in

Metro Exodus: The community is testing, here is the guide (update)

Metro Exodus: The community is testing, here is the guide (update)

tl; dr: How fast does Metro Exodus run on current systems? ComputerBase has been providing answers to this question for days – but not for any current GPU and CPU, and certainly not for older systems. To bridge this gap, ComputerBase dares an experiment: Readers test Metro Exodus for guidance.

Readers benchmark Metro Exodus

Update 21.02.2019 12:04 clock

Sometimes it's the time, sometimes the copy protection, and sometimes both factors that prevent the editor's large-scale technical tests on current games from mapping many, but far from enough, system configurations. The answer to many questions is à la “How fast does the game run on computer XY? almost on the hand or in the community.

Using the example of Metro Exodus (test) ComputerBase now dares the experiment: Can ComputerBase readers test a current title according to specifications of the editors and thus a maximum of comparable and thus make a much more comprehensive technology test?

Metro Exodus is a game that accepts savegames from outside sources for this premiere. In addition, the title does not have any graphic presets, but the number of options is manageable. What else is there to note, the editors explained in the following sections.

It's not about the longest bar

The purpose of this action is not to find the community member with the longest bar, but to get a broad view of the performance of various systems in Metro Exodus. However, this only succeeds if all participants adhere to the rules and create the benchmarks to the best of their knowledge and belief.

These are the system requirements

For this first attempt one of the most important variables is excluded: Windows. For this reason, only results created with Windows 10 version 1809 (October 2018 Update) should be submitted.

Windows 10 without background load

Users should also make sure that no relevant load runs in the background at the start of the benchmark. This should be checked via the Task Manager. In addition to classics such as virus scanners, Windows updates or the download / installation of game updates via Steam and Co. can also create a heavy burden on the CPU in the background. It is best to close all applications that are not needed at the moment.

As a graphics card driver, the appropriate game-ready driver from AMD from Adrenalin 19.2.2 or Nvidia at least the GeForce 418.91 to install. In case of uncertainty, it is recommended to use the latest Radeon or GeForce driver. The drivers can be downloaded from ComputerBase for both AMD Radeon and Nvidia GeForce.

Must be turned off optimizations in the driver or GeForce Experience, downsampling or upsampling must be turned off.

These are the settings for Metro Exodus

Readers need to test Metro Exodus in the settings where the editors have also tested the game. This allows better classification of submitted results. Raytracing and DLSS are eliminated because these features could only be tested by a few readers on four selected GPUs.

The resolutions can be 1,920 × 1,080, 2,560 × 1,440 and / or 3,840 × 2,160. Up- or downsampling should be avoided, because it slightly falsifies the result. All details can be found in the following table or screenshots.

This is the test sequence

If you have never started Metro Exodus before, you should first start the game before the actual benchmark, load the test sequence, wait a few seconds, go back to the main menu and reload the savegame. This is how the so-called shader cache is filled in order to achieve the best possible result. If you want to play it safe, you should run the entire test sequence without a benchmark, reload the savegame and then record the actual benchmark.

The test is based on this savegame (download). After downloading the ZIP archive is to unpack and the content in the folder (two files) C: User “NAME” Saved Games metro exodus “CODE” to copy. Any own scores should be saved before. In the game, the ComputerBase savegame will be called up via “Continue”.

Starting from the memory point, it is important to follow the course going (not running!), Just as you can see in the following video. During this 25-second sequence, the FPS are determined. The measurement starts with the first step. Like, explains the following section.

The Savegame system of Metro Exodus does not seem to be absolutely robust and places the player a few seconds away from the benchmark sequence despite the right savegame, namely at the end of it. This is the editors noticed in further own tests.

The editors therefore ask meticulously to pay attention to whether participants really start in the same place as in the video. If you start at the (always in the same way) wrong place, you go down the ladder again, turn 180 degrees and then arrive after a few seconds in the correct place. Then you can orient yourself on the video where the starting point lies.

Determine FPS with OCAT

As a benchmark tool, only OCAT should be used for optimal comparability. The formerly popular FRAPS no longer works in all games and is limited in functionality, which is why it is omitted. OCAT can be downloaded from ComputerBase.

After installation, the program can be found either in the Windows start menu or the executable exe file in the directory “C: Program Files (x86) OCAT” – a desktop icon is unfortunately not created. In OCAT itself, the graphical overlay has to be switched off by clicking on “Stop overlay”, because it can falsify the measurements. In addition, the time for the benchmarks under the point “Capture time in seconds” must be set to 25 (for 25 seconds) in the “Capture” tab. The actual test is then started in the game by pressing the “F12” key. This is done in the sequence shown above parallel to the first step along the corridor. An optical feedback, when the 25 seconds have elapsed, there is no overlay.


