The PC Tune-Up Super Guide
Keep your computer running lean & mean.
-- by Karen Kenworthy

RAM
Hard Drives
File System
Video
Comm Ports/Modems
What's the Problem?
A Well-Stocked PC Toolbox
Online Service Stations
Diagnostic Darlings

I t may have been your father. Or a favorite uncle. Or maybe the high school gym coach who taught your Driver's Education class. But somewhere, sometime, someone probably gave you this advice: If you're going to drive a car, make sure you know something about how it works. Maybe not enough to open your own garage, but at least enough to make simple repairs and to realize when to seek expert advice. That same counsel applies to computers. Some of you may become experts, doing your own upgrades and major repairs-or even building your PCs from parts. Even if you prefer to leave repairs and upgrades to an expert, you should still know the basics about your machine, and be able to recognize serious problems before you find your PC-and your data-in the breakdown lane. To keep your machine healthy, you must learn to recognize the symptoms of many common computer hardware problems. For those of you w ho prefer to get under the hood, we'll also discuss ways to make an exact diagnosis and some repairs.

RAM

I t's hard to imagine a component that works harder than RAM. These circuits are involved in almost every activity inside your computer. The results of every calculation performed by the CPU are stored in RAM. Every byte read from or written to a disk passes through RAM. Keystrokes, data transmitted and received by a modem, data sent across the network, even the images we see on our screens, all pass through your PC's RAM.

With all this activity, it's not surprising that undetected RAM failures can cause a wide variety of symptoms, often masquerading as a fa ilure of almost any other system component. Because many RAM failures are intermittent, the symptoms often come and go, confounding your efforts to pinpoint the problem. But such failures also provide an important clue. When mysterious problems such as system hangs, program crashes and general protection faults (GPFs), and corrupted data pop up, test your RAM.

Several RAM testing programs are available. The thoroughness of their tests varies considerably, as does their ability to identify RAM problems. The Windows HIMEM.SYS driver (which will test RAM unless its command line in CONFIG.SYS includes the /TESTMEM:OFF parameter) works by simply writing known data to each byte of RAM, then reading the data back. If the data does not change, HIMEM.SYS assumes the RAM is functioning correctly.

But most RAM defects are tougher to detect. That's why advanced diagnostic programs, such as AMIDiag from American Megatrends, perform a battery of tests on a computer's RAM. AMIDiag checks the operation of certain parts of the RAM circuitry and creates worst-case scenarios that stress the RAM and its associated circuitry. These stress tests reveal subtle or intermittent flaws.

AMIDiag first tests the address lines the CPU uses to specify which RAM location will be read or written to. If one or more of these lines malfunctions, data may be written to the wrong location-destroying overwritten data. To test the address lines, the diagnostic program first writes the same known value to every byte of RAM. It then attempts to write a different value to a single byte of RAM. Finally it searches all of RAM, looking for this new value. If it finds this value where it doesn't belong, one or more of the address lines must be defective. AMIDiag repeats the test several times, varying the address where the single-byte value is stored. This repetition fully tests the operation of all address lines.

Next, AMIDiag performs several pattern tests. These tests store and retrieve various bit patterns, checking for stuck bits (bits that are always 1), dead cells (bits that are always 0) and other data-dependent problems.

AMIDiag's walking bit tests flood RAM with all 1s (or 0s), then shift a single bit containing 0 (or 1) through all of RAM. The checkerboard pattern writes bytes consisting of alternating blocks of 1s and 0s. The pseudo-random pattern test uses randomly generated sequences of 1s and 0s to test the RAM.

AMIDiag also looks for RAM refresh failures. Almost all the RAM found in PCs today is dynamic RAM (DRAM), a type of memory circuit that suffers from electronic amnesia. SDRAM and EDO DRAM are variants of DRAM. Data stored in DRAM disappears after about 20 microseconds.

To prevent data loss, a special circuit in your PC continuously reads your RAM's contents and refreshes it by writing back the same values it reads. As long as each byte is refreshed before it loses its data, all is well. But if the DRAM's refresh cycle takes a little too long, an occasional bit can be forgotten.

