Hardware Game: Round Three

Sunday, 20 March 2005

I had promised myself, privately and in confidence, that I would never build a PC. I did not have time to go picking up all that hardware knowledge and lingo. I am an artificer of software, I calmly reassured myself, and have no time for the ways of the overclockers, who speak in tongues of mysterious jargon. I believed I would find comfort with an off-the-shelf model. Trouble was, I had learnt enough of the hardware game to realise I would not find the comfort I sought.

So in the first and second parts of this series, I discussed how I planned my custom PC build. When I proposed that building a PC is hard, I was not talking about the physical act of construction. Sure, it is a pain, but it is nothing compared to the research and planning you have to put in to get the components you want... and the components that work.

In this, the final article, I talk briefly about putting the whole thing together and then offer my thoughts on the whole shebang.

Construction

Assuming all of the components have been purchased and delivered to your door, some serious time will then be spent knee deep in silicon and wire. Be prepared to give up the good part of a day.

There is no official recommended order to putting everything together, as it depends on the motherboard, case and components, but here is what we did:

Now there were a few complications with the assembly process.

Attaching the processor to the motherboard was a piece of cake, but adding the heatsink and fan was a truly frightening experience. The processor is the most expensive component inside the PC and it is very important to make sure that it is in good contact with the heatsink, otherwise it will sizzle and fry within seconds when power courses through the processor's veins. The heatsink came with a large black frame designed to secure the heatsink solidly against the processor, but the actual act of securing required significant pressure to be exerted by my reluctant fingers. I wanted to hear a click not a crack. My God, it is not something I want to do again for some time.

I did not like the case as much as I had expected. Apart from the lack of subtlety in the design, which was of course apparent before I bought it, there were a number of issues. The case was the heaviest I have ever encountered, something I gathered as soon as I picked up the box after delivery. My arms were not good for the rest of that particular day. Wheels would have been handy for moving it around the flat, but instead I need to break my back when relocating it from point to point.

Securing the motherboard into the case was rather difficult too, the plastic tacks did not seem to work out very well and I longed for screws. Final moan; the door on the front of the case turned out to be annoying as it has to be left open to use the CD drive.

Plus points? The USB and Firewire sockets on top of the case are great and plugging in my digital camera is a lot easier. I can also control the speed of the main case fans as well as the PSU fan. The case has pretty good airflow and the fans do not need to be on top speed according to the internal temperature sensors. The top fan setting is one I refer to as noisy speed. My new, more powerful PC, is a lot quieter than my previous one, a change which is welcome.

On Layout

It took most of a day to get the system built from scratch but it was an interesting experience. Having gone through the building process from start to finish, I now have some ideas about what I would like to see in an ideal system, from a construction perspective.

A form of power bus around the edge of the case would be good, considering that power is needed everywhere and anywhere these days. This would mean a reduced web of wires stretching all over the motherboard, less confusion. While I am on the subject of power, the D-shaped power connectors are a dog to unplug. Trying to unplug the power from the AGP card is quite scary, because the force you need to exert to pull out the plugs is more likely to wrench the card from the AGP slot on the board. I pray for a redesign of the power connectors.

I would also appreciate more components with diagnostic lights. While the lights on my Corsair memory are a bit over the top (and the next version really takes the piss), it is comforting to see that the memory is in use and working. I would like to see diagnostic indicators for other components such as drives and expansion cards. We have moved on from BIOS beeps to diagnose problems. This is the 21st century, for God's sake.

And heck, could the processor heatsink and fan be a little less frightening?

Booting

A modern BIOS will usually work straight out of box without requiring any extra configuration. I did tweak a few things - I am a power user after all - but the new motherboard detected the hard drives and components without problems. Well, sort of. Let me come back to this in a moment.

Booting up Windows XP for the first time on the new system sent it into a fit. I had plugged the hard drive from my old system straight into the new one and Windows had noticed that the hardware configuration had changed, a lot. It had to quietly sort out the drivers and work out how it was supposed to interact with this new system. To the user (i.e. me) it appeared that the machine had crashed but it was simply... thinking. Once it had got through this process it then told me that I need to re-activate my copy of XP within three days.

I have two copies of Windows on the same machine - my everyday partition and my gaming partition. To avoid having to re-activate twice on the same PC, I copied the activation file wpa.dbl from the re-activated Windows to the other one before booting it up. Note that this is perfectly legal; the Microsoft EULA states that you can not use a personal copy of Windows on more than one PC.