OCAT: The overlay must be disabled
OCAT: The overlay must be disabled

OCAT: Measured by pressing F12 for 25 seconds
OCAT: Measured by pressing F12 for 25 seconds

OCAT: Measurement results are under document  OCAT
OCAT: Measurement results are under document OCAT

OCAT: The test results in the CSV file
OCAT: The test results in the CSV file

The only important measurement for the reader benchmarks are the average frames per second (FPS). There are two ways to evaluate the results.

On the drive of the Windows installation you will find under “User / Username / Documents / OCAT / captures” the measured values ​​in detailed form (OCAT filename date) for each pass and in summarized form (perf_summary). The latter is the required file and there are all the results, if several runs have been made. For example, “perf_summary” can be opened with Excel or the free LibreOffice. In order to get an overview into the file, the “comma” must be selected as the separation option of the columns. The column “Average FPS (Application)” contains the required result.

Who on the other hand would like to have a visually prettier and more detailed evaluation, the Tool CapFrameX is recommended to the heart. If you use the default installation paths of OCAT, you will automatically see one or more benchmark results (depending on how many have been created) in the program on the left-hand side of the program, and you can switch between them by clicking on them. Various evaluations are offered, such as maximum, average and minimum FPS, but also numerous frametime results as percentiles. There are no decimal places here. CapFrameX can also be downloaded from ComputerBase.


CapFrameX graphically evaluates OCAT measurements
CapFrameX graphically evaluates OCAT measurements

CapFrameX graphically evaluates OCAT measurements
CapFrameX graphically evaluates OCAT measurements

CapFrameX graphically evaluates OCAT measurements
CapFrameX graphically evaluates OCAT measurements

This is how benchmarks are reported

Once the benchmarks are done, the results in the thread for that article can be posted. It is important to name the processor and the graphics card (manufacturer, type and model). For verification, the version of Windows 10 and the video card driver should be mentioned again. Anyone who has overclocked CPU, GPU or both can like to mention the exact frequency in the post, which helps in the classification. In the diagram, however, only the name suffix “OC” is listed for reasons of clarity.

Should this experiment catch on, another system for submitting results is conceivable in the future. For this first attempt, the message about the forum should be sufficient.

Here we go

Again and again ComputerBase tests current games and applications on older hardware, but that is not the rule. Although the editors in the archive have an almost complete history with the GPUs and also with the CPUs the essential types still in the access, establishment and enterprise of each further (older) platform are connected with an enormous expenditure of time – and copy protection mechanisms like Denuvo, the Allowing only five computer configurations per game key in 24 hours is another hurdle.

ComputerBase readers usually have exactly one gaming PC at home, which is already set up. What could be more obvious than to assemble many small contributions from individual interested readers into one big overall picture? At least in theory, the approach is promising. Now he has to prove himself in practice.

Depending on the game, this will be more and less successful, some titles are not even suitable with cloud savegames tied to the account. Also, it will take a while for the process to take effect or this guide really clarifies all the essential issues here. Feedback is always welcome.

Due to the large and enthusiastic community, however, the editorial team sees a unique opportunity to successfully use the multitude of different systems and thus to make interesting discoveries and have fun while sharing information. Happy benchmarking!

The benchmark results of the readers

During the first day, many readers took part in the community benchmark and posted results on numerous different machines. In addition, there was a lot of constructive feedback that the editorial team has already partially implemented, but will primarily implement in the next community benchmark. For this, ComputerBase would like to sincerely thank you!

