BAD HARDWARE WEEK

A gem or something else? Lets check.
In two part bullshit 1 and 2 , at recently bought The Inquirer, journalist Nebojsa (Fearless!!) gives some perspectives on Woodcrest and Intel's future server chips. In first part distinguished jounalist finds that Woodrest performance are unbeatable, what is true, but devil is as always, in the forgotten details. He didn't mention that his analysis doesn't apply under all circumstances, but only for carefully selected applications that have to fit into Woodcrest's 4MB L2.
Frankly, I don't know any standard 2-4GB DRAM Vista server application that will completely fit under real conditions into 4MB L2, but perhaps someone knows?? Beside, do you know that you will need a lot of memory bandwith for Vista's 3D Aero graphics (picture below)? Some 1800 Mbytes per second at minimum.

Yes, Nebojsa separated second part of his analysis, where he showed that Intel's new generation server micro architecture is a bullshit without quite revisioned currently associated I/O architecture. But, who will garantee that all will read both of his articles?? And of course, Woodcrest's the most important part ( that one speaking on Woodcrest performance sensitivity on application locality bigger than 4MB of L2) he left for the UNPUBLISHED 3rd part of Woodcrest analysis , that one which naturally fits somewhere BETWEEN those two mentioned parts. Luckily, you can find it almost mirrored here.
In the end, let compare Intel's "New generation architecture" from its both end. Low performance Merom and high performance Woodrest use the same micro architecture, the same size 4MB L2 cache, but differs dramatically in performance, though not so much in clock (3Ghz at Woodrest and 2,67Ghz at Merom). WHAT WILL HAPPEN WITH A NEW GENERATION ARCHITECTURE PERFORMANCE AND ITS ONLY 14 STAGE PIPELINE WHEN 4/2MB L2 CACHE MISS HAPPEN? THE PIPELINE FLUSH! Do you feel something again is stinking at Intel (and its rabat partners)? Very bad. Please, DO A FLUSH, immediately.

This post link