PCs contain static RAM (SRAM) as well as DRAM. This type of memory doesn't need to be refreshed, and can store and retrieve information much faster than DRAM. The computer's SRAM always contains a copy of the data stored in the most frequently or recently accessed portions of DRAM. The computer only accesses the slower DRAM when it can't find the data it needs in the SRAM cache.

AMIDiag can test your computer's SRAM as well as its DRAM. But most diagnostic programs are limited to DRAM testing. You can perform some SRAM tests on your own. If your computer's BIOS setup program allows you to enable and disable your SRAM caches, you can identify SRAM problems using just your DRAM test software. The trick is to perform all tests with and without SRAM caches enabled. If the problems disappear along with the SRAM, the cache is failing. If the problem persists whether the cache is enabled or not, the DRAM is the culprit.

Identifying a memory problem is important. But that alone doesn't put your computer back into action. You can't repair a RAM module; you must replace it. If your computer has only one module, replacement is easy. Buy a module of the same size and speed, remove the old module, and insert the new one. Just take care to avoid damaging the old or new modules with static electricity.

But if your computer has more than one module, you'll be lucky if your diagnostic software will identify the actual DIMM or SIMM you should replace. In those cases you must buy one new module, swap it for one of your computer's modules, and repeat your memory tests. If all tests pass, then by sheer luck you've replaced the defective module. But if the memory tests continue to identify a memory problem, reinstall your original memory module and use your new memory module to replace a different old module. Continue testing and swapping modules until your memory passes all tests perfectly.

Hard Drives

H ard drives sometimes fail dramatically, accompanied by a cacophony of noise or vibration. This may indicate that the normally floating read/write heads are scraping furrows across the drive's platters, plowing under your programs and data. Or the bearings that support the platter's spindle may be failing. At 5,000- to 10,000rpm, a bearing failure can quickly turn an expensive drive full of valuable information into an overheated paperweight.

Less spectacular failures may cause unexpected delays or pauses. Most drive hesitations are benign, caused by the flushing of hard disk caches, virtual memory accesses of the swap file or other background hard disk activity. But repeated unexplained delays might signal automatic retries, which means your drive has trouble retrieving previously stored data. Flaws in a platter's recording surface can cause these delays. So can inaccurate positioning of the drive's read and write heads, erratic behavior of the drive's electronics, or even loose or defective cables.

Besides retries, your hard drive uses another trick to preserve the data you entrust to it. Much of the space on your drive isn't used to store your data at all. Instead, it contains multiple layers of error correction codes (ECC) that can recreate missing data in case of a read failure. These codes can replace unreadable bits and allow your data to survive many (but not all) mishaps.

Unfortunately, hard drives give no outward indication when they resort to error correction to recover damaged data. As far as you can tell, your drive is performing perfectly-right up until the time the drive failure becomes so severe that even the error corr ection codes don't allow recovery, Windows displays a message like "Data error on drive C:," and your data is suddenly lost.

A drive testing utility such as Windows' ScanDisk tells you a drive is suffering before it dies. When you ask it to perform a thorough scan, ScanDisk will read and write every sector on your hard disk. If an error occurs, ScanDisk offers to copy the sector's data to a previously unused portion of the drive and mark the original sector as bad (to prevent its reuse)

But ScanDisk has serious limitations. For instance, it allows the hard drive to perform retries and error correction during testing. If a sector is failing but still readable thanks to either of these recovery techniques, ScanDisk will not report an error. ScanDisk also doesn't stress the disk's read-and-write circuitry by writing special worst-case data patterns. Once again, a sector, or an entire drive, may be marginal but pass all of ScanDisk's tests.

SpinRite from Gibson Resea rch performs more thorough hard drive tests. This classic diagnostic utility, which has been around for years, is the most demanding drive tester available. It can also recover data after the drive's own retries and error correction have failed.

SpinRite starts by determining dozens of hard drive characteristics, some interesting to you and me (such as the drive's size, geometry and speed) and others of interest only to SpinRite itself (such as the angle transected by each sector). It also quickly tests the computer's CPU, RAM and drive electronics.