Reader benchmarks in Metro Exodus – 1,920 × 1,080, DirectX 12

    • i7-8086K OC, RTX 2080 Ti FE OC

    • i9-9900K, GB RTX 2080 Ti Gaming

    • i7-8700K OC, RTX 2080 Ti OC

    • i7-8086K OC, RTX 2080 Ti FE

    • CB: i7-8700K OC, RTX 2080 Ti FE

    • i7 9700K, RTX 2080 FE

    • CB: i7-8700K OC, RTX 2080 FE

    • CB: i7-8700K OC, GTX 1080 Ti FE

    • 7 1700 OC, Palit RTX 2080 Dual OC

    • CB: i7-8700K OC, Radeon VII

    • i7-8700K OC, EVGA GTX 1080 FTW OC

    • i7-8700K OC, GTX 1080 OC

    • 7 1700X OC, RX Vega 64 OC

    • i7-4930K OC, RX Vega 64 OC

    • i9 7940X OC, Palit RTX 2080 TI

    • CB: i7-8700K OC, RTX 2070 Ref

    • 7 1700 OC, GB RTX 2070 WindForce

    • CB: i7-8700K OC, RX Vega 64 Ref

    • CB: i7-8700K OC, GTX 1080 FE

    • 5 1600X, Sapphire RX Vega 56 OC

    • 7 1700X, RX Vega 64

    • CB: i7-8700K OC, RTX 2060 FE

    • 7 2700X, MSI GTX 1070 Ti Gaming

    • CB: i7-8700K OC, RX Vega 56

    • 5 2600 OC, Asus GTX 1070 Dual OC

    • CB: i7-8700K OC, GTX 1070 FE

    • i5-4670k, GTX 980 Ti FE

    • CB: i7-8700K OC, RX 590 Ref

    • 5 2600, RX 580 OC

    • TR 1920X, Asus GTX 1060 9 Gbps

    • CB: i7-8700K OC, RX 580 Ref

    • CB: i7-8700K OC, GTX 1060 FE

    • 5 1600 PC, RX 480 OC

    • Xeon 1230v3 OC, EVGA 970 FTW OC

While most of the benchmark results fit surprisingly well with the editor's own readings (marked by the acronym “CB:” in the charts), there were always results that were too good. Thus, some readers have apparently been confused that “Ultra” in Metro Exodus despite the name is just not yet the maximum level of detail. After “Ultra” follows in the game “extreme” – one may discuss whether this makes sense.

Reader benchmarks in Metro Exodus – 2,560 × 1,440, DirectX 12

    • i7-8086K OC, RTX 2080 Ti FE OC

    • i7-8086K OC, RTX 2080 Ti FE

    • i9-9900K, MSI RTX 2080 Ti Gaming X

    • CB: i7-8700K OC, RTX 2080 Ti FE

    • i7-8700K OC, RTX 2080 Ti OC

    • i9-9900K, Palit RTX 2080 Ti Gaming

    • CB: i7-8700K OC, RTX 2080 FE

    • i7 9700K, RTX 2080 FE

    • i9 7940X OC, Palit RTX 2080 TI

    • 7 1700 OC, Palit RTX 2080 Dual OC

    • 7 2700X, GTX 1080 Ti

    • CB: i7-8700K OC, GTX 1080 Ti FE

    • CB: i7-8700K OC, Radeon VII

    • CB: i7-8700K OC, RTX 2070 Ref

    • 7 1700 OC, GB RTX 2070 WindForce

    • CB: i7-8700K OC, RX Vega 64 Ref

    • CB: i7-8700K OC, GTX 1080 FE

    • 7 1700X, RX Vega 64

    • 5 1600X, Sapphire RX Vega 56 OC

    • CB: i7-8700K OC, RTX 2060 FE

    • CB: i7-8700K OC, RX Vega 56

    • 5 2600 OC, Asus GTX 1070 Dual OC

    • CB: i7-8700K OC, GTX 1070 FE

    • 5 1600 PC, RX 480 OC

In addition, the results show that the actually GPU-heavy scene with a fast graphics card like the GeForce RTX 2080 Ti can then be quite CPU-limited. A reader with a heavily overclocked Core i7-8086K and a similarly overclocked GeForce RTX 2080 Ti, including overclocked memory and a water cooler, achieves by far the highest FPS values. And a created video (an exemplary reader service!) Shows perfectly that all settings as well as the benchmark procedure and the FPS values ​​are correct.

Reader benchmarks in Metro Exodus – 3,840 × 2,160, DirectX 12

    • i7-8700K OC, RTX 2080 Ti OC

    • 7 2700X OC, Gainward RTX 2080 Ti OC

    • i7-8700K OC, RTX 2080 Ti OC

    • CB: i7-8700K OC, RTX 2080 Ti FE

    • CB: i7-8700K OC, RTX 2080 FE

    • CB: i7-8700K OC, GTX 1080 Ti FE

    • CB: i7-8700K OC, Radeon VII

    • 7 1700 OC, Palit RTX 2080 Dual OC

    • CB: i7-8700K OC, RTX 2070 Ref

    • 7 1700 OC, GB RTX 2070 WindForce

    • CB: i7-8700K OC, RX Vega 64 Ref

    • CB: i7-8700K OC, GTX 1080 FE

Help from the tool developer for hopefully further benchmark results

It is also noticeable that surprisingly many, almost all, have resorted to the tool CapFrameX (download) for the optical evaluation of the OCAT (download) measured values. Matching with this, the developer has mixed with the forum, contributed his own values ​​and helpful tips on the tool itself.

But with the first update, the community benchmark is far from over. The editors hope for many more results in order to update the article again. Throughout the weekend benchmarks can be posted on, all of which are still included in the article. Insofar, it still applies: Happy benchmarking!

This article was interesting, helpful or both? The editorial team is happy to receive any support in the form of deactivated ad blockers or a subscription from ComputerBase Pro. More about Ads on ComputerBase.

Leave a Reply

Your email address will not be published. Required fields are marked *

Google Play Store: 2018 scanned more than 50 billion apps daily

Google Play Store: 2018 scanned more than 50 billion apps daily

Hidden Side: Lego combines analogue building with augmented reality

Hidden Side: Lego combines analogue building with augmented reality