Troubleshooting and benchmarking a new PC

Features James Hunt Jan 4, 2013

You’ve built your PC. But how do you check it’s working, and what do you do if it’s not?

Once you’ve bought a new PC you’ll probably want to check it’s working properly, but how can you know for sure and what can you do if it isn’t? The answer to the first question is benchmarking; running stress tests and performance ratings on hardware will give you a numerical value that you can compare to an older machine (or a similar system to yours, via online resources) to tell whether your hardware is performing at optimal capacity. If it isn’t, it’s time to answer the second question, and get down to some serious troubleshooting.

Benchmarking and troubleshooting are both complicated, nuanced procedures that can’t really be given a full explanation - the easiest way to learn is by doing. However, we can point you in the right direction and help you get started with the basics. We’ve broken each procedure down by component, so that you can focus on the hardware most important to you.

CPU benchmarks and troubleshooting

The CPU is often the pride and joy of any new PC, but it’s also one of the most inscrutable components. How can you be sure it’s running as fast as it’s supposed to, and doing so without errors? The answer, of course, is benchmarking, which uses processor-heavy tasks to stress the hardware and check the accuracy of its work.

Tests which are designed to be processor-centric ensure that the speed and quality of other components in your system don’t factor in, to give you the true speed of your hardware. This can give you values that are even more accurate than the box. If that sounds surprising, it’s down to the way CPUs are made. Hardware is made in batches, which are tested for performance, and their quality can vary based on the materials and production process. For this reason, you can expect a certain level of differentiation between the rated speeds and the actual speeds, and benchmarking can help you find out whether your processor is dead on the speed it should be, or somewhere above or below it.

Perhaps the best program to test your CPU with is SiSoftware Sandra 2013 (available at www.sisoftware.co.uk), which is a free benchmarking application with a huge number of tests that can be applied to the CPU. They’re ‘synthetic’ benchmarks, which means that they’re designed to test the hardware’s capabilities and maximum performance, rather than resemble real-world usage conditions. The multi-threaded tests should give higher scores for processors with faster clocks and more CPUs, and you can easily compare the results for your hardware against other processors using the built-in tools.

If the results show that your processor is underperforming, though, it’s time to start troubleshooting. To begin with, check how much the processor is underperforming by. If it’s only slow by a few percentage points of the expected 100% value, then you may simply be out of luck - processors are normally rated conservatively to ensure this doesn’t happen, but manufacturing issues and stock shortages may result in slightly slower processors going out. You can try and swap it, but there’s no guarantee you’ll get a better one in return.

If the problem is that the processor is performing very badly (but remains stable) then there are a couple of potential problems. First, check in your PC’s BIOS settings to make sure it hasn’t accidentally been underclocked. If it has, reset the values so that it runs at full speed, reboot, and run the benchmark again. If that was the problem, it’ll now be running as fast as you expected. If not, it may be a genuine hardware fault or even a case of mistaken identity - a Core i3 accidentally badged as Core i5, for example. Get in touch with your supplier for a replacement.

Finally, if the benchmarks are causing the computer to crash, you’ve either got a hardware fault or a cooling issue. When CPUs are stressed, they get hot - and if they get too hot, they start to fail. The CPU’s own fan should be enough to prevent the temperature rising to damaging levels (that’s why overclocked CPUs need better cooling systems), but if it isn’t you need to check why. If you’re keeping the computer in a cupboard or the air vents are blocked, that could be the reason. It could also be that the cooling isn’t working properly. If the former is true, simply move the case to a more ventilated location, but if the cooling is broken, you should ensure that it gets replaced. If someone else built it, insist they provide a new CPU as well, in case the lack of cooling has already damaged your hardware!

RAM benchmarks and troubleshooting

Although CPU performance provides the most general measure of a machine’s performance, the speed of the memory can also become a major bottleneck, especially in high-performance gaming machines. So, before doing anything on your system, you should immediately test your new RAM modules for errors. RAM (especially budget RAM) is prone to manufacturing errors which can cause all manner of problems, including unexplained crashes and freezes. There’s often no way to diagnose these when they’re happening, so ensuring that the RAM is working is the only way to be certain it’s not the cause.

Use a free program like the excellent Memtest86+ (available at www.memtest.org) to test your RAM as soon as you’re able to. The software is easy to run, and you can even download a version that auto-installs to a bootable USB stick for complete simplicity of use. If the tests uncover any errors so soon after your purchase, you’ll be entitled to a replacement from your supplier because they’ve sent you faulty goods. If they refuse, you can at least contact the manufacturer under the product’s warranty conditions. If there are no errors, you can be confident that the RAM is working, and proceed with benchmarking.

Memory benchmarks use synthetic tests to determine the bandwidth and latency of your RAM modules and banks. The best performance is given by RAM modules with a high frequency, although the number of channels and capacity of modules also plays a part.

If you’re willing to pay for benchmarking software, AIDA64 Extreme Edition can test read, write and copy bandwidth, and latency timings. Again, though, SiSoftware’s Sandra 2013 is capable of running tests that give you statistics in real-world terms - Bandwidth in gigabytes per second, latency in nanoseconds. Higher is good for the first, lower is good for the second, and faster clock speeds on your RAM should improve the former while lowering the latter. Again, if you run into problems, you need to do some troubleshooting.

