Wednesday, November 30, 2011

Computer Care

Overall:
No matter how clean you keep your room, dust will start to build up in your machine. Every three months or so, take it open and wipe the dust off with cloth.

Static Discharge:
Always ground yourself by touching a piece of metal grounded to the floor before you touch any computer equipment. Computer equipment are designed to withstand at most around 19volts, it will not survive a static discharge at thousands of volts. You -will- fry your equipment dead if you don't take the proper precautions. This is especially important if you're working on carpet or lacquered wood flooring, where static discharge will collect at a much higher rate.

Any anti-static bag that you obtain should be kept, as these offer protection for your silicon gold bricks if you need to take them out of your case.

Motherboards:
Motherboards do not need maintenance except for BIOS Updates. and BIOS Updates are unnecessary unless something breaks. Generally, don't screw with your mobo, especially if your computer came with windows. That Windows installation will be an OEM (Original Equipment Manufacturer) edition tied to your mobo, so if you brick (break) your mobo, you lose your windows license, and those things are not cheap to have to rebuy, and you do not want to deal with Microsoft Customer Service on this issue either.

How can you brick your mobo? If you flash your BIOS incorrectly or the power gets cut for some reason mid-flash, you will brick your mobo.

So in general, clean the dust off of it if it accumulates, otherwise leave it alone.

CPUs:
CPUs are generally robust enough that you can just leave them alone provided you didn't overclock them. If you did, this post is probably too basic for you. The thing you should monitor about your CPU is its temperature.

To monitor your CPU temperature. You can use HWMonitor or Core Temp, both of which are free Utilities. HWmonitor also shows a lot of other things, but core temp has a windows gadget addon and a graphic log which makes monitoring easier, whichever one you want. I personally use both.

The general cutoff temperature for a CPU is 60 degrees Celsius. Intel chips may be able to tolerate a bit higher, but try to keep it below if you can, especially if you're not overclocking. Also, you should verify your maximum operating temperature in the CPU's specs. The general rule of thumb is your CPU should be kept below the temperature around 20 degrees lower than maximum operating temperature.

For laptops, the value may be a little higher. You should consult your chip's specifications. Note this is not the shutdown temperature. The shutdown temperature is clocked at when damage to the CPU is imminent, and damage can still accumulate even when the CPU is lower.

So what happens when your CPU temperature is too high?

The first thing to do is to immediately clean up your case. High temperatures mean that your ventilation has been blocked to a serious degree. Open it up, take proper antistatic measures, and clean it up.

If your Case is clean, the first thing you should do is consult the manufacturer's specifications. CPU's report temperatures as a -distance- to the shutdown temperature (also known as TJ max, which stands for Temperature Junction Max). Rarely, a chip's TJ Max is recorded wrong, and thus the CPU will report incorrect temperatures. This may be able to be fixed by a motherboard BIOS update, but otherwise you can manually correct your TJ max. If your CPU is more than 20 degrees Celsius below the TJ max, you're usually good to go.

If not, then you must install additional ventilation. This usually is unnecessary on a factory-produced machine, and normal CPU's rarely need anything more than the stock cooler to stay at a safe temperature. If you see additional fan grates on your case though, consider buying additional fans.

It is also advantageous to make half of your fans face inward to create a wind tunnel, which is much more efficient at cooling down CPU's.

If none of these things work, your CPU may be aging. If you cannot replace it, consider underclocking it to squeeze more longevity out of it. Like the name says, you will lose performance.

You can have windows limit your CPU power through Control Panel-> Power Options. Under advanced, you'll see an option for processor power management. Reduce the power to a level that you no longer go above the safe temperature.

If you don't have this option, you'll need to manually do so by going into the motherboard BIOS. This is the same as overclocking, so you should read an overclocking guide. I will tell you how the CPU clock is calculated:

CPU Clock = BUS Clock * Multiplier.

BUS Clock is a the basic clockspeed for most of your components in the computer. For most modern machines, it's 200Mhz. Thus, to achieve a clockspeed of 3.3Ghz, you will have a multiplier of 15.5, Most machines have the BUS speed at the lowest possible, so you will need to reduce the multiplier. Calculate your multiplier by dividing your CPU clock by the BUS Clock, and select multipliers lower than that value until you reach one where your CPU no longer goes above safe temperature.

It is a good idea (if only for the peace of mind) to stress test your CPU once in a while to see if it is stable. Problems are better addressed earlier than later. Stress Testing your CPU will introduce some wear on it, so do not do it too often.

