Hacker News Reader

Comment by tpmx | Hacker News Reader

"The United States, meanwhile, enjoyed a 5G maximum speed of 1815Mbps – about three times as fast as 4G users’ maximum speed."

This report seems to be lacking a reality check. (Is it practically possible to get ~600 Mbps from LTE? Exotic bundles like LTE MIMO don't really count unless they're actually offered as part of regular plans.)

View on HN

You pretty much need a direct line-of-sight to the 5G tower, without trees in the way, or clouds:

https://www.youtube.com/watch?v=_CTUs_2hq6Y

That video had so much bullshit I had to stop. 1ms ping? Get a clue dumbass. Physics.

Can you explain why 1ms ping is not possible because of physics? Otherwise this comment comes off as rather rude and unhelpful.

I bet he doesn't understand that 5G in the millimeter wave bands will have an effective coverage radius of about that of a street light. At those tiny distances 1ms is possible.

But not implemented. ALL current "5G" networks are 5G non-standalone in which 4G does all registiration with the basestation and most of the work. 5G is just a bit of extra data channel on top.

Speed of light is about 300km per ms, which is a bit larger than a typical street light covers. I think we're going to need a few more divisors to get 300km down to 30m.

Network latency is limited by the speed of light.

Yes but the speed of light is very fast. Light can travel nearly 200 miles in one millisecond.

Please explain how 5G would overcome the obstacles to 1ms ping better than 4G. Higher bandwidth does not reduce latency for minuscule packets with low contention.

Correct, light travels at 186 miles per millisecond, but virtually nothing we have can actually transfer a single packet anywhere near light speed. Take two high end expensive network cards with fiber SFP+ connectors and connect them with just 1 meter of cable, and I very highly doubt you’ll get a ping time of under 0.1ms. And that’s just 1 meter! If speed of light was the limiting factor, that should be down in the nanosecond region.

I hate to be that guy but I have done that test and the latency is under 30 us or 0.03 ms.

I meant to say 0.01ms :(

(Hence why my later part of the comment references “nanosecond region”)

Thanks for calling it out!

Edit:

To say my original post another way, about the fastest I’ve ever seen for a ping time is ten microseconds over a one meter fiber optic cable. That’s 3,000x slower than the speed of light. Fiber optic cable is far more efficient than a Point-To-Multipoint radio which is in a noisy environment (aka cellular), so to think you’ll get within 2 orders of magnitude using 5G vs the best fiber speeds I’ve seen seems crazy to me.

Sorry - I did not attempt to come off as rude or unhelpful.

My point is his comparison of 14ms against his expectation of 1ms. That is not realistic, given the use-case being a speed test to [semi-random] provider.

There is a physical limit in play in these scenarios - being the speed of light in a vacuum and light refraction index (fiber). Simply assuming that due to the fact that 4G is "becoming" 5G you should see an order of magnitude less latency (1ms vs 14ms) is in my view, from a physics perspective, silly.

Even including the RAN (Radio Access Network), the majority of said latency to and from the throughput server endpoint is most likely NOT the result of delay on the RAN, but rather the physical distance (proximity) from the mobile station (MS, smartphone/handset).

At the 30.72 MS/s rate, 2k FFT size, 1201 subcarriers occupied per FFT, QAM256 on each subcarrier, 4x4 MIMO, you get ~576Mbps max throughput (pure number of bits transferred per second). This ignores encoding, cyclic prefix, and all the other overheads associated with the LTE protocol.

So I guess that's where the article author got the "LTE speed". It was probably from some aussie mobile operator's "theoretical maximum speed" number or similar.

there's also carrier aggregation, you could multiply that throughput by the number of carrier components provided by the telco (rarely more than 2 or 3).

Yes, it is possible to get 600Mbps from LTE, but you need lots of mid-band spectrum - the US operators are quite challenged in that respect, plus CPE capable of carrier aggregation + higher MIMO ('flagship' smartphones generally).

This article provides a good overview of how the Telstra deployment works: https://cellularinsights.com/telstras-gigabit-class-lte-netw...

Reply on news.ycombinator.com