Originally posted by: luis infante

hello

My question is known the specifications of a computer to use with solver cplex, ¿is necessary a Intel Xeon Processor o is equal use a Intel I7? , the memory RAM ,  32 gb ¿is enough or more is recommended?, the size of the problems is more of 2,000,000 of constraints and  20000 binary variables

I actually use a computer with CPU  Intel Xeon 2.1 ghz- quadcore - two threads  with 16 gb of RAM and other computer with CPU Amd  FX 4 ghz - eight core and 32 gb of RAM  both computers with Windows seven but i think buy other computer in the future

Other question is about of Operating System ,¿ what operating system is recommended?, ¿the perfomance is better with linux?

Originally posted by: fbahr

> My question is known the specifications of a computer to use with solver cplex, ¿is necessary a Intel Xeon Processor o is equal use a Intel I7?

Most likely, true multiprocessor machines (especially when you think of taking advantage of CPLEX's new distributed parallel MIP optimization feature) will outperform otherwise comparable multicore machines.
- http://www-01.ibm.com/support/docview.wss?uid=swg21400083

> the memory RAM ,  32 gb is enough or more is recommended?

Hard to tell...
- http://www-01.ibm.com/support/docview.wss?uid=swg21399933

> Other question is about of Operating System ,¿ what operating system is recommended?, ¿the perfomance is better with linux?

[As long as you pick a 64bit system] ...(overall) performance shouldn't differ that much.

--fbahr

Originally posted by: Mark L. Stone

Per "Comparison of CPLEX performance on dual core machines with true multiple processor machines." http://www-01.ibm.com/support/docview.wss?uid=swg21400083  "A dual core processor is a CPU with two separate cores on the same die; a bi-processor machine has two distinct processors. This observation means that a multi-processor machine has two separate CPUs with their own resources, whereas in a dual core machine, resources are shared. Indeed, dual core processors still have only a single bus for grabbing system memory."

Is sharing a memory bus the primary or even only reason why use of a multicore processor may be slower than use of multiple processors?  What if the memory bus is fast enough, so that for example with 2 cores, the memory bus can accommodate grabbing memory for both at maximum memory speed and still be within capacity of the memory bus?  For instance, might the maximum rate for reading memory be < 29.85 GB/sec and the memory bus can do 59.7 GB/sec?  Does that type of scenario exist, for instance with many recent Intel processors having a 59.7 GB/sec memory bus?  I have no idea how fast memory can actually be grabbed in such a configuration.