PC power consumption in games

Under normal conditions, the issue of power consumption of a home PC under load is only related to the power of the power supply unit and how to organize the cooling system so that it can cope even in the summer heat. Of course, you can also calculate the cost of 1 hour of gaming according to electricity tariffs. But, to be honest, even with current prices, even the most avid gamers will end up with a monthly amount that is disproportionately small compared to the price of PC components. After all, budget computers consume little power at the same time, and if a user has a conventional 100,000 UAH for a powerful gaming machine, then he or she will most likely have some 500 UAH to pay the electricity bill for daily hours-long gaming sessions. So the question seems to be interesting for the Test Lab, but not very important from a practical point of view. Under normal circumstances.

However, as of 2024, conditions in Ukraine are anything but normal. Thanks to the titanic efforts of the Defense Forces and energy industry workers, at least some Ukrainians continue to have the luxury of playing games on their home PCs (it's worth recalling here that games are a great means of psychological relief and relaxation, which is very necessary to keep the "cuckoo" in its place). And nowadays, you need to be thoughtful about electricity consumption while playing games. This applies to everyone in general to contribute to reducing the load on the system. And, of course, it becomes critical when the system is powered by batteries, because the capacity is finite and it would be a shame to be left without power until the end of the stabilization shutdown due to your own entertainment...

In his recent article Energy-independent editorial office, Oleg Danilov has already raised this topic, describing the behavior of PCs in several games:

"As it turned out, the energy efficiency of games does not always depend on the quality of graphics, but sometimes simply on the skill of developers who know how to optimize their games. For example, in Windows, my PC with a monitor consumes about 80-100 watts, depending on some internal processes of Windows 11, plus another 20 watts is required by the dual-band router, plus 30 watts for the fan at minimum speed. Euro Truck Simulator 2/American Truck Simulator "consume" 340-360 watts, Dragon Dogma 2 - 350-380 watts, The First Descendant - 380 watts, and The Crew Motorfest, which is even better in terms of graphics than the previous three games, consumes only 230 watts. Victoria 3 "eats" 190 to 320 watts, depending on the map scale from "paper" to detailed with towns, forests, and steam trains. Magic: The Gathering Arena consumes only 170 watts, while TerraScape, which doesn't seem to have any super-cool graphics, consumes as much as 350 watts. But it should be noted that the consumption in games strongly depends on the specific processor and video card, as well as on the restrictions set by the developers, for example, on the fixed fps."

This inspired us to prepare a separate material with checks in more than 50 games and benchmarks, plus more detailed testing in different modes.

Test bench

All components worked in their normal mode out of the box, without any forced manipulations with frequencies and voltages. The topic of optimizing power consumption by hardware is very interesting in itself, but it requires a separate article so as not to distract attention from the main focus of this article.

The measurements were made with an Intertek BE009 household electricity meter, which powered only the system unit. No additional USB devices were connected to the computer, except for the keyboard and mouse mentioned above. Of course, at the very least, the monitor and router must also be powered in order to play games, so to get a realistic picture of "how many hours you can play from a battery with a capacity of XX kWh", they should be added to the overall calculation. However, measuring their consumption together with the system unit will only harm the integrity of the experiment.

When idle in Windows 11 Professional, this configuration consumes 96 watts steadily with almost no fluctuations. This is the starting point for all other measurements.

Synthetic tests

Before proceeding to the actual game testing, it was useful to find out how much the assembled PC consumes in a "vacuum", that is, in synthetic tests that load either the processor or the graphics core.

It is noticeable that each test provides a very even load, in many cases there is no difference between the minimum and maximum consumption at all. You can also roughly calculate the expected maximum combined CPU+GPU load: this is the maximum CPU load plus the average Blender GPU Benchmark score minus the idle consumption. That is 357+((292+303)/2)-96=558.5 watts. Looking ahead, this is how it turned out with an error of less than 1%.

