CSGO/CS2 observations (2 tweaks for you to try)

[Chief Blur Buster]

Latency increases is just like a defacto longer Internet wire (speed of wire), to try to equalize yourself into the playing field and bypass low-latency handicap algorithms in competitive games. You're downgrading your Internet connection to match the other peoples' connections on the game server. And you've got the extra opportunity to also remove some ping jitter, by absorbing some of the jitter in your intentionally-latency tapedelay.

But, one caveat: I need to remind people that only symmetric latency increases are scientifically kosher (e.g. not a cheat). Asymmetric latency increases is cheating and is considered grounds for post deletion. Do not talk about it here. Any latency increases you add should be equal in the both incoming/outgoing directions. Always keep latency-increase talk only in legal territory please.

Generally:

- A good gaming VPN (and fast VPN protocol like wireguard) can add intentional latency while bypassing bufferbloat on other backbones
- Or a Linux router is customizable to add intentional latency (packetdelay everything). Get familiar with those shell commands to add a tape delay for your packets!

The rest is the users' own homework; as we don't specialize in this topic, but latency-increase is a valid toolbox in better online gaming, because of how game servers handicap low-lag participants to prevent the "LPB" problem. But it sometimes creates major problems when goldilocks-latency participants get a better gaming experience than low-latency participants, and therefore that's why intentional latency-increase methods to "downgrade your Internet latency slightly" is a valid manoever to put you in the same level of playing field, in a way that is exactly like a higher-latency Internet connection.

[agstrato]

Another CS2 trick.

INCREASE YOUR LATENCY !!!! 65ms feels like a sweetspot (interp ingame is set to similar value). Just a little test, i was really annoyed. Faceit, 2 games (lvl6-8, im not a wondrous player)

First game with ping 15... skill OFF. Prefired at every corner. Killed before i even cross the edge. Uncomfortable aim. Top fraggers ping 105 and 65, what a surprise. Being a burden support throwing flashes and hoping allies hit something instead.

How did you increase your ping?

Hello i am currently in talks with Chief to see if he approves me sharing a way for users to safely increase their ping. I have recently wrote a very good tutorial on another forum and i asked for permission from Chief to see if i would be allowed to post a link to that but i am also open into copy pasting the whole tutorial here or in a separate post in order for all of you to test this. I will also like to give credit to user @joseph_from_pilsen as i actually think he is one of the guys that got me into trying Clumsy and eventually resulted into me writing that article, it's a small world guys. Be patient, i am just waiting to see what Chief decides as i respect him and this forum too much to spam unwanted links like a retard. @joseph_from_pilsen, i salute you, ur buddy from the other forum HEHE!

[joseph_from_pilsen]

Hello and welcome to blurbusters, dear agstrato redditer

I made a new discovery, since i posted here in another topic that there DO EXIST servers where the CS2 game works fine, and so far i still found only few third party servers with good input lag + responsive sprays, i made a new, probably critical discovery. For all those who write here that the game suffers from bad choppy sprays, input lag, delay, freezes, lags, packetloss... It seems to be really NOT on your side and also NOT on the side of game code itself. Because after many hours of casual playing of MM/DM i found suddenly a Valve official (!) DM server which stunned me instantly after few seconds of playing by its exceptional performance. The game was running here like a charm. I dont remember so smooth gameplay since beta release. And the server was FULL of 16 players so it was under a decent load and still performing like empty offline with bots game (at least for me the first and the last really smooth experience with cs2). All the time before i was convinced that valve server quality and game behavior cant leave the interval of "50 shades of trash".

Here is a short recording. https://www.youtube.com/watch?v=LZ9lfyToSRI Without nv shadowplay recording it was running and shooting even smoother (sadly my android phone is too big to place it into a stand to make a record from 3rd person view without nvidia shadowplay recording input lag so you wont see it). I cant remember when i was playing at valve server (at full capacity), without packetloss, without input lag, without desync, with clean netgraph. Look at the netgraph, its a straight red line, never seen at faceit ,never seen at any valve server before so clean run. Btw during the previous 30 minutes of enjoying playing here i gained a nice "fanbase" here, who desperately wanted to kick me out for "cheating" so im greeting them here nicely as i promised (you can see in video ).
Normally my game isnt running even far away that smooth, and what i listen, its general issue. I can make some recording from "normal" games but i think its pointless, as everybody knows how it commonly "behaves". For me personally common issues of common servers are - bad sprays, crosshair resistance, randomly floaty aim, inconsistent aim, packetloss, all-coloured netgraph lags, choppy movement, movement resistance, late counter step detection, oversteps...

