4 ms·

> The VR problem is latency. We should be able to get sense/computation/display down to the 10-15ms required, but distance introduces additional latency, due to

> The VR problem is latency. We should be able to get sense/computation/display down to the 10-15ms required, but distance introduces additional latency, due to c. If a ns is roughly a foot, 15ms is 15000 feet, gives a theoretical maximum diameter (not radius) for remote "city limits". Wow, the VR problem is latency? Have you ever had a real-time conversation with anyone outside your hemisphere? If 15ms is 15,000 feet then 1/0.015 = 66 times that should be the speed of light, i.e. 300,000 km/s. Does that (66 x 15,000 ft = 300,000 km) seem even vaguely in the right ballpark to you? You're off by a factor of 1000. It's 15 microseconds, not milliseconds. And incidentally, this might be a good excuse to go travel outside the country and meet people from elsewhere on the planet... you never know what you might learn about the world around you.

akvadrako 9y ago
His math is way off, but the point still stands. The additional lag due to distance is noticeable above 5.000 km or so. Maybe it's good enough - but in VR, it might be jarring.
- abetusk 9y ago
  For context, the diameter of the earth is 12,742 km. So it would take roughly 30ms to make a round trip. Assuming that a signal need only traverse half the diameter as you can take the shortest arc, you get down to ~6371km which brings you back down in the range of 15ms. I assume the 15ms was chosen for the 60fps that VR needs to run at to be considered reasonable. If so, having a local machine handle redraws would be reasonable and requiring a 30fps update on all "agents" and other world elements would probably be pretty unnoticeable by most people.
  - akvadrako 9y ago
    Typical transatlantic latency is way more than 15ms - more like 50 one way. To Asia and Australia is much more. Thats on top of processing and last mile delays. Anyone can notice the total latency in even a Skype conversation.
hyperpallium 9y ago
You're being somewhat uncharitable of an off-the-cuff milli/micro error, and going on to make incorrect personal assumptions.
- future1979 9y ago
  I thought your 15ms comment was correct. It is 16ms motion to photon latency. Why microseconds?
  - MaulingMonkey 9y ago
    15ms is correct (or even wildly optimistic) for VR on account of hardware buffering for GPU throughput / input filtering for stability, etc., LCD reaction times, and other local software and hardware stuff. The part that's not correct is 15ms ~ 15,000 feet @ c: 15ns ~ 15 ft 15us ~ 15,000 ft 15ms ~ 15,000,000 ft 1ns ~ 0.9836 ft per https://www.wolframalpha.com/input/?i=speed+of+light+*+1+nanosecond+in+feet https://www.wolframalpha.com/input/?i=speed+of+light+*+1+nan... Some other fun figures: https://www.wolframalpha.com/input/?i=earth+circumference https://www.wolframalpha.com/input/?i=earth+circumference Light travel time t in vacuum from t = x/c: | 134 ms (milliseconds) Light travel time t in an optical fiber t = 1.48x/c: | 198 ms (milliseconds) You wouldn't want to wait for a camera to turn in response to head movement at these distances, but we're already warping buffers in VR to reduce apparent latency - to hide it better for head rotations. You wouldn't want to try and directly control twitchy aircraft or race cars with this kind of latency, but this is still under e.g. throttle response times on a modern car, and you've probably dealt with worse round trip ping times for online gaming.
- throwaway613834 9y ago
  The issue was not the "milli/micro error", but the seemingly completely reckless lack of disregard for the accuracy of your personal conclusion, which you were presenting without any doubt whatsoever as a fact. I'm normally pretty forgiving of mistakes, at least when it comes to silly mathematical errors or typos. I'm also pretty forgiving of wildly wrong ideas when there the speaker conveys a fair uncertainty along with the idea. What I'm not going to brush off as a silly error is the sheer overconfidence in and propagation of wildly wrong information that could've been corrected with only 1-2 seconds of common sense or with any remote attempt at a citation. If it had been a typo or just an arithmetic error, I wouldn't have cared (or even noticed). But you also completely confidently declared that the problem had been identified to be c and that the boundary for VR is "city limits". It means that either [1] you have never had a real-time conversation with anyone much farther away to help you sanity-check this result (in which case it's an innocent mistake, and like I said, it's a good excuse to also go out and see the world), or (or possibly "and") that [2] you did not even care to try to think about this topic for 2 seconds before writing about it -- not when you first heard about it wherever you did, nor when you wrote about it here to teach it to other. Whatever the larger issue, the silly math error is 100% beside the point... you could've written light-years for all anyone cared, and forgotten about it the day after. However, the underlying issue is not the kind of problem to forget about the next day. Lots of alarm bells should be going off in your mind that you need to actively do something to correct whatever the root problem is (which I was trying to help you identify), because otherwise you'll keep actively making the world (and HN) a worse place with this kind of misinformation, and that is something I'm tacitly assuming you don't want to do. So: please, please, please ensure put 1-2 seconds of thought into sanity checking whatever you write in the future -- especially it's too painful to at least qualify what you write with an "I think" so that readers realize you're deriving your own conclusion and not propagating an established fact. We have enough problems in this world with the spread of misinformation that hinges on subtle nuances that we don't also need to put up with problems that should be pretty easy to spot from light-years away.