If the amount of RAM appears to be wrong, you need to check two things: first, that you’re running a 64-bit operating system (remember, 32-bit systems can’t recognise more than 4GB) and secondly, that the RAM is actually present and seated correctly in its slots.

If you’re getting crashes during tests, it may again be the result of excessive heat inside the case (although RAM doesn’t typically require active cooling, so it would have to be quite excessive) or the result of accidentally overclocked RAM. Check the speed rating (given in MHz) and make sure your BIOS doesn’t list a higher value for the RAM modules. If the RAM is operating slower than you’d expect, as with CPUs it may be underclocked, or simply performing within an acceptable margin of error.

You can use benchmarks to determine the best configuration for your RAM modules, too. Generally, you get the fastest speeds by placing the best/biggest modules in the first bank, but feel free to swap them around and benchmark other configurations to see what works best. Just remember not to accidentally zap them while you’re moving them around between reboots!

Storage benchmarks and troubleshooting

It might sound unlikely, but the biggest bottleneck of any modern PC is quite likely to be the hard drive. Processors are getting quicker and quicker, but hard drives are getting bigger and bigger. Games and applications are becoming huge, and hard drives are expected to shift more data in and out of RAM than ever before at quicker speeds. Most of the time, they simply can’t keep up, as you’ll be aware if you’ve ever sat in front of a PC waiting for a churning disk to stop so that you can continue working.

Although SiSoftware’s Sandra 2013 does have a disk benchmarking component, the best tool for the job is probably CrystalDiskMark (available at crystalmark.info), which is specifically designed to test the read and write performance of storage (mechanical and solid state) at a variety of different transfer sizes. Tests will involve sequential and random access (in both directions). If you prefer, you can also use a program such as PC Mark to perform a ‘trace’ test, which examines read and write speeds under normal usage conditions over a longer period of time (and reports latency, which CrystalDiskMark doesn’t).

As with memory, you should also check your storage hardware for errors at the earliest opportunity - you don’t want to discover that a hard drive is in danger of failing when you’ve already filled it with your stuff, after all! Most drive manufacturers offer a free proprietary testing tool that you can use, although the built-in Windows disk checker is a reasonable alternative. If you find any bad sectors, you are entitled to get a replacement.

If your benchmark suggests that hard drive access is slower than it should be and there are no bad sectors, it’s likely that you have a configuration option wrong somewhere, so look in the BIOS and check that the drive has been correctly detected. It’s also possible that your OS has, for some reason, installed conflicting drivers for the device, so look in Device Manager and check that the IDE/ATA controllers don’t show any problems.

Alternatively, if the slow drive is a secondary storage device, check the power management settings - it may be that the hard drive has been told to power down when not in use, and the slow access times are the result of a wait while the drive powers up again.

If you have an SSD, slow access times can be caused by chipset incompatibilities, so if possible, try connecting your SSD to another motherboard (either an old one, or someone else’s) and seeing if the problem persists. It’s also possible that you’ll encounter write speed problems after cloning an existing mechanical drive to an SSD. If you did this, the best course of action is (somewhat unhelpfully) to use a program like Partition Magic to do a secure erase and then start again. SSDs should be significantly faster than HDDs, so any slowness is a definite sign of problems.

Graphics benchmarks and troubleshooting

GPU performance is of primary interest to gamers, but even if you’re running an onboard GPU only, you might be interested to see how it performs against actual graphics cards in case you want to upgrade to something better. There are more ways to test the graphics capabilities of a computer than almost anything else, and it is well worth testing for peace of mind. Indeed, many games even provide their own benchmarking, and Micro Mart regularly use Crysis 2 and DiRT 3 for just such purposes.

If you’ve read this article in order, you’ll know the drill by now: failures during stress tests may be the result of inadequate cooling, failures under moderate use can indicate software or hardware faults, and under-performance likely indicates a configuration mistake. However, rather than to test hardware, graphics card benchmarking is more popularly used to fine-tune it. The kind of people who buy graphics cards at all are normally interested in getting the best performance out of them, and benchmarking tools are the best way to see whether tweaks have had a positive, negative, or indeed any effect.

For testing real-world performance, we recommend Fraps (available at www.fraps.com) - an unusual gaming benchmark tool which runs in the background while you play games, then uses your PC’s actual performance to analyse its capabilities. By taking notes about how many frames your computer renders while you play a game, Fraps can give you an accurate picture of how your PC is actually performing, rather than putting it through a series of test designed to mimic gaming performance. The program has several other features and a paid-for version, but the benchmarking is available in the free incarnation.

Alternatively, try 3D Mark, which is perhaps the most popular benchmarking tool around. As the name suggests, it primarily analyses 3D performance, so if you’re a gamer who likes to tinker with settings, it can give you a good metric by which to quickly compare your PC pre- and post-tweak. There are several versions of the software which you should choose based on your graphics hardware’s DirectX capabilities - 3DMark11 for DX11 cards, 3DMark Vantage for DX10 cards, and 3DMark06 for DX9 cards. You’ll be given a separate score for both CPU and GPU, and in both cases, higher is better.

That, more or less, is how you check your new PC is in working order. Remember that, whatever tool you use, you should keep a note of the results you got out of it - a solid record of them will allow you to refer back to them in the future, and could potentially help you to recognise failing hardware long before anything goes seriously wrong. It’s also a good idea to run them just before an upgrade, as well as after - if only so you can see how much of an improvement the new hardware has made!