Make sure to get a good
bang/buck ratio in whatever you buy.
Do you really need the latest, hottest, coolest stuff that dealers have listed ? Or would you settle with the latest, coolest, hottest stuff of three months ago ?
Many parts have either a speed or a capacity. (Some have both - or other measurable properties.)
And - double the speed will not mean double the speed.
Excuse me?
Yes. A CPU that is twice as fast as your current one will not make your computer run twice as fast.
You will most likely notice it's faster - but there are architectural limits and bottlenecks keeping your system from becoming twice as fast as before by doubling the speed of one component.
Maybe not enough memory ?
See Why running MySQL with 32 megs of RAM is a bad idea...
Okay - let's say you do not want the top stuff. Someting modern and reasonable will do, especially if your budet is limited.
Take CPUs. A typical ad might list the following:
Hoglon 500 $83
Hoglon 750 $97
Hoglon 1000 $145
Hoglon 1300 $160
Hoglon 1600 $320
Hoglon 2000 $982
What would you buy ?
Jerry Pournelle just wrote high praise about his Hoglon 2000 - but Chaos Manor is not what it used to be. Jerry is a computer journalist.
$982 is serious money, too.
Some research shows that they stopped making the 500 and 750 half a year ago. The 2000 is new - but they will release the 2500 in about two months.
(Which - as many journalists will tell you, totally obsoletes your 2000).
So - go for the bang/buck ratio, avoiding outdated stuff.
Divide bang (speed) by buck (price):
MHz b/b
1000 6.9
1300 8.1
1600 5.0
2000 2.0
You will notice that the latest model is priced out of proportion. The 1600 is only about 20% slower than the 2000 (an almost unnoticeable difference) at less than one third of the price.
At half the price of the 1600, you get the 1300, which is 35 percent slower than the 2000.
Should you go even lower ?
No, as the bang/buck ratio is worse.
Of course, you can shell out the extra cash for the 1600 - but buying a 2000 ?