A nice test is Prime95. Download it and run it for 20 minutes. It will ask you which test to use. Choose in-place large FFT. If your CPU can run it for 20 minutes without error and stay below the safe temperature, your CPU is healthy. If it goes above safe temperature, then it is a small warning sign of wear. (some factory computers are not designed to run at 100% power for that long, so if your brand new comp goes above, it's generally not a problem.) If, however, it makes a calculation mistake (and you haven't overclocked your CPU), then you must replace it.

Graphics Card:
If your card has a fan, clean it regularly along with the rest of your case, but otherwise you can't really maintain a card. You can use HWMonitor to watch the card's temperature. If it goes too high, you need to either install more fans, buy a new case, or replace the card.

A good stress test is FurMark, run it once to make sure your GPU is stable. I don't really recommend doing it more than a few times.

Memory:
Memory is usually fine as it does not wear down significantly. A good, free memory test is called HCi Design's MemTest. You can use it to verify your memory once in a while. If your memory test reports errors, you should replace it.

Hard Drive:
The most dangerous thing you can do to a hard drive is again temperature. Hard Drives above 50 Celsius will start to become damaged. To monitor your Hard Drive, use a SMART Monitor such as CrystalDiskInfo.

The most telltale signs of imminent disk failure lies within 05 (Reallocated Sector Count), C5 (Current Pending Sector Count) and C6 (Uncorrectable Sector Count). If any of these values are not 0 in your smart data, drive failure is imminent.

Another sign of a bad drive is low access speed. Download HD Tune to test speed. If your computer returns a speed lower than 30MB per second, you probably will need it replaced. (Most USB external hard drives have a speed of 24MB per second, so anything lower than 30 on a mainboard drive is unacceptable.)

Unfortunately, there's no real fix to HD failure. You must keep a backup with you at all times, or if the drive started showing signs, immediately copy all your data out. Don't forget to check your manufacturer's warranty for a potential replacement. If the warranty is over, you have no choice other than to buy a new drive out of your own pocket.

If you have multiple computers, the most hassle-free backup utility is Macrium Reflect. Images made with this utility can be restored to 100% accuracy, meaning you do not need to wast any time reinstalling. Try to keep your images up to date and keep two versions of them, just in case one of them is made after your system went bad.

If you don't, however, you should have at two drives, your system drive, and your backup system drive, which should have an image of your system drive -as well as- the most minimalistic installation of a reflect-compatible windows version preloaded, so that when your system drive does die, you can just plug in the backup and load reflect onto another blank drive that'll become your new system drive.

You should have your active drives defragmented daily. A fragmented drive will wear out your drive head faster than a full defrag ever will. Why? A hard drive must spin at least 180 degrees before it can find any arbitrary random bit of data, but it can access a sequential set of sata with one spin no matter how long it is. If your file is split into 8 pieces, your drive must rotate 4 times (as opposed to half) before it can locate all the pieces. What's more is fragmentation grows faster the more fragmented the disk is, because when you delete a fragmented file, data written later to those sectors may belong to several files next time.

Power Supply Unit:
Under HWMonitor, you can find the values for the voltages of your PSU rails. If any of them start to vary by more than 5% of the voltage, you should have it replaced. The exception is the -12V rail, which can tolerate up to 10%.

It is also recommended that you buy high-quality PSUs from a trusted brand. Low-price PSU's often fail to deliver their promised power, and some can even explode. Go to JonnyGURU.com for a reliable source of information on PSU's.

Registry Cleaners
Avoid these like the plague. They do not do anything useful.

Why is that? Your registry is an indexed database of -records-. It is not code, and because it is indexed, its size does NOT matter. A 1KB registry is just as fast as a 100MB one, and most registries will not exceed 100MB. To give an idea of how much that actually is, based on current HD prices, 100MB will cost you a fraction of a cent.

Also, there is no program on earth that can do this correctly, nor will there ever be one as long as computers still follow the Turing machine model (which it probably won't ever get out of). This the corollary to the answer to Turing's halting problem, which says no, it is impossible to detect with a program whether another program will halt (aka, finish executing) unless the detecting program is larger in size than the program being inspected. Similarly, no Program on earth will be able to detect concretely if you really need a registry entry either. Thus, the effort you spend on this does NOT outweigh the risks of destabilizing your system, so just don't do it.

Others:
Uninstall anything you don't need. A program you think you're not running may still be running in the background, or even as a service. Keep your drivers and programs up to date as much as possible. Incompatible/badly installed drivers cause 95% of windows crashes/instability.

1 comment:

  1. It would also be advisable to install an antivirus program to ensure that your system is protected from software threats. There are a lot of freeware available in the internet, any of which can be trusted to protect your computer against the more common viruses. However, if you want the more stringent versions, it would be advisable to take a look at some product reviews before actually purchasing a full antivirus program for your computer.

    Benita Bolland

    ReplyDelete

Note: Only a member of this blog may post a comment.