Power consumption in games

At this point, a dilemma arises as to how to build the material further, because the test in games revealed two trends in general: different games consume differently (which is logical, it is the details that are interesting here) and different settings in the same game consume differently. Both aspects are important, but you have to choose from something - so let's look at the first one first, i.e. the picture for different games. And here it is important to keep in mind the second aspect, i.e. the settings. All of the tests below were conducted with the best graphics settings possible and at a resolution of 2560×1600. In many cases, the results can be adjusted quite significantly by changing the settings to squeeze out less consumption. But then this would prevent an objective comparison.

Speaking of the objectivity of the comparison. It immediately became apparent that, unlike synthetic tests, consumption in games is quite dynamic and depends on what is happening on the screen. For greater laboratory purity of the study, it would be worth simulating real gameplay in different scenes for a fixed time (for example, 1 hour) and measuring the total consumption during this time, but with so many games, this would increase the time to prepare the article by an order of magnitude...

So "we have what we have" - measurements were made either in the built-in benchmarks, because they are perfectly indicative, or during 10-minute gaming sessions. Manually, we recorded the realistically lowest (conditionally close to "1% low") and realistically highest ("1% high") consumption values. From the observations, we can say that the average value in most cases is indeed somewhere between the lowest and the highest, so we can be guided by this.

As you can see, the set of games is very diverse - from the classics of the 1990s to the most modern AAA titles. Of course, it's impossible to cover the entire spectrum, so the selection is limited to the author's Steam and Epic Games libraries, time to download from the network, and common sense.

The first non-obvious trend: The quality of 3D graphics is not always crucial. The best example is Dredge and Whisker Waters, ironically two arcade fishing simulators with roughly equally simple but 3D graphics. Dredge requires 160-180 watts, while Whisker Waters is quite an "adult" 375-446 watts. On the negative side, this is also illustrated by 474-488 watts in Rise of Gun, where the graphics are frankly primitive. At the same time, the latest hit Black Myth: Wukong, which produces a great picture and is very demanding on hardware, loaded the PC only at 405-413 watts. And in the graphics benchmark, Civilization VI was the champion in terms of both minimum and maximum consumption (550-562 watts), although its graphics are not at all fantastic today.

The second non-obvious trend: games with diverse gameplay have a greater variation in consumption. If you look at Suzerain, you can see that there is almost a 100% difference between the minimum 120W and maximum 235W, while in most games this difference is between 10 and 20%. This is because most of the game play in Suzerain takes place in a window with text and answer options, and the peak figures are obtained with relatively infrequent movement on the 3D map.

The third trend (already obvious): casual games on the same PC can be played about three to four times more economically than in 3D. But there is also a noticeable difference in the details. For example, Sunless Sea, which is an older game, consumes a minimum of 108-116 watts, while its descendant in terms of idea and gameplay called Sunless Skies from the same developers already requires 145-170 watts. Well, retro gaming, represented by Baldur's Gate II, Duke Nukem 3D and Quake II, generally demonstrates champion applications for frugality.

Detailed testing

Having considered the overall picture, it is time to move on to the details, because more thorough testing in specific games helps to reveal some things that are not visible in the graph above.

For starters, Cyberpunk 2077. A game that, as a result of a long evolution, has become a model not only for the game itself, but also for the benchmark. The results show two things: depending on the settings, the performance in frames per second varies six times, and not only does the consumption jump by only 20% from the lowest to the highest value, but it also does not directly depend on the number of FPS! But that's because in each case, the system is running as fast as it can.

To confirm this fact, we checked Hitman 3, which also allows very wide graphics quality settings when ray tracing, DLSS, and frame generation are enabled. At the same time, the average consumption in different modes still remained fairly uniform. The anomalies of the minimum and maximum values are quite repeatable, this is not an error, but no significant conclusions can be drawn here except for the simple fact that the game works this way.

