Here's where the pain begins.

You should really consider a 500W+ power supply if you want a system to last a couple years and allow expansion (there are even 600W power supplies available now).  The wattage is what determines how much hardware can be powered - the smaller, the fewer.  400-450 is acceptable, but do you really want to start at a literal hard-stop and pay for it when you want to extend it?  Power supplies are relatively inexpensive.  Make certain to get a power supply with the proper connectors - my dual Xeons required a special connector on the power supply which made choices limited.

Definitely go 1GB (or 2GB) RAM modules.  Again, think ahead.  If the mobo has four slots (like most) and you have two 1GB modules, you can upgrade to a 64-bit OS (if possible) and add 2GB more without selling the four 512MB modules.  Most consumer mobos only support up to 4GB, even 64-bit.  Those that support more are going to be costly for both the mobo and the additional memory.

Consider FSB speeds, onboard chipsets, the chipset, and cpu L2 cache.  As mentioned, ASUS makes very good motherboards.

nVidia graphics - no contest here.  PCI-Express if possible (all my machines are, regretably, AGP).  Onboard graphics chipsets are usually minimal at best.  You need a separate card to get the advantages.  No less than 128MB - 256 or 512MB for 3D graphics/gaming is recommended.

I'm undecided between Intel and AMD these days.  My Intel dual Xeon (old dual cpu 2.66) is very nice, but my dual core AMDx2 4800 64-bit is sweetly fast.  Personally, I don't care as long as it performs and has the feature set useful to my needs.  Haven't had a chance to see the new core-duo Intel chips in action other than my iMac G5 (which is a hard comparison with Windows-based computers).

Get - a - full - tower - case (Antec is good)!!!  If you want a computer for the long haul, a mini tower is as good as getting a Toyota to haul a silver bullet and then upgrading it to a 34' Highline!  I always want space for some sort of expansion - drives, drive bays, memory slots, PCI slots.  Mobos that have 2 PCI slots in a case that only allows 2 harddrives are evil!  Only good for auxilliary machines that do, say, farm rendering.  For an all-purpose workstation, you want space for a floppy (maybe), a DVD-CD burner or two, up to three harddrives or more, and at least four PCI slots!  Also consider the USB/Firewire connections available.

Go SATA on the drives if possible.  Installing the OS with them is painful (requires drivers pre-designated), but performance is better than EIDE.  7200 or 10000 RPM, no less (5400 RPM drives are like floppies and you'll really only see them in old laptops these days).

My personal advice is avoid Dell, Gateway, and Alienware - expensive stuff that you can outdo for less.  Building a machine yourself is a learning experience, but you avoid all of the caveats, limitations, and expenses of pre-built machines.  Except for Macs, all of my computers since ... the mid 90's have been self built.  You can get pre-built from others at better prices (as mentioned).  My AMD is a prebuilt, but only cost $1500 for everything nearly two years ago when the usual cost was closer to $2000+.

Hmmm, what else.  Get efficient cooling.  Either quiet front and back fans or a liquid cooling system.  But don't go for the gimicky 'LED fans' or neon tube cooling - this is just show, not function.  Function and expansion are key here.  It is hard to determine the direction of the wind - AMD or Intel, this cpu socket or the other, this memory type of the other (DDR, DDRAM, etc.), PCI-Express or AGP.  So, read the markers and observe the reviews and news to see what has the most potential and support.

Definitely shop around.  I was able to upgrade my AMD 3500 to AMDx2 4800 for around $130 (Egghead.com - something near $700 less than other prices).  It was a deal not to be passed over.  This is where decisions get complex.  You may need to settle for spending less for a potential system and upgrading later when prices are lower than paying for the latest-greatest.  Latest-greatest tends to be expensive-costly!  If you have thousands of unwanted dollars to throw about, the LG principle can't fail (in the long run).  If not, you have to strategize with choices that are economical, but future expandable.

This leads to my epilogue.  Determining what current technologies will disappear and which will continue is not an easy endeavor.  I think we tend to forget this aspect of our computing choices over the long run.  It is not only a matter of 'BetaMAX-vs-VHS', but also one of new technologies being quickly surplanted by newer ones in very quick succession.  I tend to stay a few years behind just because it is not worth the trouble of 'keeeping up with the Jones' and pouring money into the 'next best thing since sliced bread'.  In essence, I'm saying that you should look for the latest-greatest from a year ago and buy that!  By this time, it will be at least half the cost and still be very powerful.

C makes it easy to shoot yourself in the foot. C++ makes it harder, but when you do, you blow your whole leg off.

 -- Bjarne Stroustrup

Contact Me | Kuroyume's DevelopmentZone