/. linked to some predictions for the next year. Three of them pertain to HPC.
Before I get into the three predictions, let me point out that predicting events that have already happened is not generally hard. This is important for the first prediction.
He indicates that Itanium is on life support, and that HP is trying to get out of its deal with Intel. Apparantly he is not aware that this appears to have already happened. Two years ago. Nothing to see here folks, we know what the next generation of money making architecture is for Intel and AMD. Itanium it is not.
Onto number two. The Top500 list will be dominated by Cell based machines. I doubt this, either in number of machines, or in total reported TFloppage. His point is pointing in the right direction. Moore’s law and other factors will drive 2x performance over a 2 year period, so for really hard transients, you need APUs. This was obvious at SC06 this past year, and we have been saying it for years. I don’t think the lists in 07 will have APU-based machines dominating. This will take a little time. Aside from that, the APU market has several competing technologies within it, and it remains to see which one will win out. I could make predictions here. Obvious ones. The chips which have the right mix of price, performance, ease of development, and cost of development. This immediately excludes quite a few. As a simple rule of thumb, I would postulate any APU that does not offer at least 10x wallclock application performance acceleration that sells for more than $10k including development tools, will be toast. There are a few of them that will come in way under this. I would bet on them.
His third and final HPC related point was that OpenSolaris will eclipse Linux in terms of size/users.
Um…. er…. uh…. . I am not aware of any ISVs targeting OpenSolaris for their applications. I am aware of many ISVs targetting a specific distribution of Linux (much to my and many others chagrin, as compared to targetting the LSB process) as a platform. Every ISV is in platform reduction mode, has been for a while. Reduces their costs. Everyone is supporting Linux on their hardware, so I (and many others, including the bulk of the market) expect that two major platforms are going to remain, with a third “boutique” small platform off on the side. None of these are Solaris or OpenSolaris based.
I don’t see it. Lots of the customers I speak with are in the process of ripping out all or most of their Unix to replace them with Linux. Its that old Pareto-optimal rule again. Doesn’t have to be perfect. Has to be good enough. And it is.
Won’t comment on Vista adoption, other than noting that it would require new hardware purchases for everyone.
Put another way, he is likely 3 for 3. Wrong that is…