So if you are suffering from trash performance/loss/red netgraph and you tuned your PC to exclude a problem on your side, it seems to be fully server sided/valve network sided issues. Using a simple math - a proposition is invalid, if there does exist at least one case which negates it. This is our case. There does exist at least one server where the game runs fine so we can negate many general statements - its not on your end, neither on the end of game code itself. The only issue is that the "working fine servers" are extremely rare, i found two, maybe three examples... since beta release.
Luckily, i made this discovery after a long playing period in a row without interrupting so many other statements can be also negated (described in following text).

Just for blurbusters enthusiasists to get maximum info: Account is relatively new (2y) but still has few hundred of cs2 hours, userdata folder was deleted approx. 14 days ago, disabled overlay, disabled steam cloud. Subticked shooting+movement (desubtick commands i removed 14 days ago).. Game running with high details + high shadows + ao + msaa2x + quality HDR + 4xAF, vsync+gsync+reflex on. Logitech G pro wireless x superlight, keyboard corsair K65 mini (this keyboard has sub 1ms key reaction to step fast btw), 280hz vg259QM, ryzen 7600+2x16GB 6000mhz ram + 3060Ti + gigabyte A620 x gaming cheap MoBo + budget silverstone 500W PSU + cloud II headset over USB, mic on, Internet: connected over 60Ghz last mile radio to local Fttx node. No fake latency added.
I have been running the PC without rebooting many hours in a row, i didnt restart game many hours before, I suffered few really bad laggy games before, i alt-tabbed many times, i played with high latency, low latency, almost everything i did during this evening playing session. (this is important information i think, considering many "mythbustering" threads here - the alt-tabbing hypothesis didnt have any impact, the game was able to run smooth after many hours - another myth busted: you (dont) need to reboot the game every few hours, its performance deteriorates with time the more the longer you run it - a myth obviously). Time: 11pm-1am, electricity under medium load (boiler heating water in next room), old house with old (i would say "vintage") electricity (copper wires without grounding wire (TN-C) ), but a new electric digital meter (i know, i will disappoint those who suspect digital meters from causing bad gameplay).

Sadly i got only steam network format address of the valve server, but here is networking history

Code: Select all

[#2191676068 SDR server steamid:90178616494069785 vport 0]:
[Networking]     End-to-end connection: connected
[Networking]         Remote host is in data center 'fra'
[Networking]         Current rates:
[Networking]             Sent:  64.0 pkts/sec  27.7 K/sec
[Networking]             Recv:  87.4 pkts/sec  81.5 K/sec
[Networking]             Ping:11ms    Max latency variance: 0.5ms
[Networking]             Quality:  100%  (Dropped:0.00%  WeirdSeq:0.00%)
[Networking]             Est avail bandwidth: 781.2KB/s  
[Networking]             Bytes buffered: 0
[Networking]         Lifetime stats:
[Networking]             Totals
[Networking]                 Sent:    210,992 pkts     115,034,358 bytes
[Networking]                 Recv:    381,066 pkts     326,862,106 bytes
[Networking]                 Recv w seq:    381,065 pkts
[Networking]                 Dropped   :         11 pkts   0.00%
[Networking]                 OutOfOrder:          0 pkts   0.00%
[Networking]                 OOOFixed  :          0 pkts   0.00%
[Networking]                 Duplicate :          0 pkts   0.00%
[Networking]                 SeqLurch  :          0 pkts   0.00%
[Networking]             Ping histogram: (173266 total samples)
[Networking]                      0-25    25-50    50-75   75-100  100-125  125-150  150-200  200-300     300+
[Networking]                    173264        2        0        0        0        0        0        0        0
[Networking]                    100.0%     0.0%     0.0%     0.0%     0.0%     0.0%     0.0%     0.0%     0.0%
[Networking]             Ping distribution:
[Networking]                 5th   50th   75th   95th   98th
[Networking]                11ms   11ms   11ms   11ms   12ms
[Networking]             Connection quality histogram: (659 measurement intervals)
[Networking]                 perfect    99+  97-99  95-97  90-95  75-90  50-75    <50   dead
[Networking]                     657      1      1      0      0      0      0      0      0
[Networking]                   99.7%   0.2%   0.2%   0.0%   0.0%   0.0%   0.0%   0.0%   0.0%
[Networking]             Connection quality distribution:
[Networking]               50th  25th   5th   2nd
[Networking]               100%  100%  100%  100%
[Networking]             Latency variance histogram: (359612 total measurements)
[Networking]                       <1     1-2     2-5    5-10   10-20     >20
[Networking]                   357883    1033     650      33      13       0
[Networking]                    99.5%    0.3%    0.2%    0.0%    0.0%    0.0%

Take it as an evidence that its possible to enjoy a working game even in current state, without common hassles, its definitely not an unicorn / myth, its a real scenario. Now only one detail is remaining - to make all servers to behave like this one does tonight...

