Hacker News Reader

Comment by gst | Hacker News Reader

Same in Austria (but 20 EUR instead of 10). LTE is a competitor for ADSL and Cable here and LTE providers sell WiFi routers that use LTE instead of ADSL/Cable as uplink.

I recently replaced my home ADSL connection with LTE (for 20 EUR a month: https://www.yesss.at/tarifoptionen/unlimited) and get approx 2x the download bandwidth and 10x the upload bandwidth for the same price as my previous ADSL connection. Even when there's congestion (which doesn't happen a lot) it's still significantly faster than ADSL.

View on HN

How is latency for you? When watching a stream or downloading latency which playing an online game jumps up to ~1s (1000ms) in the game. Croatian TELE2 provider, indoors at -92 dBm. Usual latency is about 15-20ms doing ookla speedtest but "real life" it's about 40-50ms.

This issue is likely to be bufferbloat [1] and not directly related to LTE. If you can put a properly configured router in front of the modem that may help. I haven't looked into it but it seems recent OpenWRT have the necessary support to mitigate bufferbloat.

[1] https://www.bufferbloat.net/projects/bloat/wiki/Introduction...

Depends on the server you want to reach. Same country you can get latency ~20 ms. If you get lucky also lower. But haven't seen anything below 10 ms. Also "-92 dBm" doesn't say much. There are 4 relevant values for LTE: Rssi (probably your -92, the most irrelevant one), Rsrp, Rsrq and Sinr.

Especially the upload makes it so much better. Usually nobody is uploading much.

>get approx 2x the download bandwidth

What about latency?

In a LTE network that is not over its capacity the latency is a bit under 20ms these days in my experience. With 5G this should go down to 1 to 2ms.

What accounts for the difference in these figures? Is it understandable for someone without in depth knowledge or not really?!

Latency in LTE is, I think 5 ms. The rest from the delay is normal network delay and not an issue of LTE. So with 5G (with everything else being the same) will reduce the delay by about 4 ms.

Not really a network engineer so cant really help with that. The 20ms is just my experience and the 1 to 2ms is the theorical number given by the people designing 5G. Found this writeup but still not really sure beyond improvements to the slot-based framework and higher frequency as being the reasons

https://www.edn.com/electronics-blogs/5g-waves/4460346/5G--T...

All these latency numbers vary a lot based on connection quality. As I understand you basically need line of sight for 5G to work at its maximum frequency and will fallback to what is basically 4G when such a connection is not possible. But then again the antennas/stations for 5G should be much smaller and thus easier to deploy closer to the end users.

Latency will go down with 5G if you pay for it. How much is the big question.

There are 3 sides to 5G: - eMBB: massive boradband --- high throughput, mmWaves; - URLLC: ultra reliable and low latency, for industrial applications typically; - mMTC: massive IoT. Cheap and low power.

The network can handle all 3, but as a device or user that doesn't mean you'll get all the benefits at the same time, some choice will be required. The reduced latency always has some cost impact.

Today the LTE air latency is 4-5 ms, the rest is in the core network. 5G has some air interface changes to reduce the air contribution to 1 ms, in the context of URLLC. For basic users, it'll likely be like LTE (already pretty good). There may be some general improvement if they optimize the core latency for all, but TBC.

Reply on news.ycombinator.com