Effects of assymetric latency on online FPS.

[Maelstrom]

I play quake and overwatch and probably have thousands of hours in both games by now, my main symptoms in this game with regard to network latency is:

hitreg - I can miss the enemy entirely and score hits, and vice versa, I can land exactly on their heads and get noregs in cases where it doesn't make sense for the server to reject my hit (i.e. in some cases in these games, if you died first you will get the noreg, this is just latency)
movement - certain players just move faster while it seems like my movement is comparativevly slowed down

I've posted my spec elsewhere but this is on a fresh windows install, 7800x3d c states off, 240hz monitor and FTTH. I do not seem to have input lag in offline games - so in my mind this is purely an online only issue. My PC in general feels very responsive, and slow motion footage of quake and overwatch with my mouse hand in frame shows a 1 to 1 correspondence with minimal delay. (unfortunately no measurement here as I don't have the equipment, but as a quick test to surface any issues I thought it sufficed.)

This post isn't to ask for a solution, but to see whether other people with FTTH experiencing these symptoms can do a speed test.
I've noticed my connection has assymetric latency - i.e. my download is significantly quicker than my upload.

Can anyone else experiencing these symptoms clarify whether their connection has this trait?

[Puff]

Netcode in different games varies a lot. I'm not sure about Quake, but Overwatch's simply sucks.
You can try disabling coalescing in driver settings. As well as disable RSS or bind it to a specific non 0 core.
^ HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\Ndis\Parameters (RssBaseCpu dword)
If that doesn't seem to do a trick might try disabling MSI mode for the card and/or using /useplatformtick yes on top of it. This combo helped a lot on my system in CS2 and Valorant and a bit in OW.

Good luck.

[d-unknown]

I play quake and overwatch and probably have thousands of hours in both games by now, my main symptoms in this game with regard to network latency is:

hitreg - I can miss the enemy entirely and score hits, and vice versa, I can land exactly on their heads and get noregs in cases where it doesn't make sense for the server to reject my hit (i.e. in some cases in these games, if you died first you will get the noreg, this is just latency)
movement - certain players just move faster while it seems like my movement is comparativevly slowed down

I've posted my spec elsewhere but this is on a fresh windows install, 7800x3d c states off, 240hz monitor and FTTH. I do not seem to have input lag in offline games - so in my mind this is purely an online only issue. My PC in general feels very responsive, and slow motion footage of quake and overwatch with my mouse hand in frame shows a 1 to 1 correspondence with minimal delay. (unfortunately no measurement here as I don't have the equipment, but as a quick test to surface any issues I thought it sufficed.)

This post isn't to ask for a solution, but to see whether other people with FTTH experiencing these symptoms can do a speed test.
I've noticed my connection has assymetric latency - i.e. my download is significantly quicker than my upload.

Can anyone else experiencing these symptoms clarify whether their connection has this trait?

same here with 7950x3D, bad hitreg in every online fps shooter, and this is most likely an internet issue, these symptoms belongs to something called "server Desync", when you shoot first you die first, you die behind cover, your character is slow and heavy and so easy to kill , and most annoying thing is that you feel like you're 2 sec behind every one

imho it's not related with latency, it's related with the routing of your isp, you have to find the closest vpn server with wireguard protocol and try to ping your game server with pingplotter, and compare the routes the packets takes with and without vpn
it has to be different to be effective and of course, no guaranty to work everytime, but for me it's 50/50

it's not a 100% perm solution, but it can make your games more enjoyable with less hitreg issues.

as you've a ryzen 7xxx cpu , don't use DDR5 expo, use default ram speed but with the advertized timing of your kit (i believe it's 6000mhz cl30), you'il loose 10-15% fps, but you'll feel the mouse so different

[Maelstrom]

Netcode in different games varies a lot. I'm not sure about Quake, but Overwatch's simply sucks.
You can try disabling coalescing in driver settings. As well as disable RSS or bind it to a specific non 0 core.
^ HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\Ndis\Parameters (RssBaseCpu dword)
If that doesn't seem to do a trick might try disabling MSI mode for the card and/or using /useplatformtick yes on top of it. This combo helped a lot on my system in CS2 and Valorant and a bit in OW.

Good luck.

The fundamentals of netcode remains quite consistent between different games imo. mostly around client side prediction, an interp buffer, and some form of antilag/reconciliation. There are lots of parameters that can be tweaked ofc, such as ticks, packets, packet sizes, and lots of logic around specifics of each game but hit detection and movement are consistent, and so are my symptoms. (e.g. specifically feeling like you're the one guy the enemy hanzo can instagib at all times, or in quake where you're a magnet for any projectile damage).

If you've had these symptoms, are you able to check whether your connection is assymetric?

I appreciate the tips you've provided but as a long time lurker of this forum and how it handles technical info wrt this issue, I usually think they're either a very slight improvement or placebo. Mostly as the people who suggest them often can't justify them.

Are you able to help with the question in the OP? If you have these symptoms, does your latency seem symmetric or assymetric?

[Maelstrom]

same here with 7950x3D, bad hitreg in every online fps shooter, and this is most likely an internet issue, these symptoms belongs to something called "server Desync", when you shoot first you die first, you die behind cover, your character is slow and heavy and so easy to kill , and most annoying thing is that you feel like you're 2 sec behind every one

Server desync is a nice way to put it, but that's only putting a name to the issue (and one that is quite overloaded in meaning) rather than explaining what it is or how it is caused. I agree that I'm reaching the point where I don't think the issue is PC or EMI related, I've had different computers in different places and this has always occured to some degree.

imho it's not related with latency, it's related with the routing of your isp, you have to find the closest vpn server with wireguard protocol and try to ping your game server with pingplotter, and compare the routes the packets takes with and without vpn
it has to be different to be effective and of course, no guaranty to work everytime, but for me it's 50/50

what can routing affect, other than the latency of each individual packet? And all other metrics are derived from there, e.g. jitter or ordering.
I've seen people discuss this as though having a consistent 10ms ping would alleviate all issues, but here I am, with a ping that varies between 15-20 in quake live, and it feels like complete crap. I ask the question in the OP, about outgoing packets being significantly slower than incoming packets from the server, as that's the last thing that seems "undesirable" with my connection, and it happens to be one that i've seen 0 discussion of online.

it's not a 100% perm solution, but it can make your games more enjoyable with less hitreg issues.

as you've a ryzen 7xxx cpu , don't use DDR5 expo, use default ram speed but with the advertized timing of your kit (i believe it's 6000mhz cl30), you'il loose 10-15% fps, but you'll feel the mouse so different

6000mhz cl 30 is what you get with expo, I think default is cl24 with 4800Mhz. I'll give it a try but I'll let you know up front that mouse latency is quite good as it is right now, and additionally, that this doesn't answer the question in the OP.


If you have the issue just try to spot whether you have an assymetric connection in terms of latency,

[cursed-gamer]

In my opinion this is internet issue and providers doing some fishy stuff. Most likely routings from a server to client are very poor and what you see it a game trying to interpolate everything, that's why your enemies move irrationally fast.

[Maelstrom]

In my opinion this is internet issue and providers doing some fishy stuff. Most likely routings from a server to client are very poor and what you see it a game trying to interpolate everything, that's why your enemies move irrationally fast.

the problem with this is that I have a netgraph available in both games and the overall metrics from the netgraph don't seem to indicate anything fishy. Unless I'm missing something, if routing were an issue, you'd expect to see very high jitter or packet loss as the server rejects out of order packets. This seems to me another type of received wisdom that hasn't been verified properly.

[d-unknown]

Server desync is a nice way to put it, but that's only putting a name to the issue (and one that is quite overloaded in meaning) rather than explaining what it is or how it is caused. I agree that I'm reaching the point where I don't think the issue is PC or EMI related, I've had different computers in different places and this has always occured to some degree.

giving a name to an issue is a good start, and you can't be 100% sure that emi is not involved , cpu & ram sticks , fans, motherboards generate emi

what can routing affect, other than the latency of each individual packet? And all other metrics are derived from there, e.g. jitter or ordering.
I've seen people discuss this as though having a consistent 10ms ping would alleviate all issues, but here I am, with a ping that varies between 15-20 in quake live, and it feels like complete crap. I ask the question in the OP, about outgoing packets being significantly slower than incoming packets from the server, as that's the last thing that seems "undesirable" with my connection, and it happens to be one that i've seen 0 discussion of online.

15-20ms ping variaition , looks like you've to fix you're bufferbloat , check it with dslreports.com & post results
bufferbloat can cause the issues that you're describing

6000mhz cl 30 is what you get with expo, I think default is cl24 with 4800Mhz. I'll give it a try but I'll let you know up front that mouse latency is quite good as it is right now, and additionally, that this doesn't answer the question in the OP.

i didn't answer your question because it's not clear to me, you're assuming that your latency is not symetric and your download is quicker than your upload , well, can you please clarify how did you manage to compare upload latency to download latency ?

[Maelstrom]

giving a name to an issue is a good start, and you can't be 100% sure that emi is not involved , cpu & ram sticks , fans, motherboards generate emi

the problem is it's too broad of a term IMO. The reality of online gaming is that server desync is always present to some degree, all of netcode is an implementation to hide as much of the desync as possible. And the fact that while naming server desync you not only bring in network performance as a possible cause, but also the kitchen sink of EMI/PC spec, you can see why it pays to be much more specific.

15-20ms ping variaition , looks like you've to fix you're bufferbloat , check it with dslreports.com & post results
bufferbloat can cause the issues that you're describing

I already have a bufferbloat solution in place. Also, the variation is FROM a ping of 15 TO a ping of 20 i.e. around 5ms which does not strike me as anomalous.

i didn't answer your question because it's not clear to me, you're assuming that your latency is not symetric and your download is quicker than your upload , well, can you please clarify how did you manage to compare upload latency to download latency ?

Process of elimination essentially. I have a FTTH connection, the advanced netgraphs do not show anything anomalous w.r.t. ping, packet loss, or jitter in both overwatch and quake. My performance is quite steady and locally there doesn't seem to be anything wrong by checking kovaak's responsiveness.

I also happened to notice while doing some speedtests that latencies for the download were often quicker than the upload. It makes me wonder if lag-comp implementation isn't able to handle this correctly.