What is it about these POWER servers, such as the 750 Express, that makes their hardware so much more expensive? When comparing specs with their lesser serversi do see more performance, but not $35k more!
These appear to be utterly different machine classes. I’m not sure how you think there’s some valid apples to apples cost comparison here. It’s like cost comparing a car and a dump truck based on raw horsepower.
Other than the POWER7 sockets, what else is different? DDR3 is DDR3 isn’t it? NICs are NICs, SCSI and SAS and USB are all universally followed “protocols” aren’t they?
The POWER7 processors themselves are wildly expensive. Back in last September, systems using them were described as triple the price of ones with Intel Nehalem processors.
(from http://www.theinquirer.net/inquirer/review/1532674/power7-vs-nehalem-ex)
Aside from that, the POWER7 was designed for use in massively scalable supercomputers - the design was for all practical purposes commissioned by DARPA.
Comparing this to an i386 processor such as the Intel Nehalem inside your cheaper reference is like comparing a bicycle to a train. Yes, they both have wheels and yes, they both can carry people, but one can move a whole lot more and a whole lot faster, and it’s easy to add a car or two to a train to add capacity.
Reason i asked my upgrade from Tivoli 5.5 to 6.2 later this year. I “inherited” this project and found out that the IBM p520 is outdated so i was going to replace the server as well. I just poked around a little bit at the latest series of IBM servers and hence the question.
Regarding the x86 versus Power7 architecture:
I just have to decide if i want to keep AIX 5.3 or move it to Linux or Windows. The p520 won’t run AIX 6, so there is a lot i need to consider. Thanks for the info.
I think other people have already addressed most of the issues, but I can add something to this. It is not necessarily the end-points (NIC, PCI slot, etc) that make the difference, it is the interconnects. High-end server class systems will have dedicated buses and/or controllers for each of those I/O connections while lower-end systems will share them across multiple ports. This allows each connection to run at maximum capacity concurrently providing far greater aggregate throughput.
I’m guessing (based on age/size of p520, etc.) you could get by with the x86 box. If you need more power in the future, then the Power server will scale much better than x86 servers.