Menu Home

Building PC’s 101

I’m often asked “Ra – what xyz should I get?” or “how do I build a PC?”

There’s a lot of knowledge out there about it, so why should I bother reinventing the wheel? Well, there’s many different types of wheels and sadly there’s a lot of usage of the old wooden carriage wheels instead of some meaty Pireli goodness. There’s more than one way to skin a cat and I’m here to sell you some Pireli’s.

Today we look at the most overrated and misguided question:
What CPU should I use? AMD or Intel?
*sighs longingly*

Let us start from scratch. Who remembers the story (or was it a song?) about the man who built his house on sand and the man who built his house on rocks and which house went bye bye first? The one who built his house on sand right?

Building a PC by choosing the CPU first is much like that – the CPU is important but in reality is superficial in the larger scheme of things. This brings us to our first maxim:

Every house should be built on a good foundation. Every computer should be built on a good foundation. Every system should be built on a good foundation.

When I say system I dont mean computers, I mean systems in the broader sense of the word. Your body is a system, a car is a system, a business is a system. Inputs, Processing, Outputs. That’s what a system is at the fundamental core.

So where am I going with this? Many people misguidedly choose their CPU based on propoganda: “x is faster than y, y runs hotter than x and is unstable” etc Look, it’s very hard for a CPU to be unstable in and of itself. Ludicrously good engineering and QA/QC tends to make sure of that. An Intel CPU in and of itself is no more stable than an AMD, VIA, Transmeta, Alpha, ARM…

So what could it be? What could give the likes of AMD such a bad reputation for being “unstable?” Let us look back to the house on sand analogy. The house was fine (the cpu is fine) but the foundation was not (the motherboard is not). The house ultimately failed (the cpu ultimately failed) and got the bad rep for its troubles.

And Intel propoganda was right for some time.. chipsets that supported AMD CPU’s were not that crash hot for a long time, and even now are prone to the occasional hiccup. I certainly believe that AMD CPU’s are extremely capable, but they’re often held back by the chipsets they’re paired with.

This isn’t an Intel fanboy talking either.. I’m an unbiased realist as you may soon realise. I use both AMD and Intel CPU’s without bias or judgement – they’re chosen on their merits and paired with the best possible equipment that can be afforded.

Intel itself is not innocent – like AMD their CPU’s are manufactured to insanely good levels, but there have been a few bad chipsets out and about.. some of the earlier VIA and SIS chipsets were real horrors. And so the CPU decision should be less of a factor in the systems design process than it generally is. The thing is that CPU’s are all engineered and manufactured to an insanely tight specification. Motherboards on the other hand could be engineered by a pack of monkeys or engineered by world class engineers. Yet you would prioritise a CPU decision over your motherboard decision? What use is the fastest whizpop wopbopaloobopXP+5000 gold edition xxtreeeme CPU when you’re putting it into a motherboard that’s about as stable as soggy bread?

When designing a system, first get a pricerange. Then get a list of NEEDS. What will it be used primarily for? Every good engineer plans a system with a structured and well planned approach, and so you must first diagnose the requirements of the system and design for that. Factor in longevity and scalability – we’ll touch on these later.

Chances are you’ll be looking at either Intel or AMD. If you’re looking at anything else you’ll probably be after something specialised, so you’ll more or less know what you’re doing. So on this tack we will design, on paper, two systems side by side: One Intel, One AMD. Then we compare those systems against our list of needs, budget, compare bang for the buck value, take into account platform maturity and projected longevity and then we will make our choice.

So you dont build a house with the roof first do you? You start with the foundations, in this instance the CASE and PSU. DO NOT SKIMP HERE, ESPECIALLY ON THE PSU. A cheap or flakey PSU can be the difference between a stable system and a potentially dead system. So buy a quality case – it’ll cost more but it’ll come with a quality PSU (or in some cases you’ll have to buy a PSU at an added cost.) Some good brands to watch out for are Lian-Li, Aopen, InWin, and Antec. PSU manufacturers to keep an eye out for are Enermax, Fortron, Sparkle, Topower, Antec, and Aopen. There are others but I’m not going to compile a huge who’s who list. Make sure the PSU has all the required connectors (ATX-P4 connectors, SATA power if possible)

Once you’ve got your case and PSU lined up, it’s time to move on to the motherboard. Get the most recent motherboards to suit both CPU’s – One for Intel, one for AMD. Some of you may have a preferred motherboard manufacturer, most of you wont have any clue so it pays to be safe: Stick to the big manufacturers – Asus, Abit, Gigabyte and MSI. Alternative goodies include Soltek, Shuttle and Epox. There will obviously be several contendors for the Motherboard spot, the advantage with doing it at this stage is you tend to have more of a budget for the Motherboard, instead of buying it as an afterthought with what little peanuts you have left.

