isn’t in full bloom yet, and I’m still waiting for Gary and Bob to be fired so I can be interested in the NHL again.
I was watching news from Intel’s Dev Forum, and the amazing thing was – no sooner did Jim Allchin step off stage at IDF then his comments about Windows x64 timing made headlines. Now there’s an example of Intel’s Centrino technology shortening the race to publish first 😉
Too bad one outlet confused reality by reporting that the desktop and server versions of x64 are on different release trains. Based on the CPP and RCs alone, it’s obvious that these trains are leaving the station together.
So we should see x64 server and client in time for WinHEC. In fact, the WinHEC team says 64-bit will be everywhere at the show. So what’s that mean to developers and ISV partners? According to Jim’s comments today, that means get your code ready — make sure 32-bit apps are tested and supported for AMD’s Opteron64 and Intel’s Xeon EM64T.
For those interested in Windows Server x64 performance numbers, Jim highlighted internal tests that showed ERP apps get 20% more users when running x64 Editions vs. 32-bit Window Server 2003. And the x64 Editions gets nearly 2X more users per server [insider note: with 4-way single cores] for Terminal Services.
Jim also spoke of Windows on Itanium. He pointed to a recent NEC benchmark for 32-way systems running SAP’s sales and distribution standard app benchmark. Just further proof that IPF is best for scale-up database and business applications.
Aside from 64-bit computing, Intel was pushing I/O routing technology called I/OAT. According to Intel, I/O Acceleration Technology “speeds up data-flow between server applications and the network.” To this, Jim said that we’ll support I/OAT in future versions of the Windows client and server.
Intel also pushed dual-core/multi-core chips under the banner of Moore’s Law lives [the last part is a separate issue]. Multicore chips are expected to deliver greater performance than single core chips, but how much greater or slightly better varies. This is great stuff for Windows, SQL, BizTalk and other customers since Microsoft will license per proc, not per core. It’ll be interesting to see if other large software vendors change their app licensing policy to price dual-core chips as a single proc. If you have insights, let me know.