Then the fun begins. If possible (some SCSI drives prevent this), SpinRite will disable a drive's retry and error correction features. This allows it to detect errors these features had masked. Next it begins an extensive series of read-and-write tests designed to reveal weaknesses in the drive's recording surfaces or electronics.

SpinRite rigorously tests the hard drive's RLL encoder and decoder. All modern drives i ncorporate a type of data compression known as run length limited (RLL) encoding. A circuit called the RLL encoder compresses data before it is written. When hard disk data is read, another circuit-the RLL decoder-reverses the process.

SpinRite performs five suites of RLL encoder/decoder circuit tests, each tailored to the strengths and weaknesses of a particular brand of circuit. If a sector fails during these tests, SpinRite grabs its data (by re-enabling retries and error correction if necessary) and moves it to a new region of the hard disk. If repeated retries and error correction attempts fail to recover a sector's data, SpinRite resorts to DynaStat analysis. Using this technique, SpinRite repeatedly reads the damaged sector and compares the results. After hundreds of reads the program then uses a knowledge of disk-read failure modes and statistics to predict the missing data's most likely values. SpinRite then relocates the rebuilt data to a reliable region of the disk.

SpinRite does have a shortcoming, however. Its current version can't test drive partitions formatted with the new FAT32 file system. This file system, which first appeared in Windows 95 SR2, is required when creating partitions larger than 2GB. And it is often used with smaller partitions to keep cluster sizes as low as possible. Fortunately, Gibson Research says that when Windows 98 comes out, a new version of SpinRite will be released that supports FAT32.

In the meantime, if you have FAT32 partitions, you'll have to use ScanDisk or another third-party disk diagnostic utility, such as Norton Utilities.

File System

E ven if your drives work perfectly, you're not immune to disk headaches. Buggy software, viruses and power failures can cause your hard disk to lose data. These problems can damage a volume's file system, the normally hidden portion of a hard disk where the operating system tracks files, directories and free space.

The original DOS file system-still used by Windows 95-consists of the directory hierarchy and the file allocation table (FAT). The directory hierarchy holds information about each of a volume's files, including name, size, creation date and last modification date. If a file's entry is accidentally overwritten, several calamities can occur. At worst, the file may be deleted or its name may change; then programs won't be able to find it. The file's size might change, causing data to vanish; or the file's dates might change, confusing backup and installation programs that rely on this information.

The directory listing for each file contains a pointer to a location in the FAT, where the operating system ke eps track of disk space. The FAT indicates which file owns which regions of the disk, and which regions are free and available for assignment to new or growing files. As the acronym implies, the FAT is a large table, with one entry for each of the volume's clusters (also known as allocation units). Bad things can happen when the FAT is corrupted: A file may be truncated, free space may disappear, two files may be assigned the same disk sectors, or a defective drive region previously marked bad may be inadvertently returned to service.

And, file system problems tend to spread like the flu. One corrupt entry can confuse the operating system, leading to several additional errors as the file system is updated. In no time, one isolated glitch can grow into an epidemic of data corruption and loss.

That's why it's important to detect and correct file system problems before they become disasters. Fortunately, several utility programs simplify this process. You can run Windows' ScanDi sk and Norton Disk Doctor (part of Norton Utilities) daily, or immediately after system crashes or power failures, to keep file system problems from running rampant.

Both programs check the logical consistency of a hard disk's partition table (where information about volumes and driver letters is stored), boot sector (where information needed to boot Windows or DOS is stored), FAT and directory structure. Both programs can make repairs if they find errors, though some data may be lost.

ScanDisk and Norton Disk Doctor can quickly test the drive's recording surfaces and drive electronics, by reading and writing each of the hard disk's sectors. While less thorough than SpinRite's tests, these tests can spot serious damage.

The Norton Utilities suite includes other tools to ease the pain of disk problems. One stores a copy of a hard drive's boot record, FAT and root directory information in a special file named IMAGE.DAT. Later, you can use this file to repair these items if they become damaged. Norton Utilities' UnErase Wizard makes it easy to restore accidentally deleted files.