Research on the best chipsets at the time, look around on forums for feedback but be aware that people do not tend to post about positive experiences – they just go on with life. People that are having problems however will be more inclined to post, either to ask for help, or to state their opinion on the chipset or motherboard. (These opinions almost always end with the words “fucking sucks.”) Choose a chipset or chipset pair wisely – go for the one(s) that have the least bottlenecks and the best features. But too many features can be a bad thing; One of the prime problems with early nVidia nForce2 southbridges (MCP’s) was that there was just too much cramped into such a small square of silicon.. and so one section of the chipset tended to mess with other sections and the whole thing also tended to overheat a bit.

One silly argument against this stage is “duh you could mismatch the cpu to the motherboard! I put a socket370 CPU into a socket7 motherboard and it didnt work!” Well, first of all if you do that, you’re an idiot and you should not be within a 5 mile radius of civilisation, because you are making our species stupider just by existing. And of course natural selection is not allowed to take its natural course these days so you’ve been saved from certain death more times than you realise. Secondly, you’re going to buy one of the latest motherboards out, it’s virtually impossible for you to mismatch an equally new CPU to that! If in doubt, ask on a forum.

So now we have our foundation in place, we come to our next maxim:

A well designed system has the least bottlenecks possible

At this point you’re probably itching to choose your CPU, but hold your horses Sonny-Jim. You must identify the potential bottlenecks of the system, because again, what use is the fastest CPU money can buy if data is getting to it through a bottleneck. The CPU may have a bus interface comparable to a garden hose, and your poor design has only a drinking straws’ worth of data going into that garden hose. Sucks to be you.

The Front Side Bus is not the biggest bottleneck in a standard PC architecture. Most often these days it’s the IDE/ATA bus followed closely by the PCI bus. That’s right – I’m suggesting you select your hard drive before your CPU! This follows the bottom-up approach taken by choosing the motherboard first.

Now we come to RAM. Fairly simple here – some ram is good, some ram is crap, some ram manufacturers are good, some ram manufacturers are crap. First point to note is to find out the largest possible size that ONLY ONE of your ram slots on your motherboard can handle. If each slot can individually handle 512meg memory sticks, then get 512meg memory sticks. If each slot can individually handle 1gig memory sticks, then get 1gig memory sticks.

Why is this? Longevity. Two to three years down the track your computer will probably be running dog slow.. the latest version of winblows has taken care of that.. so you decide to plonk in some more ram to ease the situation. With a full bank of say 256meg sticks, in 4 slots, adding up to 1gig, you will have used up all your slots and will have to remove one to put in a larger stick. What are you then going to do with the spare stick? On the other hand if you put in one large capacity stick now, you’ll have slots to spare when you go to upgrade.. make sense?

Get plenty of ram too, take a look at what the current joebloggs computerstore is speccing out at and get double that. Again this helps with longevity because multitasking is becoming more and more commonplace and ram helps considerably there. I remember when 512megs of ram was crazy talk, hell I remember when even 128megs of ram was crazy talk (I purchased 128megs of ram for my parents computer, a K6-2, back when 128megs of pc100 was $550 per stick.. it took the scent of my credit card to convince the sales guys that I wasn’t insane) and hell, even 512megs now is a realistic operational minimum.

At this stage if you’re a gamer or a pro graphics editor or something similar, you’ll want to move on to your video card. Nothing major here. Just avoid PINE and EAGLE video cards if you can.

Now you can choose your CPU. It’ll no doubt be FAR slower than you originally anticipated but that’s not a bad thing. Your system will be more stable, possibly faster, and will have a greater lifespan meaning you’ve made a smart investment. You may have a slower CPU now but that doesnt stop you from upgrading in the future! And by having a good motherboard at least you know that bios releases will be made to further support new CPU cores and speeds, as well as having some semblence of support should you need it.

Some of you may find this advice hard to swallow, because many people habitually design CPU-first in a brute force method. The problem with this method is it opens up the potential for things to go wrong. Worrying about the CPU over everything else is akin to worrying about the roof of a house being built and whether or not that roof will keep you warm and dry. A roof is effectively useless with a poor or nonexistent foundation and walls.

Some of you are saying “you’re full of it! you’re nuts! you HAVE to choose the CPU first!!!” But who says so? Where is this a written law of PC Architecture Design? Nowhere. So think outside the box. Don’t choose your curtains before the foundations are laid.

So to summarise on todays lesson:

Plan from the foundation up
CPU is not everything
Build two simultaneous PC’s on paper, compare them, and make an informed and unbiased choice
Reduce bottlenecks as much as possible
Plan for longevity/scalability with a clear upgrade path
You get what you pay for – try to buy the best possible gear!

Try it next time you build a PC.. you might be surprised

Categories: Write Ups