How many people see Zwifters disappear from the screen?

Since 4-6 weeks ago, over and over again, I see the following problem: from time to time, all other avatars disappear from the screen and “Zwifters nearby” list goes blank. I just ride alone for anywhere from 10-15 seconds to a minute, then everything returns back to normal… until it happens again. This happens during free rides and workouts (and is not a problem because, well, who cares about others during workouts), but also happens during races, e.g., Tour de Zwift. I unavoidably get dropped when other riders disappear and I have noone to draft behind, and I have no chance of catching up with them when they reappear some 15 seconds ahead of me.

Zwift has a standard advice which boils down to a sketchy internet connection. The official recommendation to check cable modem and/or wi-fi router. In my case, this does not appear to apply. I checked the log file in Zwiftalizer from this morning’s Tour de Zwift #6 - and found no reported internet outages whatsoever. I installed a Chrome add-on “Internet connection log” on the same PC which runs Zwift and asked my son to tell me when riders around him disappear while he is Zwifting, so that I could immediately check the internet log. The avatars were gone, he told me right away, but there were no internet or wi-fi outage since he started Zwifting.

This makes me fairly convinced that this issue is likely on the Zwift servers / software side.

I opened a ticket with Zwift today and waiting for their reply. My past experience with Zwift customer service taught me to keep my expectations low.

I wonder how many of you see the same issue, and if you also found that it started some time around Christmas or New Year, more or less? If you stumble upon a solution which you are willing to share, that would be even better!

Not seen it to the same extent (where ALL riders disappear). Yes, standard stuff about networks, and drops out and blah blah.

Pretty sure that in recent updates to Zwift, there were tweaks to ‘riders seen on screen’. You say there are no drop outs, but what about outright Internet speed? What does a Speedtest tell you for Download and Upload speed?

Could be PC power (limiting number of riders on screen helps with lower-powered machines) What CPU, GPU and RAM specifications do you have?

Do you have another device to try Zwift on? iPhone, Android? iPad? anything else to see if the problem is there too?

There are plenty of threads about this at this point. Yeah, it seems to happen to me or one of my teammates in pretty much every race, and I’m starting to have a whole library of scenarios where the circumstances make it obvious that it was not a simple connectivity issue.

When this happens in a race, the best you can do is to significantly increase your power and hope that you can maintain the speed of the group without having anyone to draft until the other riders show up again. Overall, this kind of crap together with all the recent weirdnesses in drafting, riding on the wrong side of the road at start and what else make Zwift racing an increasingly frustrating experience.

2 Likes

Brett, your thought about variability of internet speed is an interesting one.

Tests show that my connection meets the specification of the plan which I am paying for, 45 Mbps down and 6 Mbps up.

The speed tests which I know of do not run continuously (and if they would, they would take all available bandwidth), so I am not sure if it is possible to get a log of changes in internet speed while I am riding. Yet, as we all know, internet speed actually depends on speed (and load) on both sending and receiving end. If Zwift servers get overloaded, their speed may get throttled and may lead to interruptions or delays with sending the data. Those would be random and will affect only a subset of Zwifters, which I think is what we are seeing.

The computer is a desktop with Core i7-6850K @ 3.60 GHz, 64 GB RAM, Nvidia GeForce GTX 1080, Windows 10, multiple SSD drives in the computer, no random antivirus scans, no continuous backups, no connection to the cloud which could consume bandwidth or resources. I have my doubts that it can be a computer limitation, at least in terms of its parameters. ANT+ as of two days ago has been brought right next to the trainer using USB extender over ethernet cable, but this had no impact on dropouts of riders (as expected, because dropouts happen in the Zwift program which runs on the computer). There were no disconnects of the trainer from the PC, either, according to Zwiftalizer.

Humm… Those Internet speeds, and PC specs sound good. Good enough (i would say) for Zwift to not drop avatars from the screen. On these forums, @Dave_ZPCMR is a legend for PC performance related questions :slight_smile: Maybe he can help.

Also mentioned is that there are a lot of other posts about this - so it’s not just you.

Server load is a key factor too - maybe i’m luck and connect to one of their less-used servers?

