Tested broken? Of Monday models and night tests

Final tests, headaches and a happy ending. That sums up my test odyssey with the Surface Book 3. But it's not just the fact that I'm testing three versions of the Microsoft Surface Book 3 that's to blame. It's also the fact that two books got into trouble during the test.

In my triple review of the Book 3, there is no mention of the fact that I had to run night tests due to defective devices and therefore actually had not three, but five devices on the test bench. And for good reason.

Anything that can go wrong will go wrong

What happened?

The first device with strange behaviour is a 13.5-inch book with i7 SoC and dedicated nVidia graphics card. After unpacking, the initial setup and a long wait to install Windows updates, I want to sound out the performance of the three test devices right at the start. To do this, I start with the 1080p gaming benchmark Fire Strike from 3DMark. After that, the aforementioned device no longer runs smoothly. I can now see its mouse pointer dozens of times on the screen when I move it. Programmes only start and close after a long wait. Just like when we were using Windows 98. I can't get into the task manager, but I do manage to restart the device. Which takes an unusually long time and does nothing.

Resetting to factory settings without deleting data takes an abnormal five hours and doesn't help either. Only days later do I have to deal with this device again. A second reset, which takes a normal amount of time, seems to have helped. This could be the end of the problem - but the device still needs to be repaired. Welcome to the warranty and a replacement device, which is now allowed to jump in for all tests and then works perfectly.

The second problem device takes much longer to start acting up. When I test gaming-capable notebooks, I save the gaming and the in-game benchmarks until the end. And so I'm surprised when, days after the first problem, I come across another notebook that isn't running smoothly. Diesmal die 15-Zoll-Version. When playing Far Cry 5, I notice that the graphics card does not exceed 30 FPS at ultra quality settings.

In the picture below you can see two results of a 3DMark benchmark in comparison. It is the identical test that was run at different times with the same device. The first result comes directly from testing after the initial setup of the product. The second was taken immediately after the stress test, which keeps the notebook's hardware at maximum performance until it runs out of battery. Or until the graphics card is fried. Note not only the different score of the two measurements, but also the average memory clock frequency and temperature of the graphics card.

Here are the two results:

• https://www.3dmark.com/3dm/48200077?
  - https://www.3dmark.com/3dm/48474617?

Well, 7000 points less. The stress test may be to blame for the fact that the nVidia GeForce GTX 1660 Ti Max-Q only delivers a fraction of the performance. More precisely, the software HeavyLoad and FurMark are to blame in a double team.

Or else, I am to blame. Because there are also warnings. For example, the Chrome browser has not allowed the FurMark software to be downloaded for some time now.

In addition, FurMark itself also warns against the use:

Caution
FurMark is a very GPU intensive application and instability or system shutdowns may occur.
RUN THE BURN-IN TEST AT YOUR OWN RISK
. Possible causes of instability: excessive overclocking settings, insufficient PSU (power supply unit), problems with the VGA cooler (GPU or/and VRM)

Of course, I also reset this book after noticing the defect. Even twice - but the graphics card doesn't get back up to speed. I have no choice but to repeat all the tests with a night test device. Luckily, this one doesn't make a fuss.

When I look through and compare the results that the 15-inch book delivered before it broke, I notice something else: With the broken device, the volume of the fan measured from my sitting position at maximum power is 42 dB. With the night test device it is 52 dB. An indication that the stress test or I am not to blame after all?

Failure to reproduce the error

It can happen that a notebook has a defect. It's not the first time that I've sat frowning in front of a product that no longer runs smoothly. In such situations, the first thing I ask myself is whether there is user error - whether I myself am to blame. If this is not the case, I think about where the test device comes from. Our test products either come from our own warehouse or directly from the manufacturer. If the latter is the case, it could be a pre-sale model that is plagued by teething problems. Or it could be a model that other testers have already held in their hands. In any case, I try to reproduce the error with a product in our warehouse. If I succeed, I'll let you know. Otherwise, I assume that the devices you buy from us will work perfectly.

I only assume that a product has not been damaged by a pre-tester if it has never been in anyone's hands before and it comes from our warehouse. Furthermore, if only a single product is defective, I do not consider this to be an indication of a more common defect. Monday models are also not limited to individual manufacturers.

The reason why I am still writing about the defects is because, in this case, two out of three devices "broke" during testing. Not only is this new, but these are the first defective Microsoft devices I've come across. It's quite possible that none of this means anything. But it is still remarkable enough not to keep quiet.