Video

S ome of us spend more time looking at our computer's display than we spend looking at our immediate family. Maybe we'll take a few days to notice that new moustache, hair color or nose ring. But we'll spot a dead pixel, color smearing or image jitter immediately.

Determining a video problem's cause is another matter. It takes some of your PC's most complex hardware and software to put an image on screen. When a component fails, it's often unclear where the failure has occurred.

Third-party diagnostic programs such as AMIDiag or TouchStone Software's CheckIt Professional Edition can help. First, they test your video adapter card's RAM. This RAM stores the image currently being displayed and may also cache other video information. Diagnostic utilities also display a series of test patterns on the screen. If all patterns display correctly, it's likely your video adapter card, video cable and monitor are working correctly.

But if any of the test patterns don't appear normal, it's time to play PC detective. If images shrink or expand as the displayed image changes, you probably have an inadequate or failing monitor power supply. Unchanging or mis-colored pixels or regions usually indicate problems with the video adapter card's RAM. If the entire screen is off-color, it may reveal a malfunctioning digital-to-analog converter (DAC). These circuits convert the digital information found in the video adapter card's RAM into the analog signals that control the monitor's electron guns. Or it may simply mean you should adjust the monitor's contrast, brightness or color balance controls.

A loose or defective cable connecting the monitor to the card may cause images that swim or wiggle. Faulty synchronization circuitry on the video card or in the monitor may create a jittery image that jumps or flickers.

Misbehaving video hardware can cause system hangs, too. If you suspect this, power off your computer, then remove and re-seat your video card and cables. Before reinserting the video card, clean the contact along its edge with a soft pencil eraser. This removes any oxide that might distort the signals between the PC's bus and the card.

To test your theories, temporarily replace your adapter card, monitor or video cable with one borrowed from a working computer. (If you replace the adapter card, you might have to install new drivers.) If your problems go away during the swap, it's time to spring for a new card or cable, or your monitor may need a trip to the shop.

Comm Ports/Modems

T elephone lines weren't designed to carry data. But thanks to the hard work of many ingenious engineers, trillions of bytes of data travel across phone lines daily. The journey is complicated, and those traveling bytes must clear substantial hurdles.

The main roads on this trip are the modem (and cable if it's an external modem) and the serial port circuitry, which includes a Universal Asynchronous Receiver/Transmitter (UART) to help send and receive data.

Much can go awry on this trek. Data can arrive corrupted, or not at all. Or the transmission may take much longer than you expect. Fortunately, diagnostic programs such as CheckIt can help find missing data.

CheckIt starts by asking the serial port's UART to perform its own internal diagnostic. If this test is successful, it proves the computer can communicate with the UART. But it doesn't prove that the UART or the rest of the serial port's circuitry is actually functioning correctly. The only true test of the serial port is data transmission.

That's why CheckIt contains two gizmos known as loop-back plugs. These plugs fit over the serial port connectors on the back of your computer (there's one for 25-pin serial connectors, another for 9-pin connectors). The wiring inside the plugs route outgoing data back to the serial port from which it came, allowing your computer to send data to itself. After you install the plug, CheckIt fully tests the serial port by asking it to send and receive several different types of data at all possible speeds. If the port receives all bytes intact, that port's circuitry is exonerated.

Now it's time to connect a modem to the port and test the modem itself. This test phase starts by sending several common commands to the modem and checking for correct responses. Next, the modem is ordered to dial a special modem bank operated by TouchStone just for CheckIt users. These modems carry out a predefined dialog with your modem, testing its ability to communicate over the phone lines at various speeds and under different conditions. If this final test phase is successful, your serial port and modem hardware are functioning correctly. And you'll know that any remaining problems probably belong to the phone company or the person operating the computer at the other end of the line.

CheckIt also includes a loop-back plug that attaches to parallel port connectors so you can test your parallel port circuits. It can also test the motherboard's direct memory access (DMA) circuits. These circuits allow some parallel ports and other adapter cards to copy data directly to and from the computer's RAM without CPU interven tion.

Now you know the basics of how the box on or under your desk works. Armed with this knowledge and the simple tools detailed here, you can keep that new model humming along or save the old clunker from the computer junkyard.