akvadrako 9y ago

His math is way off, but the point still stands. The additional lag due to distance is noticeable above 5.000 km or so. Maybe it's good enough - but in VR, it might be jarring.

abetusk 9y ago
For context, the diameter of the earth is 12,742 km. So it would take roughly 30ms to make a round trip. Assuming that a signal need only traverse half the diameter as you can take the shortest arc, you get down to ~6371km which brings you back down in the range of 15ms. I assume the 15ms was chosen for the 60fps that VR needs to run at to be considered reasonable. If so, having a local machine handle redraws would be reasonable and requiring a 30fps update on all "agents" and other world elements would probably be pretty unnoticeable by most people.
- akvadrako 9y ago
  Typical transatlantic latency is way more than 15ms - more like 50 one way. To Asia and Australia is much more. Thats on top of processing and last mile delays. Anyone can notice the total latency in even a Skype conversation.

hyperpallium 9y ago

You're being somewhat uncharitable of an off-the-cuff milli/micro error, and going on to make incorrect personal assumptions.

future1979 9y ago
I thought your 15ms comment was correct. It is 16ms motion to photon latency. Why microseconds?
- MaulingMonkey 9y ago
  15ms is correct (or even wildly optimistic) for VR on account of hardware buffering for GPU throughput / input filtering for stability, etc., LCD reaction times, and other local software and hardware stuff. The part that's not correct is 15ms ~ 15,000 feet @ c: 15ns ~ 15 ft 15us ~ 15,000 ft 15ms ~ 15,000,000 ft 1ns ~ 0.9836 ft per https://www.wolframalpha.com/input/?i=speed+of+light+*+1+nanosecond+in+feet https://www.wolframalpha.com/input/?i=speed+of+light+*+1+nan... Some other fun figures: https://www.wolframalpha.com/input/?i=earth+circumference https://www.wolframalpha.com/input/?i=earth+circumference Light travel time t in vacuum from t = x/c: | 134 ms (milliseconds) Light travel time t in an optical fiber t = 1.48x/c: | 198 ms (milliseconds) You wouldn't want to wait for a camera to turn in response to head movement at these distances, but we're already warping buffers in VR to reduce apparent latency - to hide it better for head rotations. You wouldn't want to try and directly control twitchy aircraft or race cars with this kind of latency, but this is still under e.g. throttle response times on a modern car, and you've probably dealt with worse round trip ping times for online gaming.
throwaway613834 9y ago
The issue was not the "milli/micro error", but the seemingly completely reckless lack of disregard for the accuracy of your personal conclusion, which you were presenting without any doubt whatsoever as a fact. I'm normally pretty forgiving of mistakes, at least when it comes to silly mathematical errors or typos. I'm also pretty forgiving of wildly wrong ideas when there the speaker conveys a fair uncertainty along with the idea. What I'm not going to brush off as a silly error is the sheer overconfidence in and propagation of wildly wrong information that could've been corrected with only 1-2 seconds of common sense or with any remote attempt at a citation. If it had been a typo or just an arithmetic error, I wouldn't have cared (or even noticed). But you also completely confidently declared that the problem had been identified to be c and that the boundary for VR is "city limits". It means that either [1] you have never had a real-time conversation with anyone much farther away to help you sanity-check this result (in which case it's an innocent mistake, and like I said, it's a good excuse to also go out and see the world), or (or possibly "and") that [2] you did not even care to try to think about this topic for 2 seconds before writing about it -- not when you first heard about it wherever you did, nor when you wrote about it here to teach it to other. Whatever the larger issue, the silly math error is 100% beside the point... you could've written light-years for all anyone cared, and forgotten about it the day after. However, the underlying issue is not the kind of problem to forget about the next day. Lots of alarm bells should be going off in your mind that you need to actively do something to correct whatever the root problem is (which I was trying to help you identify), because otherwise you'll keep actively making the world (and HN) a worse place with this kind of misinformation, and that is something I'm tacitly assuming you don't want to do. So: please, please, please ensure put 1-2 seconds of thought into sanity checking whatever you write in the future -- especially it's too painful to at least qualify what you write with an "I think" so that readers realize you're deriving your own conclusion and not propagating an established fact. We have enough problems in this world with the spread of misinformation that hinges on subtle nuances that we don't also need to put up with problems that should be pretty easy to spot from light-years away.