Cloud Gaming

One of the topics that attracts the most interest in the "green gaming" space is cloud gaming. Does it help? Does it hurt?

We took a long look [see article in Computer Games Journal] at this in our lab, running scores of tests with different hardware and software configurations, and different games. Here's a high-level summary. More details in our reports. Our "Cloud-gaming" scenario for the entire United States assumes 75% of all gaming hours are on cloud (vs 20% in Baseline). It results in gaming energy demand rising 17% above baseline in the five years between 2016 and 2021. Note that this is in isolation from other changes that may be happening in parallel (e.g., shift from consoles to PCs, or visa-versa).

The two charts below tell the story. The on the left shows the "Strong Uptake of Cloud-based Gaming" scenario, in context with other possible scenarios. These are book-ended by the dotted green line (low case) with high efficiencies across all platform types, and a transition to greater market share of consoles, and no improvements at all in efficiencies or changes in the mix of systems people use to game, i.e. the "Frozen Efficiency and Market Shares" scenario. The chart on the right shows the effect of cloud gaming at the individual system level of cloud gaming versus purely local gaming.

<Click images to enlarge>

Caption: Values shown include all user modes for gaming devices: gaming, non-gaming, video streaming, web browsing, idle, off. Cloud gaming and video-streaming values include network energy and energy used in the data center. Lower values for Entry-level systems reflect the relatively high proportion of “Light gaming” user types (fewer hours spent gaming). There is currently no cloud-based gaming option for PS3, Xbox 360, Nintendo devices, or Apple TV. Display energy not included. These values are averages, and are weighted by the proportion of "light", "moderate", "intensive", and "extreme" gamers in the user base for each system type. "Light" users of "Entry-level" systems consume 287 kWh/year while "Extreme" users of "High-end" consume 2,111 kWh, which is more than a 7-fold variation. For laptops, the variation is almost 17-fold.

Our project consistently looks at all uses and modes of gaming devices, so we consider gaming (local and cloud) but also modes such as video streaming, web browsing, idle, off. We're interested in total energy use of the equipment, across all it's modes of use. These modes are more fully defined in our reports. We also break down the market and "fleet" of gaming devices into several user types (reflecting intensity of use).

"Streaming" in the figure above refers to video streaming. The red strip in the bars is the network component of energy used when in video-streaming mode. "NonCloud" is energy use when no cloud services are being utilized. That's the reference point we compare against to see how much more energy, in aggregate, these systems use when connected to cloud services.

Of course given systems will use more or less than these averages. Significantly more in many cases. The bookends of "Light" users on "Entry-Level" systems versus "Extreme" users of "High-End" systems represent a 7-fold difference in energy use for desktops and 17-fold difference for laptops. See caption. Even more differences would no doubt be found in the wild, as our numbers reflect 'only' the 23 representative systems we tested in detail.

Here are the questions that come up most often:

  • Does cloud-based gaming save energy?

    • Unfortunately, no. In fact, in all of our testing we found that cloud-based gaming requires significantly more energy than similarly powerful equipment located in the gamer's home -- three-times as much in the most extreme cases we identified (see Fig. 5 here). A key reason for this is that data centers hosting the high-power gaming servers require very substantial ventilation and air conditioning. To this one must add the non-trivial network energy between the server and the gamer. And, of course, the connected user-side equipment still uses some energy, even when the heavier lifting is shifted to a remote server. The added energy is even greater if the home system is less graphically powerful than the cloud-based system and/or the games run in the cloud are more compute-intensive than what the gamer would select on their own local equipment. The impact will be minimized by very "thin" clients on the user side. New research from other scientists in our lab has found that data center energy efficiency continues to improve in leaps and bounds. That said, increased throughput is largely offsetting these gains -- which isn't good news for the climate.

  • How much of this extra energy is associated with the network as distinct from the data centers?

    • In our calculations for PC cloud gaming, the data center is responsible for about 340 Watts of power per user and the network an additional 180 watts. So, both components of the overall gaming environment are quite important. The corresponding values for console cloud gaming are 180 and 120 watts, respectively. Internet network energy is covered more deeply in this article. It is important to note that efforts to make networks more energy efficient have achieved dramatic improvements in recent years and can be expected to continue to do so.

  • Does cloud-based gaming (potentially) provide environmental benefits by reducing the need for local hardware, which means less e-waste, etc?

    • We haven't evaluated the solid-waste side of this topic, but it's a good question. Of course, if componentry is shared among multiple gamers, that can translate into less gear per gamer (although, only in the first generation of remote implementations one gamer can use a given GPU at a time, so those servers are enormous -- typically 8 GPUs). This will hopefully change for the better. Two countervailing factors come to mind. On the one hand, if centralized technology lasts longer there is less solid waste per hour of gaming use. However, turnover could well be faster in data centers, given the pressure to keep with the latest equipment, and longer operating hours than home-based systems. Also, the data center facility itself and the associated infrastructure--not to mention land use--as well as the networking gear represent equipment that is not needed for distributed gaming in existing buildings.

  • How can this new source of gaming energy use best be managed?

    • Like most energy-using enterprises, a given amount of work can be achieved with widely-ranging amounts of energy input. That holds for refrigerators, light bulbs, cars, and servers alike.... Herein lies the main opportunity of energy efficiency in general and green-gaming in particular. Energy use for cloud gaming will depend on the specs of the servers, how fully the servers are utilized, efficiency of the data centers, efficiency of the network equipment used to link gamers to the data center, and the energy used by the gamer's local device. The local device hopefully employ power management that minimizes that particular load. The extent of power management varies widely across gaming devices (see Fig 22 in our report). Cleaning up the grid that serves data centers is important, as it is for home-based gaming. When cloud gaming, the client on the user's side should be an "thin" and efficient as possible and displays as efficient as possible, as they both of course use power even when the rendering is performed remotely.

For more discussion, see the various articles in the trade press and popular media: BBC, Digital Trends, New Scientist, and Wired.