[dervu]

Would be useful to connect through different VPN servers to same game server to exclude routing as issue and confirm it is server quality. Eventually, other people should confirm.
Might be also that this specific server had more resources delegated at this time and next time it won't behave the same.

[joseph_from_pilsen]

Last days official cs2 servers run extremelly badly, first day of christmas weekend (friday) it lagged in EU even at 1am CET, around 1:30am it finally got fixed ("a bit" late...), the whole evening netgraph one big mess, game running choppy, even hard lags every few dozens of seconds causing interrupts of movement and crosshair. I switched to different games, its unbearable. The biggest issue is typical for high server variance, cant shoot fast, the shooting at server is absolutelly uncontrollable + all inputs are delayed, i have sub 1ms keyboard but at server im still moving and shooting is also behind input (it shoots even a half of meter off crosshair, the server crosshair is behind too). On the other hand, some games in league of legends im "casting" an item to be activated 2 times and still not used at server so... A dilemma, to play 30 tickrate LoL where QWR + hourglass = either R without hourglass or hourglass without using ult... or laggy cs2...
edit: even reddit is complaining massively https://www.reddit.com/r/GlobalOffensiv ... ers_valve/

[joseph_from_pilsen]

Just an ordinary report.
CS2 servers run much better during working days, today it was about 50:50 to meet good/bad servers. Weekends still a disaster, much laggier, whenever i play at friday evening, its like a nightmare, its so obvious difference.... And it gets better again after weekend.
As my game runs fine at the time when server is good, i dont have much what to complain about (except the obvious flaws of the game itself which everyone suffers from), i dont suffer from any input lag feeling + when there are only low ping opponents + the server is good = then the game runs fine. Sadly, even at working days, its still a minority of games. Nothing is more annoying than a matchmaking game where it hard lags 3 times during 30 seconds warmup and i cant simply dodge the server like i would do at faceit.
And hell yeah, one important hint. I had broken keyboard... a whole fu...ng year i have played with broken K65 Corsair which double clicked. It was my friend who pointed at this issue, cuz i thought its normal and I only badly set repeat interval in windows keyboard settings, it was a malfunctional keyboard instead. So my movement was absolutelly wrong due to this fact as the keyboard overstepped me. Now i play with my old mechanical 40 USD garbo and i just realized how much false impressions i got (like that there is always an inaccuracy, in reality it was keyboard moving me). Now i ordered an analog "cheating" keyboard, im very curious to try it (I ordered the budget 129 USD drunkdeer one directly from the producer).

[joseph_from_pilsen]

Another interesting thing. For ryzen owners. The game engine is flawed in many ways, this flaw is really interesting. I tried to turn off SMT to try what will happen. And... The game is really badly optimized, there must be some elementary level mistake in game core, because the game hogs double of CPU resourcew with SMT off, but giving the same performance (or worse if it hits 100 CPU limit). FPS are capped at 260fps and game runs at 260fps stable in both scenarios.

2 scenarios, only difference is SMT ON/OFF, full DM server, screened after map ended (process lasso). With SMT OFF the game simply consumes double amount of CPU resources... Just consuming 2x more CPU for nothing.

[joseph_from_pilsen]

I tried to solve my feeling that after every lag the game desynces a LOT and client buffers too much (so im behind others as fu..k).

It wasnt only feeling, i desperatelly was looking which parameter causes this behavior and it seems i found it. Its parameter cq_buffer_bloat_msecs, in default state its set to 150, 150 fu..ing miliseconds of buffer on YOUR side which is no way compensated by server putting you up (another) 150ms behind others (you dont need even to bother with reacting then, 150ms is more than my avg. reaction time). When there are no lags, it seems that the game doesnt use the full interval, but whenever it lags for a while, it starts buffering and game puts you into very "funny" single digit kill games where everybody is prefiring you.

