Mike's general stuff ~ Random

Recent article subjects: [Linux] [Hardware] [Browsers]

How not to build your own PC

I'm writing this article mainly as a guide and a warning to those who build PCs for themselves, or are thinking about it for the first time.  My advice is mainly aimed at people interested in gaming, people who need keep their PC spec parallel to the current 'latest'.

My current home PC isn't that bad now, all things considered, but it has been expensive and a pain in the process of getting it to be good enough for my needs.  The more annoying part is how its predecessor was reliable and fast from day one, and happens to now have the job of serving this website, my company website, and my email. There was a reasonably valid reason why I was upgrading, but that's a long story.

What mistakes did I make?

One big mistake really, which covers a multitude of sins: I didn't research the subject. I know a fair bit about computer hardware, and it is my business (and was my previous employment to a certain extent) to keep on top of what is currently going on in computer hardware innovation.  Admittedly the company I was working for at the time wasn't in the habit of buildings its workstations from bits, unlike the company I worked for before that, so I was just 'out of touch'.

Researching the subject involves finding out:

To explain these points - point one - finding out about what hardware is currently available, how long it has been available to end users, and how long it will continue to be available (how long before something better/incompatible comes out).  For example, someone thinking about buying a PC now to gradually upgrade in all respects for say the next 2 to 4 years would be unwise to get an AthlonXP or Pentium 4, because they're both at the end of their product lines. It would also be unwise to get a motherboard that supports a maximum FSB of the CPU you're going to put in.  The two times this is pretty much unavoidable is at the start and end of a product line.  My advice though usually is never to buy at the beginning or end of a product line, as one usually has the least idea of what is going to come next.  Hardware reviews are reasonably decent resources for this sort of information, though do not rely on any single source for your research.

Point two - the reputations of PC hardware manufacturers change all the time.  Few companies have been fortunate enough never to have had a single screw-up (I can't think of any). Screw-ups usually last for a significant length of time, six months or more, usually in the shape of a particular series of hardware (like for example the pre 2GHz Pentium 4s, which were prone to overheating and barely performed better than the last versions of their predecessor, the Pentium 3). In the hard disk market, one manufacturer may justifiably have a poor reputation (like IBM with their "DeathStar" drives a couple of years ago) for a couple of years or more, then customer confidence will return in time.  Seagate had a similar screw-up with regard to a contract with IBM for drives to go in their PC range, but now Seagate's reputation is one of the best in the market.  Though it is worth noting that everyone can have (and do) significantly conflicting experiences with a particular piece of hardware.  It is important to read up on user experiences (not reviews) as much and as widely as possible.  Do not stick to a particular forum, search the Internet for the hardware that you're interested in.

Point three - compatibility issues are few and far between, and can be an indicator of how temperamental/fussy/unstable a particular piece of hardware is going to be.  Hardware reviews can be a good source for this information, though user experiences are infinitely better.

Point four - for example, that CPUs nowadays tend to run at over 40 centigrade, as opposed to say the Pentium 3 range running between 30 and 40 centigrade.  Factors like this have a knock-on effect. The effect of this example is that system cooling needs to be handled differently.  It is now standard to have an additional case fan (so one on the case, one or two in the PSU, and one on the CPU - excluding devices that have their own built-in).  This means more noise, but it also means that there are more items to research.  Choice of power supply has also become important due to the increasing demand for power from computer hardware.

I screwed up on all points except the third :-)  But really that was down to luck.  My choice of motherboard manufacturer (ECS) was particularly poor.  I found out I couldn't run NT4 anymore after I got the hardware, due to the motherboard chipset being an SiS (735 chipset) one with no NT4 IDE drivers (the default NT4 IDE drivers are very poor and old, barely supporting DMA properly even after hacking). I was to find out later that even though my CPU supported SSE, my motherboard BIOS did not.  ECS still haven't resolved this issue, and ignored my emails to them on the subject, so I had to resort to an unofficial, third-party BIOS update (which thankfully has been reliable, and provided a few other desirable features of the motherboard that ECS neglected to enable).

I also made the mistake of buying (though at the time, this was the latest revision at the time) a Palomino core AthlonXP, which required a noisy CPU fan (running at over 5000 RPM) and still ran at a temperature of over 50 centigrade even when idle, getting up to a maximum of over 60C when busy. While the CPU is designed to handle temperatures this high (which is a good thing to have come out of the current crazy generation of x86 CPUs), the heat it produces will take more effort to be exhausted from the system, and will heat up other components which may not like being heated that much.  I am sure that hard disks last longer if they are kept cooler, though due to circumstantial evidence of every failed disk in my experience typically being too hot to hold or even touch.

My current CPU is exactly the same speed as the Palomino core, but the newer Thoroughbred core, which runs at an average of about 43 - 44 centigrade, and only needs a quiet cooler (~3000RPM, solid copper). My PC is still too loud for my liking, but the noise is a lot less grating to the ears than it was. High-pitched whines from computers are very irritating.

How I came to be using my current CPU and cooler is a story in itself.  After some research (though not enough to save me the following purchases) In an effort to quieten the total system noise, I spent about £60 on different heatsinks, then about £50 on the new CPU, about £30 on a new (aluminium alloy rather than lead (alloy?)) case, and £30 on the new heatsink resolved the issues. That's £140 wasted on things I should have got right in the first place.

I very much hope that Intel and AMD will realise sooner rather than later that people need to sit reasonably closely to computers, and don't want them sounding like jet engines or vacuum cleaners. In my opinion, this is more important than performance, and is currently a factor in influencing me to investigate buying an AmigaOne (a PPC-based machine).  AMD seem to have the right idea, as their current 'next generation' CPUs (the Athlon FX / 64) generate approximately as much heat as the later AthlonXP CPUs, which in my view is borderline acceptable.  Intel's latest P4 (Prescott core) allegedly has a heat dissipation of over 100W, which may damage other system components as well as sounding like a vacuum cleaner.