Everything seemed to be fine and I was really pleased that the transition to the new PC hardware had gone so smoothly. However, there were two problems which were not apparent at first.

The Details, The Devil

CPU-Z is a free program that displays information on the CPU and memory. I investigated the memory timings and the results came up 2-2.5-2-5, despite that it was expensive memory and supposed to run at lower latency timings of 2-2-2-5. However, CPU-Z also showed me the SPD table, which is the set of default timings programmed onto the memory for the BIOS to use when the system is turned on; they showed 2-2-2-5. My next step was to go into the BIOS and disable the auto-detection, setting the timings manually to what they should be. Booting up after that change, Windows went to the blue screen of death very, very quickly. Hmm.

Eventually I hit upon the thing that was wrong. The fast memory that I had bought at Corsair had to operate at a voltage of 2.75V, whereas the BIOS was defaulting to 2.7V. Once I stepped up the voltage, I got the memory working at the correct timings without the blue screen of death. This took a couple of days to figure out but there was another, more serious problem.

The reason my PC was super-charged was to play 3D games in as much visual glory as possible, so it was a surprise when games were crashing randomly. Occasionally, Far Cry would stop and I would hear a high-pitched whine through the headphones. The only resolution was to reboot the PC. It was not possible to ignore the problem in Half-Life 2, though, as it went down much more frequently. Fortunately, I did manage to find an old Half-Life deathmatch level where it would crash within seconds without fail; I had a test case.

I played around with the BIOS settings but got nowhere, so I searched on the web for similar problems. There were people out there with a similar problem, but the causes were so varied and different that I came to the conclusion that this was a standard graphics card crash - I just did not know my particular cause yet. The same graphics card had worked on my previous PC, so something in the new PC was causing the graphics card to crash. But what?

I became desperate and tried out just about any suggestion I found. As I trawled through the newsgroups and forums I became aware that there were countless other self-built victims out there, with expensive hardware that crashed too frequently to use. Recommendations come thick and fast, of course. Have you tried swapping expensive component X for expensive component Y (that you don't have)? Have you installed the latest drivers? You know, I am sure people building their own PC will have the latest drivers. There was, of course, another special category of victim: the overclockers.

These days, overclocking is quite straightforward; you just make some changes to the BIOS, to make processor, motherboard or graphics card run faster than the defaults provided out of the box. Personally, I have no interest in overclocking. I am not interested in getting a few more frames per second by risking the integrity of my processor or graphics card; I destroyed my last graphics card by using the "safe" automatic overclocking feature. I do not want to stretch the system in that way. During my web search I came across many overclockers that could not resolve system instability problems and were asking other overclocking gurus for help. I thought that maybe something in there would be relevant to my situation.

And there I found it, one of many black magic solutions to fix overclocking problems. Some AGP card overclockers recommended increasing the voltage on the card to improve stability when overclocking. Even though my card was not being overclocked, I tried increasing the voltage from 2.6V up to 2.7V through the BIOS. Surprise, surprise, the problem went away completely.

My current thinking is that the motherboard was not consistently sending 2.6V to the graphics card and it would crash when under load as a result; pushing the voltage up meant that the graphics card was always getting enough power to function. Case closed.

Take Away Thoughts

This article will get out of date very quickly but I think some of the information here will continue to be useful in years to come. As I mentioned in the first article, there are few guides to the hardware game out here on the web.

The bottom line is that building a new PC is fun but does require quite a lot of research and, even if you get everything right, you can still have significant problems once you have constructed your perfect PC.

There is always some technology in transition, meaning somewhere you have to decide whether to buy the product that will be "legacy" tomorrow or the product that is the expensive "cutting edge" with limited buying options.

Your wallet may be intrigued by the fact that a self-built PC will typically cost more than a standard build from a PC manufacturer due to economies of scale. Go for it if you are really determined but think hard about how much time you want to invest. Also consider that the money invested can be wasted if mistakes are made in the planning or during the construction.

That about wraps it up. For those less experienced in hardware, I hope you might have learnt something from this series of articles. If you found this all rather passé and would prefer to read about software development instead, don't worry, I'm not about to start Joel on Hardware.

Off you go now, someone told me the sun is shining outside.

Update 19 July 2005: My colourful PSU went bang yesterday, with less than a year of use. Luckily, it didn't take anything else with it, but I was concerned it might have. Useful PSU article to be found at Silent PC Review.

End