2 GHz is quite good for a foxton-less quad-core 65nm Itanium, PA-Risc/Alpha users will be more than happy to join IPF at that clock and with the new high bandwidth CSI interconnect.
True up to a point.Performance wise it will rock simply because the IPF core is stalled most of the time waiting for memory.Quad-channel FBDIMM should really really rock barring a major screwup like Blackford.
But , IPF doesn't exist in a virtual world.
In the real world there is Power6+ , SUN Rock ( the lesser threats ) and Nehalem EX Beckton.
Beckton has 8 cores , 24MB of cache , CSI.It will run at over 3GHz and has SMT.As x86 get more and more RAS features , the pie IPF has to content with shrinks rapidly.
Why ? What's changed except a process shrink to 65nm to warrant such optimistic expectations.
Well Montecito ran in the lab at 2.3GHz in FFM , TDP was probably 250w.
With Foxton Montecito was expected to reach 2.3-2.5GHz , Montvale 3GHz at 65nm.Obviosly , 2008 =! 2004 hopes and Montecito/Montvale ended up at 1.66GHz.Truly pathetic.
Tukwilla was done by the same people that did Montecito , obviously they knew back then ( late 2004 ) what were the problems with Montecito@Foxton.
Since CSI , IMC were done by exDEC , the HP+Intel team had to put 4 modified Montecito cores on the same die and link them to the router.It wasn't easy , but it wasn't designing a core from scratch either.
Since Tukwilla involved little uarch work I expected them to fine tune the Montecito core , make Foxton work and so on.Apparently they did squat on all accounts.
They had 3 teams : HP Fort Collins , Intel's own and exDec.That's probably over 500 of the finest CPU engineers.Why weren't they able to fix Foxton and make Tukwilla rock ?
Market Data 
Markets 
