Purchase ram up until you perceive no augment in presentation. Start with 4 GB.
You may not yet need to sprint the dual Xeon for what you’re doing. Post some more detail about what you will be operating on this.

My purpose is to run parallel MPI PDE solver / linear solvers on that. Xeon 54xx does not have QPI, so why you recommend this. You recommended RAM 800MHz; is not pretty slow?
Please give instance of both 'shitty' and ''non-shitty' HDD.

The Xeon 5400 series doesn't create much sense anymore. Go with the much quicker 5500 series. The 5520 has 4 cores plus Hyperthreading. I'd leave with Intel simply because you can promote to much faster CPUs if require be. If disk performance actually is important, leave with 15K SAS drives or Intel SSDs.

I deduction my skill on solvers is only in Ansys and Cosmos. I’m quite engineer. For examine flow and linear calculations. I have acquired by on a 2.4 xeon quad core with 8 gigs of ddr2. This all being place in a workstation atx MOBO.

Ah, you're doing similar number crunching with MPI on this machine. This denotes you're probably operating Linux or some other nix as your OS and you're most likely going to be operating the machine at full load on all cores for a long time. I'd leave for the dual Opteron 2427s in this container. Those extra four cores in the dual Opteron system should allow it to be significantly faster than the dual Xeon system in such parallel code. I wouldn't count on much extra Turbo Boost over the stock 2.26 GHz from the Xeons as they're going to be operating flat-out for a long time and limbering up so the Xeon probably has only a 67 MHz clock speed benefit over the Opterons. The Xeons are a little faster clock-for-clock than the Opterons, but you can really affect how certain code runs on a certain CPU by deception with the optimizations of the program binaries and as a result, Opteron 23xx/24xx CPUs aren't that much slower clock-for-clock than Xeon 55xx CPUs when well-optimized.

You said: 800 MHz RAM is the greatest grade of server DDR2 obtainable and it's what the Opteron 24xxs use. That's a bit slower than DDR3-1066 that the Xeon E5520 uses, but unless your code is greatly memory I/O-bound, it should be quick enough. That’s what the point: CFD/PDE solvers memory jump. So it is the matter for me! Yet you are with 2 Opteron 2427 (12 cores) with 800 MHz? If yes recommend motherboard? Tyan is not good one as the attuned one only support 667MHz.

If your application is memory I/O-bound, then you'd be enhanced off with the Xeons. Each Xeon E5520 with its three channels of DDR3-1066 has approximately twice the memory bandwidth as an Opteron 2427 with its two channels of DDR2-800 (25.6 GB/sec vs. 12.8 GB/sec.) Both have considerably more effective memory bandwidth than your current Xeon quad-core unit does, excluding the Xeons would be an improved choice if you are at all concerned about memory bandwidth.

That’s okay.

Than your present Xeon quad-core unit does. Which Xeon quad core you unspecified I have??Can you please inform me memory bandwidth of Xeon3220?

The Xeon X3220, since it's the only quad-core Xeon that sprints at 2.40 GHz and you said you have a quad-core Xeon operating at 2.40 GHz.

The Xeon X3220 has a 1066 MHz effectual FSB, fine for a maximum of 8.53 GB/sec of bandwidth. Though, that bandwidth needs to be shared not only between the CPU cores and RAM but also between the two couples of CPU cores on the two dies in the CPU, so you possibly won't be using all of that bandwidth just for RAM accesses.