The initial analysis: Detect PC problems at an early stage and get justice in the event of a dispute | Practice

Although my topic has been preached by the old hands in the PC field in forums for decades and is often the approach of any solution to performance problems, the same questions keep popping up from less experienced users. I hope to reach them with this article on our professional platform and maybe there is one or the other info for everyone else. Because it is about the analysis that must be done first to notice a problem with the computer and also to identify it clearly.

Only if you know what causes the problem, you can solve it specifically. And PCs in particular have become so complex that you can’t get very far without a structure. Although there are some particularly tricky cases that even experts can rack their brains over, in this article I would like to illustrate the basics, which in the vast majority of cases are already the solution. Many problems are not as complicated to solve as they first seem, grasping them with the right tools and experience or guidance is the real hurdle.

Of course, it is best not to have any problems with your hardware, but in order to recognize them in case of doubt, you should regularly query some performance data and run a purposeful monitoring. Of course, it is also important to understand the data collected in the process, whereby a rough classification is sufficient and you can get help from the expert again for more persistent problems. But the measured values will also help him.

Documentation, witnesses, input measurement and problems

If one buys e.g. a new PC or single parts, no matter whether new or used, I recommend, after an unpacking under witnesses and a photo documentation, always to make also under witnesses an input measurement, in order to check whether the hardware runs at all in the context of the standard configuration. Even if you are already planning to undervolt or overclock your chips and thus want to make the original consumption, performance and temperature values obsolete anyway. The initial measurement is always good for orientation and may already reveal problems that you would otherwise not notice until later and would then also have suspected them in the wrong place.

A customer of mine was experiencing severe fluctuations in FPS in his games and as a result he swapped his very fast, 3 week old GPU for the very latest model, which cost him a relatively large amount of money but did nothing. The issue remained until he asked me to look at the machine and I quickly noticed that his RAM was only running at 2133 MHz, instead of 3800 MHz as it could have. The system integrator from whom he had bought the complete PC had simply not loaded the XMP profile and the customer had relied on the advertised service. So the GPU was never the problem, because he had been in the RAM or CPU limit.

Another replaced all his case fans and CPU because his system was too noisy and not fast enough for him. With the actually good, but not outstanding case fans and the sufficiently fast CPU, this seemed strange to me according to his description, and so I also took a look at his system afterwards. In the process, I found out that there was a mounting problem with the GPU cooler from the factory and his card reached 116°C on the HotSpot and 94°C as the average temperature within 2 seconds. The card always ran in the temperature limit and the card’s telemetry had a hard time squeezing clock, voltage and power consumption into the temperature window. The card only achieved about 40% of the expected performance, and it was loud at that. Again, a system integrator had not done a comprehensive final inspection prior to delivery.

We exchanged his card and unfortunately received a used or refurbished card for the new one. He wouldn’t have noticed, but please take a close look not only at the readings, but also at the hardware itself before installation. Preferably with witnesses and photos of the component, so that you can get justice in case of damage. There is nothing to be said against a refurbished card for the time being, as long as it is equivalent or better than the replaced card.

This was new, but was already considered used by the dealer due to its few days of use until the problem was detected. The card we received was unfortunately damaged. Several bolt heads were visibly shot through and there were paint defects, which would have clearly resulted in loss of value at resale. The seals were still on the press-on cross, but they are easy to forge or renew, so they are not really a guarantee that the card has never been opened. Practically no one wants to have such a tinkered card. The performance of this card matched the benchmark, but it still went back because its visual condition did not inspire much confidence. Only because I insisted, there was then a third card and it was also new.

I then satisfied the mistrust I had towards the dealer by submitting the serial number to the manufacturer. For most, these numbers are cryptic, but they actually contain quite a bit of important info. These can help in classifying what you have received.

 

Comparison with similar systems from reputable sources

I then verify proper operation by looking at performance results from reputable sources as a reference and then comparing them to my own tests. For example, Igor’s Lab, ComputerBase, PCGH, Gamers Nexus, or Hardwareluxx have many measurement results on points / scores, power consumption, and temperatures of the hardware that is being tested in various benchmarks. Different drivers, cases, coolers and room temperatures naturally cause fluctuations, so every source and also your test will differ slightly. Normality, however, turns out.

For example, if 5 sources state a maximum temperature of 76-78°C and your test produces 91°C, this may already indicate that something is wrong because the deviation is too large. Or if a clock of about 4.8 GHz is reached in all under load and you only get 4.1 GHz and significantly less points with the same model, then you should start looking for the cause.

Boost behavior and clock rates can vary greatly today depending on the motherboard and cooling, which makes testing more complex. So provide your hardware with the best possible conditions during testing, such as a clean case with good air flow and fresh thermal paste. However, because the conditions can be so different, the manufacturers also only formulate with “up to” instead of a guaranteed specification. It can still be found in the base clock, but it doesn’t really matter anymore when the system is running correctly.

Suitable tools and the right behavior in the event of damage

Tools that help you with the calibration are monitoring programs such as HWInfo, HWMonitor, FPS-Monitor, CPU-Z and GPU-z. With them you can already learn a lot about how your system actually runs and whether everything is okay. You can then do the load, i.e. the benchmark, with CineBench, OCCT, 3D-Mark and many games, some of which are free and either offer an elaborate scene or have an integrated benchmark themselves.

For example, here is a screenshot from FPS Monitor with my configuration of the overlay. All values relevant for me are displayed there.

With new systems or components, check the values first and then use the system with a clear conscience if everything fits. That is, whether the CPU and GPU reach normal and generally known temperatures and performance points or FPS.

If problems occur afterwards, it will be much more difficult for you to find the error or to get right with the seller. Also, please refrain from making vague assumptions until you have gone into the analysis. Follow a process of elimination, have an up-to-date bios and operating system, and drivers. It is true that new software also contains new problems, but in general you can increase compatibility by being up-to-date and, with some exceptions, you can be sure that the problem is not with the software.

Occasionally, a look into the TaskManager can also help to check the processes, so that the background load of the system is not excessive. And here is also a video of me testing the graphics card with the contact problem of the cooler or overheating. Actually, 10 seconds is enough to realize that something is wrong here.

I hope my post helps you and I look forward to comments, as well as your experiences. Maybe then we can add to this short reader article step by step 🙂