Tony,
You wrote "Do you think S390 will still be here in 15 years though?"
Yes, not one little tiny shred of doubt about that.
You also wrote "IBM S390 processors are not slow."
I disagree. I work with a couple of S/390 mainframes and benchmark programs against the PC on a regular basis. The fastest mainframe I work with runs code about as fast as a PIII 550.
You wrote ". It's hard to compare MIPs with whatever in a Dell."
I just run the actual application that will be used in production on the mainframe and the PC and compare the numbers. It is the same program, same input files, same output files, different machine.
You wrote "So what's with the recent articles about a single S390 running Linux and replacing 50 or 60 Sun boxes?"
I believe you are confusing the S/390 with IBM's new Z900 line. I believe the z900 mainframes have as many as 4096 processors in them. No Sun box has even remotely this many processors in them. A Z900 can replace 50 Sun boxes because it has 50 times as many processors in them.
Now THAT could bring in a pretty penny for Intel. A box with 4096 $1000 Xeons in it would be very cool. I've not heard any talk about 4096-way IA-64 motherboards though.
You wrote "A 16, 24 or 32 way McKinley box will not go cheap, should be well up in the hundreds of thousands each. The processors should scale up in cost also, as the big cache Xeon does today."
Which takes us back to my current Dell example. The machine might cost $500,000, but Intel might see only $50,000 of that. And if large cache chips are not needed, then Intel might only see $5,000 of that.
You wrote "AND, you may get to replace all those old 370, 390 and all the clone boxes."
Without backward S/390 compatibility, you probably don't get to replace a single one of those old boxes. If they were not running non-portable code, the programs would have long ago been moved to PCs and the mainframes turned off.
Mike |
