USB to NVIDIA GPU

    Information about VirtualLink, which appeared in Anand last year (and was partly also duplicated on Habré ), did not allow to fully compose the impression of one of the important innovations. The other day, the presentation of the ASUS Zephyrus S gaming laptop, where VirtualLink technology has already been introduced , made it possible to get acquainted with the features of the new interface.

    image

    So, VirtualLink is just another implementation of the USB bus in the Type-C construct. What is the feature? Unlike the classic implementation, where two USB 3.x ports and one USB 2.0 port are connected to the USB-C connector, the bus variant customized by NVIDIA uses three USB 3.x ports. Alternative functionality (or, if simpler, the ability of system logic to transmit signals that are not related to USB management via USB bus conductors, such as digital video signals, for example) is inherent in both the classic Type-C implementation and the advanced VirtualLink solution.

    Consider the pin assignment of the USB-C connector - the basic construct for the NVIDIA VirtualLink interface:

    Pin assignment of the USB Type-C connector, modified for the needs of the NVIDIA VirtualLink interface

    The green color shows the differential pairs of the two USB3 ports defined in the classic Type-C variant. The abbreviation DP (Display Port) means that in this context these lines can be used not as USB3-signals SSTx, SSRx, but as Display Port lines. This means that the signal multiplexers serving the port are switched for video transmission - the port is used in the alternative functionality mode.

    The red color in the center of the connector shows the four signals that were the subject of customization of VirtualLink. In the classic Type-C, these are USB2 lines of a single port, the preservation of which is possible when the connector is flipped by switching.

    In a custom version of the port that supports the VirtualLink interface specification, these 4 lines are used to implement another portUSB 3.x. Thus, we get a total of three USB 3.x ports instead of two - six differential pairs instead of four . At the same time, we lose the USB2 port. The cable now requires the transmission of signals over these lines in a bandwidth of 10 Gbps, against 480 Mbps as it was before. Old cables will not work.

    Why did you need another SuperSpeed ​​port? NVIDIA claims it to service the motion sensor: link to the headset for motion tracking . It is clear that such a solution has better prospects in comparison with the legacy USB 2.0 port, for which it is difficult to find a worthy application today.

    Also, the classical approach to the implementation of alternative functionality turned out to be insufficient for the needs of virtual reality. VirtualLink provides two options for using green 3.x USB lines. We consider an option in which both USB3 ports (all 4 differential pairs) work in alternative functionality mode ( four lanes of HBR3 DisplayPort ) or one port in alternative functionality mode, and the second in USB3 port mode ( two lanes of HBR3 DisplayPort + two lanes SuperSpeed ​​USB 3 ). In any case, the ultimate video mode will support an 8K display with a 60 Hz sweep frequency.

    And about the main thing. The USB ports used in VirtualLink are formed not by the Intel processor, not by the system logic of the platform, but are located directly in the GPU. Which is quite justified in terms of the use of USB bus lines as an alternative for transferring video to VR devices. Do not forget that this functionality will require additional costs - the node of the NVIDIA Turing TU104 video adapter, capable of providing power to the connected peripherals, will consume 35 watts more. You have to pay for everything.

    Also popular now: