Last post on stutter from me! - DX9 Vs DX11 differences explained!

[Shane]

OK testing it now.

Yes a minor spike at 10 mins and 20 mins.

There were 3 ports operating to 47.190.119.207:
 - UDP 5110 (very tiny data size, tx/rx 1.6kb/1.6kb
 - UDP 5117 (looks like this is the game data) tx/rx 441kb/579kb
 - TCP 5111 tx/rx 224kb/100kb

I also see a connection to 47.190.119.209 on tcp 3011.

Firewall logs showed no dropped packets.

All packets are tiny in size.

So yes I do see your 10 min spike, but no I do not experience your stutter issues.

I'm running Windows 11 on a i7-7700k, GTX4060, an Intel i219-V ethernet adapter. I have 1Gbps fibre, and run a Sophos XG firewall.

What kind of ethernet adapter do you run and what is the driver?

edit: just had a middle of the night thought. I didn't see any ICMP traffic so the netstat test is encapsulated in one of the connections above. Secondly it's entirely possible the 10 min spike is resetting/restarted a UDP ping (i.e. the built in ping in netstat may have a 600 lifecycle).

Maybe because the language we've been using isn't providing clarity...   that spike *is* the "micro-warp/freeze" that creates the visual micro-"stutter/freeze" more a freeze than a stutter, per se; evidenced on the in-game netstat as a visually confirming spike as well... hope that helps?  (There is a noticeable updated plane position when that microfreeze ends, not overly large, but definitely present.)

[Eagler]

When you have one of these does everything,  all planes, zig zag for a second or just the plane you are flying against...

I get the occasional split second zig zag from a plane I am flying against but it does not affect the other planes flying nearby and doesn't happen every 10 seconds..more like just the rare interruption of usually smooth interactions..

I have always thought this was just the odd internet connection hiccup though sometimes it happens just at the right time as if the zig zag can be caused with a connection interrupt but don't have any evidence such a device/ setup exists...

My in the boonies starlink connection has more oddness like this than the cable connection I use at home..

Eagler

[Animl-AW]

Question
Do these spikes 10 from when you log into MA or 10 min from your first sighting?

[Vulcan]

Maybe because the language we've been using isn't providing clarity...   that spike *is* the "micro-warp/freeze" that creates the visual micro-"stutter/freeze" more a freeze than a stutter, per se; evidenced on the in-game netstat as a visually confirming spike as well... hope that helps?  (There is a noticeable updated plane position when that microfreeze ends, not overly large, but definitely present.)

That is my question. I don't see these stutters every 10 minutes. I always play on VR and things like that are usually more noticeable on the VR big screen. Yet I see the netstat spike. We really need HT to tell us what the netstat is measuring, is it a linux type UDP ping or is it an actual measure of the traffic being received.

Then again you say the position change is "not overly large", and if it's only occurring once ever 10 minutes maybe I wouldn't notice. It makes me wonder if this is even the same issue as the other stutters/freezes people are complaining about - especially given some people say they've eliminated them?

[Shane]

That is my question. I don't see these stutters every 10 minutes. I always play on VR and things like that are usually more noticeable on the VR big screen. Yet I see the netstat spike. We really need HT to tell us what the netstat is measuring, is it a linux type UDP ping or is it an actual measure of the traffic being received.

Then again you say the position change is "not overly large", and if it's only occurring once ever 10 minutes maybe I wouldn't notice. It makes me wonder if this is even the same issue as the other stutters/freezes people are complaining about - especially given some people say they've eliminated them?

There are a variety of screen freeze issues, most card related, some DX9 or VR related and generally have fixes....  this 10 min cyclic minifreeze just seems to point to being systemic to traffic as opposed to localized to pc/situational. 

Many won't even notice or not looking at the right time or just dealing with it, not knowing we're doing all this discussion.  If you're in tower or a gv it may not even be noticeable unless you're looking for it.

I don't experience it in SEA2 during MNM... so this also coincides with what Steely is finding.

I tried switching back to dx9 before, it did remove the micro warps, but came with too many other gfx mehs so I went back to dx11 and just deal, knowing it'll be there.

It seems to happen around the xx:09 mark.   I'll do some more captures and see if that's a consistent pattern.

Animl's dx11 tweaks haven't affected it, but did help in some other areas.

[steely07]

Vulcan, thank you Mate, that's excellent news!, you'll be getting the stutter, but not noticing, I'll check in a sec, but I've been watching the flag or the tower clock, and you can see them stop flapping/ticking as you get the spike

Watch the clock :

I think I know where you're headed with the network card questions, and this machine (where i'm doing most of my testing) has a RealTek Semiconductor RTL8168/8111 PCI-E Gigabit Ethernet NIC, all the "green" ethernet options disabled, in driver as well as power management etc, Gaming machine is using Intel Ethernet Connection I219-V (green settings/power management disabled) and is Intel Core i7-8700, 3080Ti and 32GB ram

Home network is gigabit fibre 1000/400 going through a Ubiquiti UniFi Enterprise Gateway Router (USG-AU).

As you've got such good visibility of your network with the XG, and if I may be so bold, and if you have time, a 30 Min DX9 test would be valuable, and appreciated, it'll be the last testing I'd ask from you, promise!

I did also try some linux utilities to do UDP ping/trace (hping3 for example), but couldn't get a reply, I had the idea that maybe a dynamic table of IP's was being created in the alleged firewall, via a (totally imagined!) process of

Start AH game, and log in, some handshaking is done checking login details are a current subscriber and if so, add the IP to an "allowed IP's table" and let user into game, so I launched the game first and then tried hping3, but still no response to UDP ping, I suspect that as you say Vulcan, it's encapsulated in the game traffic and avoids the bot scans etc etc

[Animl-AW]

When it starts matters

In this concept, if its the port (on HTC or Frontier end) everyone should get them at the same time?
If it starts 10 min after you log in it could be on the player side of things?

[steely07]

When it starts matters

In this concept, if its the port (on HTC or Frontier end) everyone should get them at the same time?
If it starts 10 min after you log in it could be on the player side of things?

In my theory, there's a 600 second (10 min) timeout or reconnection based on the client's connection time, so it's going to be 10 mins after you start game, and 10 mins ongoing for individuals, not the whole arena

[Animl-AW]

In my theory, there's a 600 second (10 min) timeout or reconnection based on the client's connection time, so it's going to be 10 mins after you start game, and 10 mins ongoing for individuals, not the whole arena

So it’s the player side that’s triggering the timing?
I may consider that a flag in process of elimination to look deeper into where it starts.

Call me ignant, but if the port had issues it would be the trigger and it would be the same timing on all players in unison?

- Devil’s advocate.

[steely07]

So it’s the player side that’s triggering the timing?
I may consider that a flag in process of elimination to look deeper into where it starts.

Call me ignant, but if the port had issues it would be the trigger and it would be the same timing on all players in unison?

- Devil’s advocate.

As I've said, I'm just poking at it from the outside, but it looks like each player starts a session when they log in, session being another word for successful connection to a server, from what i've read while doing this, I'm a bit rushed this morning, so excuse the AI explanations below, but they make sense when compared to what we're seeing

1. Connection State Table (the 600-second timer)

This tracks "is this a valid session I should forward traffic to AH game server?"
Even though packets are flowing, the session's STATE expires after 600 seconds
It's like a parking permit that expires even if you're still parked

or

The UDP inspection theory:
Some load balancers do stateful packet inspection where they:

Examine each UDP packet
Maintain a session state
Expire that state after idle time OR after total time (even with traffic!)
Force renegotiation periodically

Again, guessing there is a load balancer/firewall, this based on the fact we know what ports are "open" (because we see game traffic to these ports), but we aren't able to get any response from them using UDP ping to same ports, if they were just 1 to 1 open to the internet, they'd be scannable to NMAP (https://nmap.org/zenmap/ linking to the windows port, as it's got better screenshots, but NMAP is a linux port scanner, well known, and many features to it) or other low level tools (Vulcan has mentioned quite correctly that ping etc aren't the be all and end all, but quick and dirty checks, but there should be some response if the ports aren't firewalled)

[Dadtallica]

My stutters do not happen at any type of regular cadence. It’s 99.99999% driven by what’s going on around me. Whatever one  of the last maps was had extra clouds and it caused a few hiccups.

[Vulcan]

steely07 UDP is stateless.

A firewall will maintain a session table for UDP, and that will have a timeout value. But that is different to aging out. I cannot find anything about UDP aging out. Also I would expect if a UDP session aged out you would see the AH UDP error switching to TCP message.

The clock thing does my brain in. I would not have thought the clock hands would act like an tracked object in the arena?

[Animl-AW]

My stutters do not happen at any type of regular cadence. It’s 99.99999% driven by what’s going on around me. Whatever one  of the last maps was had extra clouds and it caused a few hiccups.

Is your V-Sync Off? If so try Fast instead. Off has drawbacks, it can create lag

[Vulcan]

In my theory, there's a 600 second (10 min) timeout or reconnection based on the client's connection time, so it's going to be 10 mins after you start game, and 10 mins ongoing for individuals, not the whole arena

Yes definitely 10 min intervals starting from when the client gets into the MA.

I tried DX9, no spike. 3 ports open: UDP 5110, UDP 5131 and TCP 5111 - similar to what I saw with the DX11 client.

I think this is something in the AH3 DX11 client, not the server end.

[Shane]

As I've said, I'm just poking at it from the outside, but it looks like each player starts a session when they log in, session being another word for successful connection to a server, from what i've read while doing this, I'm a bit rushed this morning, so excuse the AI explanations below, but they make sense when compared to what we're seeing

1. Connection State Table (the 600-second timer)

This tracks "is this a valid session I should forward traffic to AH game server?"
Even though packets are flowing, the session's STATE expires after 600 seconds
It's like a parking permit that expires even if you're still parked

or

The UDP inspection theory:
Some load balancers do stateful packet inspection where they:

Examine each UDP packet
Maintain a session state
Expire that state after idle time OR after total time (even with traffic!)
Force renegotiation periodically

Again, guessing there is a load balancer/firewall, this based on the fact we know what ports are "open" (because we see game traffic to these ports), but we aren't able to get any response from them using UDP ping to same ports, if they were just 1 to 1 open to the internet, they'd be scannable to NMAP (https://nmap.org/zenmap/ linking to the windows port, as it's got better screenshots, but NMAP is a linux port scanner, well known, and many features to it) or other low level tools (Vulcan has mentioned quite correctly that ping etc aren't the be all and end all, but quick and dirty checks, but there should be some response if the ports aren't firewalled)

I did a few tests just now... it seems the 10 min cycle starts at arena entry; when I first experienced it, it reminded me of the old ISP activity checks dinging your connection.

I did tests at the xx:x 0, 3 and 5 min marks, 10 mins each time.