Intel's metaverse plan starts by "borrowing" graphics cards

Source: Lei Technology Internet Group

Editor: TSknight

As one of the hottest topics in the tech field recently, the existence of the metaverse has provided countless companies with the opportunity to boast about their technological reserves and layouts. Following announcements related to the metaverse from semiconductor giants like AMD and NVIDIA, which have driven stock prices to soar, Intel has recently also decided to take action.

In a recent report, Intel's Senior Vice President and head of Intel's Accelerated Computing Systems and Graphics Group, Raja Koduri (former Chief Architect of AMD's Graphics Division), stated: "The metaverse may be the next major computing platform after the World Wide Web and mobile internet." He also noted: "Our current computing, storage, and networking infrastructure is simply not sufficient to realize a true metaverse," and "We need cluster computing performance that is 1000 times or more than current servers."

Although Raja Koduri believes that the metaverse cannot be truly realized at this moment, it does not prevent Intel from laying out related technologies for the metaverse. Without mentioning the infrastructure like processors and servers, Intel has also announced a new technology in the software domain, which some commentators believe may become one of the cornerstones of the metaverse.

Intel's Metaverse, Starting from Software?

Recently, Intel unveiled a brand new gaming technology called Continual Compute, aimed at solving the problem of low-performance devices being unable to effectively run high-performance applications, allowing any device to run AAA or even metaverse-level games and applications.

During the technology demonstration, Intel successfully ran the AAA game "Hitman 3" smoothly on a laptop that did not have a dedicated graphics card. As a AAA game released this year, "Hitman 3" has relatively high PC configuration requirements, and even typical dedicated graphics cards struggle to run the game smoothly.

With the help of Intel's Continual Compute technology, the laptop called upon a high-end dedicated graphics card from another device on the same network, then transmitted the computational data over the network to that device, and finally retrieved the computed results via the network to present them on the user's display.

Does this sound familiar? Players familiar with PC gaming might immediately think of cloud gaming or streaming, both of which are no longer novel technologies but have not been widely adopted over the years for various reasons.

It should be noted that, while it sounds somewhat similar to these two technologies, Intel's Continual Compute technology has a fundamental difference in application. Both cloud gaming and streaming run the game on a cloud device or another device and then transmit the game visuals to your device over the network.

In contrast, Intel's Continual Compute technology calls upon the high-performance hardware of other devices to participate in computation, directly transmitting computational data rather than visuals via high-speed networks and special software protocols. The difference lies in that the computation results in the former are actually completed on your PC, which is part of the entire computing system, while in the latter system, your PC is merely a display, and you are just watching a "video."

In poor network conditions, the latter can experience noticeable issues like visual stuttering, while Intel's new technology can temporarily reduce data transmission requirements by lowering frame rates to ensure that the user's gaming experience remains smooth.

Moreover, because it directly calls upon devices, some hardware-specific features can also be supported, such as NVIDIA's DLSS 2.0, which are difficult to realize under cloud gaming or streaming technologies. With the help of Intel's Continual Compute technology, there will be opportunities to directly utilize these advanced AI technologies on one's own laptop in the future.

So, where will this device with high-performance graphics cards and processors come from? Intel's answer is to "requisition idle devices on the same network," provided that the other party "agrees to have their device requisitioned." In simple terms, to enable this feature, two prerequisites must be met: both devices must be on the same network, and the other party must agree to have their device requisitioned. Only when both conditions are satisfied can the relevant features be activated.

Theoretically, any hardware that meets the software protocol can be included in this "rented device" network, regardless of whether the other party's hardware is from AMD or NVIDIA, as long as they are willing to support it. However, in the current technology preview demonstrated by Intel, only protocol handshakes between Intel processors are supported, and there is no specification regarding the types of graphics cards.

From the gameplay videos shown by Intel, there is no significant difference in smoothness between games running via Continual Compute technology and those running directly on the local machine at the same visual quality. However, it is worth noting that Intel ran "Hitman 3" at low graphics settings, and Intel did not demonstrate the performance of this technology at high quality. According to the documentation provided by Intel, this technology is still in the development stage, and what has been shown is just a phased achievement.

Will Continual Compute Technology Be One of the Cornerstones of the "Metaverse"?

After Intel demonstrated the Continual Compute technology, why do some people refer to it as a cornerstone of the future "metaverse"? When talking about the metaverse, we must mention virtual reality technologies like AR and VR, as only with the help of these technologies can we achieve a true "metaverse."

However, whether it is AR, VR, or more comprehensive XR, the premise for providing services to users is that they need to have a device with sufficient performance to support the software. Taking VR games as an example, a typical VR game currently requires a device with performance that exceeds at least that of a PS4 to run, and the experience is not immersive enough for players.

Considering the rendering visuals and real-time feedback required for an immersive experience, even the highest-end personal PCs currently cannot meet the demand. This is why Intel's executives pointed out in related articles that we need to increase computing performance by over 1000 times to possibly meet the requirements for running true "metaverse" software.

Although from a technological development perspective, we will inevitably be able to manufacture devices that meet these performance requirements in the future, their initial size and power consumption will certainly not be low, making it unlikely for individuals to carry and use them at all times. If future metaverse users want to have a complete gaming experience more easily, they can leverage Intel's Continual Compute technology to directly call upon a high-performance PC located on the other side of the room or rent a high-performance PC from the network.

With the use of Continual Compute technology, the user's device only needs to have a reasonably performing processor, a display device, and a network device to run applications that require high-performance hardware. In simple terms, Continual Compute technology allows users to access the performance of high-performance PCs anytime and anywhere to assist them in gaming or launching applications, and in the world of the metaverse, this technology will enable players to obtain a complete experience at a lower cost.

So, what are the advantages of Intel's technology compared to cloud gaming and other cloud computing servers? One advantage is that the individual or service provider providing the service does not need to build a large online server; they only need to connect devices equipped with relevant hardware to Intel's shared network. Secondly, users can obtain data from the nearest device, which can significantly improve issues like latency and packet loss caused by long-distance data transmission.

Moreover, in Intel's plan, this technology will eventually form a computing resource pool at the network level, allocating computing resources through intelligent resource allocation mechanisms in systems like Windows. Every idle computer connected to the network will be part of this resource pool.

At this point, I am reminded of idle PC resource utilization plans like the Phoenix Project, where participating users can receive varying degrees of rewards based on the resources they contribute. In the future metaverse, perhaps we will be able to purchase computing resources from the resource pool like buying bandwidth, without having to bear the costs and expenses associated with high-end hardware.

Additionally, according to Intel, Continual Compute technology is just one of their technological reserves for the metaverse. They are exploring how to reduce the computational performance requirements for immersive content like VR through techniques such as deep learning and neural networks to meet user needs.

However, Intel also admits that in the next five to ten years, chip performance will still struggle to meet the needs of running true metaverse applications. Real-time interaction, ultra-fine real-time rendering, and other issues pose significant challenges for performance and network infrastructure. Therefore, this Continual Compute technology may initially serve as another "cloud gaming" project. At least for now, in Minecraft-level "metaverse" games, this technology has not found much application.