Some sources

https://www.reddit.com/r/GlobalOffensiv ... r_1122023/

https://www.reddit.com/r/cs2/comments/1 ... _shooting/ (last comment in comments)


Now playing with these settings (in autoexec)

sv_clockcorrection_msecs 0 - its a tradeoff, great for awpers, worse for riflers, default is 30ms, my observation: the game shows opponents a little earlier, but for a price of very unpolished movement of opponents (its harder to predict their next position - 0ms doesnt allow client to smooth out the movement based on previous 2 ticks). Thats why its great for snipers, not so great for riflers. Versus high ping opponent its even worse as they tend to move even more erratically than normally. The parameter is possible to set in client also at server (set to 0 at BOTH sides at LAN to reduce peekers advantage at LAN game!).

cl_cq_min_queue -1 - this parameter is again a trade off, you trade hitreg for reactivity of client. With 2 the hitreg is better, but there is obvious queue delay in netgraph (the netgraph will permanently show 2 green boxes line = 20ms which are obviously buffered (read: delayed) somewhere). I didnt make any deeper investigation yet, whether the delay is or is not compensated any way (i think its not, as its client sided).

cq_min_queue_size 0

cq_buffer_bloat_msecs 0

cq_buffer_bloat_msecs_max 0 (people use usually their ping for not perfect connection so 15 or 30 may be optimal)

- these all parameters disable the annoying buffering on client side. Since i disabled it, i didnt have a game with such high behind feeling. Of course, it doesnt work so well vs high pingers as the next issue is bad netcode of cs2 and lagcompensation and interp which are all set to absurdly high numbers but this is different story (spectator vs client differ so much that its common that spectator sees no shot while i emptied a half of magazine). Im gonna to dehonest valve crap by newly bought action 240fps camera to count exactly the difference server - client based on many scenarios (low ping enemy, high ping enemy etc.). By comparing server replay vs client POV off screen.

[ahead]

SV_ are server-side commands. But who cares?

[kriegsnake]

I tried to solve my feeling that after every lag the game desynces a LOT and client buffers too much (so im behind others as fu..k).

It wasnt only feeling, i desperatelly was looking which parameter causes this behavior and it seems i found it. Its parameter cq_buffer_bloat_msecs, in default state its set to 150, 150 fu..ing miliseconds of buffer on YOUR side which is no way compensated by server putting you up (another) 150ms behind others (you dont need even to bother with reacting then, 150ms is more than my avg. reaction time). When there are no lags, it seems that the game doesnt use the full interval, but whenever it lags for a while, it starts buffering and game puts you into very "funny" single digit kill games where everybody is prefiring you.

Some sources

https://www.reddit.com/r/GlobalOffensiv ... r_1122023/

https://www.reddit.com/r/cs2/comments/1 ... _shooting/ (last comment in comments)


Now playing with these settings (in autoexec)

sv_clockcorrection_msecs 0 - its a tradeoff, great for awpers, worse for riflers, default is 30ms, my observation: the game shows opponents a little earlier, but for a price of very unpolished movement of opponents (its harder to predict their next position - 0ms doesnt allow client to smooth out the movement based on previous 2 ticks). Thats why its great for snipers, not so great for riflers. Versus high ping opponent its even worse as they tend to move even more erratically than normally. The parameter is possible to set in client also at server (set to 0 at BOTH sides at LAN to reduce peekers advantage at LAN game!).

cl_cq_min_queue -1 - this parameter is again a trade off, you trade hitreg for reactivity of client. With 2 the hitreg is better, but there is obvious queue delay in netgraph (the netgraph will permanently show 2 green boxes line = 20ms which are obviously buffered (read: delayed) somewhere). I didnt make any deeper investigation yet, whether the delay is or is not compensated any way (i think its not, as its client sided).

cq_min_queue_size 0

cq_buffer_bloat_msecs 0

cq_buffer_bloat_msecs_max 0 (people use usually their ping for not perfect connection so 15 or 30 may be optimal)

- these all parameters disable the annoying buffering on client side. Since i disabled it, i didnt have a game with such high behind feeling. Of course, it doesnt work so well vs high pingers as the next issue is bad netcode of cs2 and lagcompensation and interp which are all set to absurdly high numbers but this is different story (spectator vs client differ so much that its common that spectator sees no shot while i emptied a half of magazine). Im gonna to dehonest valve crap by newly bought action 240fps camera to count exactly the difference server - client based on many scenarios (low ping enemy, high ping enemy etc.). By comparing server replay vs client POV off screen.

Well, you are right at some point. Most of all I am playing on faceit servers , but I've tried about 40dms in 2 days, and all of them was bad for me, like 2% packet loss and stuff like that, today I've come across the server where I've been kicked in dm because each of my shot has been way too accurate (i had 0.5-1% pl ), even though I've plaid with settings you wrote above, sv_clockcorrection and cq_buffer_bloat does nothing in my case , but there is some kind of a differnce when I've tried different numbers on cl_cq_min_queue.

My conclusion is the same as in valorant, server deciding whose shots will be accurate and whoes not. You always come across the person who carrying his team with 30+ frags while his teammates below 10 kills.

And here is the screenshot of dm that I've been talking about. Please consider that I've been running across the map, not sleeping in lower tunnels or something like that.