Skip to content
The Chat Archive
Unspecified NVIDIA model (HelpSteer2)· nvidia NVIDIA HelpSteer2

Let's say i have 3 devices. device 1 connects to device 2 and device 2 connects to…

  1. User

    Let's say i have 3 devices. device 1 connects to device 2 and device 2 connects to device 3. Device 1 needs to pass data to device 3. this data is time sensitive and there is a field that provides the time at which the data was generated. this comes through fairly irregularly. there is another message that comes through stating what the device 1 time is, and this comes through much more frequently and can be handled with low latency by device 2. Device 2 will pass the time sensitive data recieved through to device 3, but not at the same rate it is recieved. Device 2 also has it's own latency internally to handling and passing this data. before the time sensitive data goes to device 3, device 2 will insert the latest time from device 1, such that the time sensitive data is stamped with the time it was generated and the time that it was sent to device 3. understood so far?

  2. Assistant

    I think I understand the basic idea. But can you clarify how the time values are actually communicated between devices? Does device 1 have some sort of clock that is communicated to device 2? Or is the time value just a raw integer that is passed between devices?

  3. User

    device 1 has an internal clock. this updates every 64 microseconds, and is communicated in form of an integer such that each value represents 64 microseconds. This integer range is only small, and will roll over once it gets to 255 (starting at 0). this clock value is communicated to Device 2 asychronously, but at least once before every roll over. understood?

  4. Assistant

    Understood.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.