photo credit: kennymatic
Help diagnosing hardware problems
It's a scenario that everyone dreads: You build a computer, power it on, and then... nothing! Or you install the OS and the system starts acting up before you've even made your first foray onto the Internet. Or you're using your computer and all of a sudden, something just doesn't work right and you're sure you haven't made any major changes. Maybe it's hardware, but how do you tell? This guide will help you figure out whether you truly have a hardware problem, and if so, which piece of hardware is the culprit.
General Tips
Before we get to advice on specific hardware, here are some tips to sharpen your troublesome-hardware hunting skills.
Make sure it's plugged in
It's happened to the best of us. Push the power button and nothing happens. Then you look over to the wall socket and realize that your rig isn't even plugged in. Or the PSU switch is set to the wrong position. If you've been touching the hardware, the first thing you want to do before digging down further is to make sure that:
All cables are plugged in.
All components that are seated in slots (RAM, expansion cards, etc.) are inserted properly.
If there are switches, they are in the correct position.
Try another port
Sometimes, you have a bad port or slot. If a hard drive or SSD isn't working on one SATA port, try another. If a USB device isn't quite working, try another USB port.
Ensure software isn't the problem
Software can be responsible for a lot of issues. Try out a few blanket fixes before diagnosing further. Some examples are:
If you've recently installed an update, try rolling back to an older version.
Ensure the configuration is correct. One misconfigured BIOS setting can cause your system to not boot. One slip of the slider can cause your fans to go full throttle.
If all else fails, uninstall the software tied to the hardware and clean up any remnants, such as if the program left its folder in C:\Program Files, and then reinstall the software.
Keep a PC speaker handy
If you don't have a higher-end motherboard with some kind of readout that display POST errors, a PC Speaker can at least help you troubleshoot what's going on if the computer doesn't BOOT.
This type of speaker normally plugs into the front panel header of the motherboard and is the only way for a computer to tell you something is wrong. If the system doesn't boot up, the computer will do something other than a single short beep. While these beep codes are generic, it at least helps point you in the direction of where a problem might be.
No beeps: There's a power issue. Check to make sure all of the required cables are connected.
One long beep followed by two or three short beeps: The video card has a problem. Make sure it's seated in the slot all the way, and that the PCIe power plugs are installed.
Repeated long beeps: Memory issue. Try reseating the memory.
Any others: Most likely a processor issue, either due to overheating, improper seating, or the processor is defective.
Eliminate variables
When you aren't sure what's causing issues, it's best to start eliminating the variables. One easy variable you can eliminate is the cable. While external cables are often prone to abuse and can fail readily, internal cables aren't designed with repeated connections in mind. If there's a cable coming out of a suspected component, try swapping it first.
One other method of eliminating variables is to do a cross-hardware test. For example, if your Wi-Fi adapter appears to be broken, get another Wi-Fi enabled device in the area and see it can connect and do network-related tasks. This will at least tell you if your Wi-Fi network is working in that area.
For system building, there's the extreme end of this: the bare-bones build. Remove all of the system components except the processor, motherboard, one stick of RAM, and video card. If you have a PC speaker, you can also install just the processor so that you can invoke the POST code beeps. If the system behaves as expected, start adding components until it breaks. The last component you added is most likely the problem.
A Handy Cheat Sheet
The following is a list of symptoms and the most probable hardware that's failing.
USB device or add-on card isn't working properly
Thankfully, this is easy: it's the device in question. Peripherals are often not tied into the system. If they start acting up, they won't take down the system. It should be easy to start trying potential fixes. With some devices, it's easy to know when to throw in the towel, like if a mouse button no longer works. Others may take some further investigation.
Blue screens
Blue screens are often a sign of a hardware problem, so note the error it gives. For example, if the error is one of these:
IRQL_NOT_GREATER_THAN_OR_EQUAL_TO
IRQL_NOT_LESS_THAN_OR_EQUAL_TO
PAGE_FAULT_IN_NONPAGED_AREA
MEMORY_MANAGEMENT
It's most likely RAM that's the problem. Though a bad boot drive (HDD or SSD) can also cause issues like this. Thankfully, there's already an article explaining how to survive and troubleshoot BSODs.
Video artifacts
If the artifacts are random, such as textures becoming discolored, streaks of color going around, in general, it looks like a really bad JPEG in a club, the video card is having trouble. It's most likely a sign it's overheating.
If the artifacts affect the whole screen and are of a consistent pattern, like discoloring every 4 lines, then chances are it's the cable or the display itself.
Video card driver keeps crashing
If trying several versions doesn't work, then the video card is definitely having issues.
Clicking noise from the hard drive
It's a classic sign the hard drive is going to die. Start backing up the data if you can!
Read/write errors that pile up on a drive
This is another sure-fire way to tell that a drive is going to die.
Sudden shutdown/power loss
This could be an indication of two things:
A vital component overheated. This could be the processor, the VRMs (the housekeeping circuitry around the processor socket), or the power supply itself. You can check the processor's temperature with a utility like HWMonitor. The VRM or power supply can be checked by touching the heatsink or case respectively.
