Redhat! Nice, this is big time support. Isn't Redhat the biggest linux vendor?
Nice catch! Linux is a natural fit for Sledge, given the iconoclastic nature of the beast, and that linux users (read developers) are very price sensitive.
Microsoft is a smart company. They know that the key to their success has been legacy software support, and cheap platforms. I'd expect them to hedge their bets.
I have always believed that HP is really trying to pull a fast one on Intel and the rest of the IA-64 crowd. HP after all is giving, as in free, these cpu designs to intel. Why would a company do that? HP says that their familiarity with the architecture will give them a competitive advantage in making such systems. Sounds a little vague to me. I think it's because HP is only giving half the VLIW key to intel.
VLIW is philosophy in which the software (read compiler) is smart, and the cpu is dumb. The cpu is dumb as in that it just executes the code as it comes, without silicon doing some run-time out of order executions, or cache control, or branch prediction, or otherwise think for itself. The idea then is that the maximum amount of silicon will be used for actual data crunching. All the stuff that was left out is now supposed to be the job of the compiler. Thus the compiler now needs to be very smart. This could work if the code were very simple in its execution- say a benchmark or certain numerical apps. But I think it's absolutely impossible to adequately predict the future at compile time with fancy self morphing code, like run time binding (scipting languages used on servers), or Java (JIT excluded). For big supercomputing tasks, and numerical simulation, the smart compiler could be ok. But for servers, I don't think VLIW can ever work (ie give superior performance), without run time code optimization.
The other half of the puzzle is what Transmedia has included with their VLIW chips- the morphing software part of it. Run time profiling and adaption. The task is similar to developing a JIT JVM, except much more demanding.
HP has this technology for their RISC chips. They are doing it for IA64. This is their wild card. Only HP will be able to combine it into an IA-64 platform solution for customers. And as both the chip designers, and as a company which has pioneered the technology, HP will be way ahead of the curve. I doubt HP will give this part away to Intel to include with their chips. Microsoft, Sun, linux, and others will be SOL.
I bet Intel tries to buy Transmedia. |
