Opteron and A64 are based on the same architecture. The Opteron can only be used on a Socket 940 MB, and requires ECC Registered RAM. If you get the 2XX series of Opterons and a dual CPU MB, you can run a multi-processor setup.
The Opteron is almost the same as the A64-FX series, but the FX has a fully unlocked multiplier that allows overclocking. For a business application, the overclocking probably won't matter and the Opteron is a much better choice since it is less expensive than the FX. The Opteron 150 and FX-55 are equivalent, while the Opt 148 and FX-53 are equivalent.
If your programs support multiple processors, or if you do multiple things on a single computer (for example, run an FEA analysis in the background and still hope to do anything else), then you may want to look into a Dual Opteron setup. If you don't care so much about the dual processor stuff, then you have the option to go with an A64, although you may still want to go with an Opteron. The main reason to go with the Opteron is that the 940 Mb's typically support more RAM than the 754 MB's. Most 754 MB's are limited to 3 GB of RAM, while the 940's can handle 8 GB (and some dual CPU versions can handle 16 GB). Other than those differences, the normal A64 and the Opterons perform very similarly.
Both the A64 and Opteron will run on normal windows.
So, now that you know the differences, you may want to provide a bit more details about the work you are doing. For example, what is your solid modeling program? Are you running FEA? If so, what FEA program are you using?
I am a mechanical engineer and run Pro/E and Mechanica (FEA) on a P4 1.8 GHz with 1GB RDRAM (and I have one of the faster computers at work. Until a few months ago I had a P3 700 with 512 MB RAM) using Win2000 as the OS. I use an Elsa DDC Video Card (Quadro based, GeForce 2 Equivalent, I think)
I design robots, and can easily run assemblies with 100 or so parts, but with more than that the CPU has a very difficult time with Hidden Line Removal. The video card is much, much faster than the CPU and can rotate the shaded model better than the CPU can remove hidden lines from a wireframe version.
I can run FEA with about 40,000 elements, but that is limited by RAM. A 40,000 element analysis will usually take a couple hours, and while it's running I can't really do much else on the computer (I need to close pretty much everything to free up as much RAM as possible). A 15,000 element analysis will finish in about 1/2 hour, and since it doesn't max the RAM I can usually do some minor work on the side (Excel, Word, etc.).
Anyhow, based on all that, my recommendations are as follows:
If you need use the system while number crunching:
Dual Opteron 246 or 248 (this will allow you to work on other things while the number crunching is going on in the background, or may speed up the analysis if it supports multi-threading)
2-4 GB of RAM (total - 1/2 for each processor)
WD 74GB Raptor HDD (unless you can go for SCSI, in which case go for that).
Tyan Dual Socket MB (not to familiar with Dual CPU MB's, but I think Tyan makes a very nice one)
If you don't need to use the system while number crunching:
Single Opteron 146 or 148 or A64 3000-3400
2 GB RAM
MB Depends on socket choice.
Anyhow, it will depend on how big your models and analyses are. As you can see, I am able to do quite a bit with a fairly low end computer (by today's standards), so if you can give an idea of the scale of your designs that might help. For video cards I would definitely recomend an Nvidia Quadro based card, but that's because I have had good experience with them and compatiblity problems with 3D Labs (the only other card I've had experience with).
Also, to you have a price limit? That might help narrow down the options and the tradeoffs between what CPU, how much RAM, what HDD's, which video card, etc.
-D'oh!
Facebook
Twitter