The power supply is old/defective and could not handle a high load.
Computer does not power on, period.
Definitely the power supply.
Computer powers on, but does not boot
The computer is failing POST. Refer to the "Keep a PC speaker handy" section above; otherwise, check the power connections and whether video card, RAM, and processor are seated properly.
General instability
Assuming there's no issue with power or temperature, it could be the motherboard, RAM, or processor.
Testing Your Hardware
Power Supply
The best way to test if the power supply is running normally is to stress the computer as much as possible. Higher component loads mean the PSU has to supply more power. You can also purchase a power supply tester, but this only measures the output voltages. While something that's not close to 3.3V, 5V, or 12V is an obvious problem, the power supply can still have issues even if the voltages check out fine.
If the system doesn't turn on, you can check to see if the power supply itself turns on. Turn off the hard switch on the power supply and unplug everything, including the cord to the mains socket. On the 24-pin motherboard connector, stick a paper clip or something similar into the plug on pin 16 to any of the ground pins (see the diagram below). Then plug the power supply back into the mains socket and turn the switch on.
With the plug facing toward you and the notch to the right, pin 16 is the fourth one down on the right side. Connect this to any of the ground connections...
...like so.
If the power supply turns on, it could still be fine. Plug the power supply back in to the computer and turn on the system. If it still doesn't turn on, try taking out the case's power switch on the motherboard and carefully touch the two pins for the switch with metal. If the computer turns on, the case's power switch is bad. Otherwise, the power supply is bad. If the power supply does not turn on, the power supply is bad.
RAM
One of the best ways to diagnose a RAM issue is to run Memtest86. It's a boot-time program, so you'll need a USB thumb stick for this to work. The files you download from the website come with a program that will turn the thumb stick into a bootable program. Once you've created it:
If you bought more than one set of RAM, take out all but one set. If you have two dual-channel kits, for example, take out one of them.
Normally, it's suggested to try one stick of RAM at a time, but if you're going to have to return the RAM, the manufacturer usually wants the whole set back.
Plug the thumb stick with Memtest86 on it into the computer, and power it on.
Boot onto the thumb stick by either using the motherboard's boot menu, or you can cheat by unplugging the main boot disk.
Let Memtest86 run its course for at least several hours. It's recommended to do an overnight run just to be safe.
Memtest86 running tests on RAM
After a Memtest86 run, note how many errors there were. If the RAM is functioning normally, it won't have any more than a handful of errors, if any. If errors start piling up in the thousands, it's definitely a sign of bad RAM. When in doubt, rerun the test.
Keep in mind that you may experience issues when you run Memtest86. Since you're trying to run a program on RAM that's potentially broken, don't be surprised if Memtest86 itself crashes or breaks in other ways. Thankfully, if Memtest86 breaks, it usually does so quickly. If it does break, reboot and try again.
Hard Drive/SSD
To see if your hard drive or SSD is showing signs of imminent doom, you can use a tool to read its SMART data. Some examples of utilities that will do this are:
PassMark's DiskCheckup
CrystalDiskInfo
SpeedFan
These programs may provide you a laundry list of data. Thankfully, Wikipedia has a list of SMART attributes with those highlighted red being the ones to look out for if you suspect the drive is failing.
If the boot drive is giving you trouble, try the following:
Install the hard drive in another computer, either internally or externally, if possible. You can purchase a hard drive dock to make external connections easier. If you do install the boot drive internally, make sure that computer doesn't try to boot from it. Either way, the goal is to get the drive running on another system to run your diagnostic tools of choice.
Use a recovery-oriented Live OS. These OSes are installed on removable media such as a CD/DVD or thumb disk and can be run from there. One example of this Ubuntu Rescue Remix. This also comes with a SMART reader tool called smartmontools. Unfortunately this tool isn't user friendly, so we'll defer telling you how to use it to either this or this tutorial.
Processor/Video Card
We've lumped these two together because the best way to test whether they're working properly is to run a stress test. For the processor, there's Prime95. For video cards, there's FurMark. Run the test for at least a few hours and see if the system or programs start showing signs of issues.
Motherboard
Unfortunately, there's no real way to test the motherboard. Some suggestions are:
Do the bare-bones build, with just the motherboard, processor, memory, and video card, and see if the system boots or can be played with in BIOS.
Swap parts in from another system if possible.
Don't forget: weird things can happen
Keep in mind that all of these are suggestions to help pinpoint what might be the problem. There's no guarantee that system problems are always hardware-related. And as strange as it sounds, hardware can sometimes just act weird. we once had a failing power supply make a hard drive to give the "click of death" sounds, even though its SMART data came back fine. Another time, we got a set of RAM that couldn't wake up the computer from standby, yet it passed multiple Memtest86 runs.
Troubleshooting hardware and finding the exact problem isn't quite a science, but we hope this article helps in your process of learning more about your rig, and keeping your setup tight.
