Guide: Performance test your graphics card

The daily SweClockers visitor probably has a good grasp of the editorial board’s graphics card tests and the software used in these. For the more inexperienced, however, it can be difficult to know what each software does, and why we have chosen these particular game titles and respective test loops. For that reason, we go through some tips and aids for how to test the performance of your graphics card.

What you should first examine is of course the system itself and make sure that everything works as it should and rolls in its intended frequencies. For the editorial system, the processor has been overclocked to 4.7 GHz, while the primary memory uses Intel’s XMP profile with slightly tighter latencies. Your system may vary from that of the editorial staff, but the important thing is that the system is stable and that no part of it acts as a bottleneck.

A quick explanation of the term bottleneck is when a component, such as the processor, does not keep up at the same pace as the rest of the system and therefore reduces the frame rate in games. When it comes to games, the graphics card should be the bottleneck, that is, the load on the card is at, or close to, 100 percent. The higher the resolution, the more the graphics card has to struggle and a quick example is to set the game’s resolution to 1,280 × 720 pixels (720p) and see when the load is transferred to the processor.

Software

Before it is time to start a performance test, a couple of software is needed to, among other things, load the graphics cards, but also to be able to capture the frame rate in different game titles. Most of the above programs are usually used by the editors in tests and reviews. Most are free to use, or are available as a free version with some restrictions regarding function. A strong recommendation is to turn off as many programs as possible, such as Steam, Epic Games Store or similar that do not need to be running during the test.

Synthetic test software such as 3dmark and Superposition is a perfect place to start if you have never tested performance graphics cards before. These are fully automatic, which means that the user only needs to press a button, let the program run and then note down the end points presented. The points can then be used to compare one’s system against other users in, for example, SweClocker’s forums or one of the many performance databases that exist.

Ocat.png

If you want to take the next step and instead measure the performance in actual game titles, you usually need additional software to intercept the frame rate. Here, the editorial staff uses OCAT, which can be downloaded for free from Github. We use the older version 1.1 which is the last stripped edition before a lot of extra functions were added that are not needed for our purposes.

Settings that should be reviewed before the performance measurements are for how long the measurement itself should take place and which key starts the measurement itself. Since many games use the keys F1 to F12 for, for example, screenshots and save functions, we recommend changing the key for measurement start to something else. The editors usually use the “Å” key for this purpose.

Last but not least are drivers that can be downloaded directly from AMD or Nvidia’s website. For performance tests, it is recommended to always have the latest version installed. This is where the Display Driver Uninstaller (DDU) comes in, where the software can remove all previous drivers so there will be no conflicts with the latest version. It is also recommended that the software run with Windows in safe mode, which can be set via the settings menu in DDU.

Game titles and test loops

This part is of course dependent on your game library, but there are also some demo versions that can be used for lack of other things. It is important to keep in mind to use a game title that is relatively new, or that can still put a lot of pressure on the graphics processor. Another point to try is to use games found in reviews to compare your system with others.

Read This Now:   Nvidia Geforce GTX Titan X

Regarding the latter point, however, it is important to understand that direct comparisons between your own measurement results and reviews on SweClockers and other hardware sites can be misleading if you do not use the exact same hardware configuration, test loop, resolution and detail settings. Achieving an identical test environment is often difficult, and therefore that type of comparison should be used more as an extremely rough estimate of performance.


Built-in test loops are fortunately becoming more and more common again after several years of absence, and are often the obvious choice to use in performance tests. Should not exist, it will be to find a suitable loop in the game, where this can often take some time. The editors’ test loops last for 55 seconds, but can just as easily be a full minute if the loop allows it.

GPU-load.jpg

It is important to keep in mind there is to choose a loop that can be recreated almost exactly every round, and preferably without a lot of preparatory work because it will probably include a couple of rounds. Here, MSI Afterburner fits very well as the user can log the graphics card load in the background and ensure that the loop is heavy enough.

Results from the performance measurements

Once the results are in, these can be found in the “Results” directory located in the “OCAT” folder, which in turn is found in the Documents section of a Windows environment. Here, each individual round is saved in a separate file and in addition, all the results are collected in a summarized file, which is the one we are going to use. Here, all games are presented with their file name, which can sometimes be a little difficult to keep track of.

Each line provides a lot of information and what is of interest now are the first and last digits. The first part is the mean value that can be rounded to the nearest integer, or take the exact numbers if you really want to be accurate. To be completely sure that the measurements are correct, several measurement rounds are recommended one after the other and a reconciliation of any differences between the runs.

Read This Now:   Gigabyte makes Mini ITX of Geforce GTX 1070

The last part of the number line is the rendering time in milliseconds for the 99th percentile frames, which means that 99 percent of the measured frame rate is higher than this value. This value will represent our “minimum value”. Since what is presented is not our final “minimum frames per second” value, we can take 1,000 and divide this by the value presented in the log file.

When all the numbers are in, these should be saved somewhere, where the editors’ first choice now is Google Sheets for the simple reason that more people can work in the document at the same time. Should it not fit, Microsoft Excel works just as well to keep track of performance figures in a smooth and easy-to-understand way.

Summary

When everything is completed and the numbers are in, it is up to each one what to do with the results. Either the purpose is just to check that the system is rolling as intended, or to actually present the results in a video or some article. In any case, this is a good start on the road, but there are plenty of methods for performance testing components. The editors know that SweClocker’s members have a lot of knowledge in that area and are happy to come up with their own tips and experiences.

Do you test the performance of your system in a different way than those described above? Share in the comments!


Notice: ob_end_flush(): failed to send buffer of zlib output compression (1) in /home/gamefeve/bitcoinminershashrate.com/wp-includes/functions.php on line 5373

Notice: ob_end_flush(): failed to send buffer of zlib output compression (1) in /home/gamefeve/bitcoinminershashrate.com/wp-includes/functions.php on line 5373