And for the next experiment, there are two arcade racing simulators - The Crew: Motorfest and Forza Horizon 5. The former has a forced 60 FPS limit, which cannot be removed in the settings, but can be reduced to a "console" 30 FPS. In 60 FPS mode, this game, if you look at the overall chart, already stands out for its lower consumption compared to other modern games - 325-350 watts. If you forcefully limit the number of frames per second by half, the consumption drops, of course, not linearly, but very noticeably. It's very debatable whether it's pleasant to play a 30 FPS, but at the same time a completely modern 3D game like The Crew: Motorfest in this mode is already approaching casual games in terms of power consumption.

Forza Horizon 5 allows you to study the issue in more detail, as it has both a mode with no performance limits at all and a more varied setting for the frame rate limit. Accordingly, even switching from the "unlimited" mode to 144 FPS (synchronized with the monitor, in this case) can reduce consumption by 20%. The quite decent 72 FPS mode allows Forza Horizon 5 in this configuration to run with a lower CPU and GPU load than The Crew: Motorfest in 60 FPS mode.

Further reduction also helps to save money, with 172-190W at 48 FPS looking more attractive in terms of playability than 160-179W at 36 FPS. There's also a separate mode without DLSS, which loads the system more than the "unlimited" DLSS option, while delivering fewer frames per second.

Another demonstration of the usefulness of the FPS limit is Far Cry 6. Switching to 60 frames per second allows you to save electricity very significantly. Just for fun, testing the HD texture pack at least on a GeForce RTX 4070 Ti SUPER graphics card with 16 GB of memory revealed neither a change in performance nor a change in consumption. By the way, it would be interesting to additionally test a few games with the ability to limit FPS on other PC configurations, say with GeForce RTX 3060 and GeForce RTX 4090 - most likely, the difference will be more noticeable the less resources the PC will use to perform the necessary calculations at 60 frames per second, but at the same time, the hardware itself consumes differently.

Anno 1800 allows you to check performance in two modes using DirectX 11 and DirectX 12. It's not worth making any far-reaching generalizations from this test, but this particular game in DX12 both runs faster and consumes less.

In general, the behavior when changing graphics settings for each game is individual and depends on how well the engine is optimized for a particular technology. Here are two examples: in Skull & Bones, enabling DLSS moderately speeds up the game and even slightly reduces consumption, and in the case of Red Dead Redemption 2, the same DLSS gives an almost threefold increase in speed, but at the cost of a significant increase in energy consumption.

Total War: Pharaoh and Wolfenstein: Youngblood are valuable for our test because they offer not just one built-in benchmark, but two or three with different conditions within the game. The extent to which the consumption varies is a good reminder of the unevenness of the game play, so the performance during a real game session may differ from the measurements in the benchmarks.

This is most clearly demonstrated by Civilization VI, which surprised us with its exceptional load during the graphics benchmark. After all, during the second built-in test, which checks the computing speed of artificial intelligence instead of graphics (but when displaying a normal 3D picture for the game at the same time), consumption dropped by 2.5 times!

Conclusions

In modern games, without any additional optimizations, you can expect consumption in the range of 400-550 watts on this configuration. This gives a general answer to the question "how to prepare for a game in a non-volatile state" and the answer itself is inconvenient, because even with a full-fledged autonomous power supply system at home with a 5 kWh reserve, the PC alone will use it up in about a 10-hour game session. There is a certain difference in that this session in Diablo IV will be an hour longer than in Assassin's Creed: Mirage, but in general, the situation is realistically quite unfortunate. From a small charging station like Ecoflow River 2, such a system will generally last less than 1 hour...

As for other games, we can't make clear generalizations either. Yes, simpler games consume less, but not everything is so obvious. The most systematic approach would be to buy a meter (it is still an inexpensive and useful thing to have around the house, especially now) and measure yourself in the games you are interested in. This is the only way to get real personalized answers to the question of what to play during blackout. At the same time, the best way to reduce consumption within the game is to set the frame rate limit to at least 60 FPS. The less CPU and GPU resources the game actually consumes, the lower the wattmeter reading will be. Well, more radical methods of power saving (such as underwriting and downlocking, among others) already require a separate study and should be the subject of a separate article.