(Also, never heard of a USB being extended over Ethernet… interesting. All i did (to get Ant+ dongle near trainer) was use the ‘recommended’ USB-USB extension cable from Amazon. Rock solid. ( AmazonBasics USB 3.0 A-Male to A-Female Extension Cable 3 m / 9.8 Feet)

Update: got reply from Zwift. The reply was good. They said, basically, that for one thing, they found some network issues in the log file which Zwiftalizer did not pick up, and secondly, brought up a thought of network latency issues which are a little harder to track. I found on Google that Xfinity/Comcast users have reported latency issues which Xfinity struggles to resolve. I am going to install a trial version of Ping Plotter to see if I can monitor latency and learn something out of it.

I am also going to track the timing of the issues better than I did before, e.g., I will take screenshots to document the time when the issues happen to try and match it with Zwift logs, network logs, latency logs.

We will see where it will take us.

Today I showed up for Tour de Zwift Stage 6. Was warming up with a bunch of others, engaged in the pre-ride chatter, when about 5 seconds before the gate opened, everyone else disappeared. I rode away all by myself, completed the hour-plus ride. What’s great is after finishing I wore the green jersey after my TdZ kit transformed back to my everyday kit. First and last time that I’ll be #1 in a sprint!

Are you running Zwift mom a machine that is hardwired to your router, or on WiFi? If WiFi, is your router on a static channel, or set to ‘best channel’? I used to have similar issues until I changed to a static WiFi channel. Apparently, when the router changes to a new channel there is a tendency for Zwift to lose its connection to the Zwift server while the router is searching. You haven’t lost your internet connection, per se, just the connection to Zwift. When this would happen to me all of the other riders would disappear until the router reset on the new channel, then everything would get back to normal. Once I changed to a static channel, the issue disappeared completely.

Please scroll to the very bottom for a summary of the investigation and solution found. I am changing this post to make the thread easier to read.

Good find Andrei.

Symantec should allow a way to open a path from the outside world to your device so look for some functionality around “opening inbound UDP for a specific device”.

The reason why the log file wont likely show much is that UDP is usually used for oneway communications (likely its used in this case for sending other rider location data) and doesnt require handshaking (proof of receipt) unlike TCP which does. You may have some other errors that result from that, but you wont see an inward UDP based error.

Looking at your log, instead of opening up all inward UDP to your Zwift device, it looks like you could reduce to just two subnets which would ofeer better security that “all external udp”

Some more thoughts:

Often providers turn off the ability to ping their servers for security reasons. Pings add overhead that impacts important time sensitive packets too (as Zwift mentioned to you re latency).

Symantec should give a reason why its blocking those packets - fixing that problem will (should) sort your problem out. It is blocking some packets from 54.190.114.139 but not all, therefore the packets are triggering some firewall reason to reject the packet compared to others coming from the same source. EDIT: it may take a few days of observations but it wont be a massive list I suspect.

A workaround is to open up UDP from each source IP that you know is related to Zwift (by relating to similar IP networks. Its not the smartest way longterm but eventually you will have them all and be able to keep Zwifting.

The right way of addressing this is digging into why those inwards packets are getting rejected and that can take time (may include working with Symantec).

What is the range of latency you get, roughly?

Please scroll to the very bottom for a summary of the investigation and solution found. I am changing this post to make the thread easier to read.

20-30ms is fine, and probably even good compared to what some Zwifters would get around the world!

Please scroll to the very bottom for a summary of the investigation and solution found. I am changing this post to make the thread easier to read.

1 Like

Good detective work Andrei. I’ll have to read it a few times to digest however :slightly_smiling_face:

Really curious how you get on.

My son rode TdZ #7 tonight with the new firewall rule. The race was smaller than mine, though - only 150 people. No issues. No instances of riders disappearing. Firewall shows no rejected incoming UDP packets related to Zwift. So far so good!

2 Likes

There was at least one instance of all riders but one disappearing from view even in the Zwift Racing League Premier Division stream on Monday (made with the broadcast graphics engine), the director was quick enough to change to another “camera” so it was only visible for maybe a second. Probably can’t blame the firewall for that…

Like you mentioned in another thread I think, so many issues in play at the moment, hard to know what symptom is being caused by what issue.

1 Like

Very few apps developers do that nowadays, the infrastructure cost is just too high. That’s why AWS makes so much money…