NaS, for a server to server comparison, you can easily look at the technical specs for both units. What you seem to be asking from me is a complete education on the topic, and I simply don't have the motivation to give one. RAS (or RASUM - Reliability, Availability, Serviceability, Upgradeability, and Management) is a complicated topic, and it could take a long time to even scratch the surface. I've worked extensively with these kinds of features, and it took me months to learn the finer points of even a few of them; so, suffice to say that an Internet forum is not the appropriate medium to learn about this topic. I gave you some links so that you can research some features by yourself. It is up to you to find out if Opteron uses similar features. Newisys came closest to offering enterprise level RASUM (on the DP level, of course), but as I said before, their business model overestimated the demand these kinds of features would have in the whitebox market (where low cost is preferable to expensive and feature rich).
I think I can say more conclusively that Opteron does not have the same RAS features that Itanium has, simply because no one has put in the R&D to add these features. IBM hasn't. HP hasn't. Sun hasn't. AMD put some minimal RAS into Opteron's memory controller, but they will need to add more as time goes on, since they will lose differentiation as enterprise level RAS features make their way down into mainstream offerings (as they typically do - I can see the future of mainstream RAS by researching high end RAS). Decoupling the chipset allows more flexibility when it comes to adding differentiating features - something AMD does not have unless they sacrifice performance by disabling the on-die memory controller. I highlighted these things in my previous post.
Part of the benefit of getting high profile design wins from major OEMs is to have these OEMs put in the R&D dollars to generate some differentiating features. I know that Itanium is getting plenty of R&D dollars, both from Intel, and other OEMs. HP has been on board for two years already, and Fujitsu will be fully on board by 2005. IBM is still lagging, however, because they put their research into Power4, while selling Itanium and Opteron into generic compute nodes. This is supposed to change eventually, if IBM gets their own Itanium chipset to ship, but so far, I haven't heard much more on this. Though neither have I heard much enthusiasm over Opteron. As usual, IBM is hedging, while putting their money on their own horse.
AI
Market Data 
Markets 
