Back

Network Guide

CS2 Sub-Tick & Network Buffer Guide: Telemetry & Hit Reg

Why you die behind walls in CS2 and which parts of it you can actually control. Tune cl_net_buffer_ticks (None vs 1 vs 2 packets), read the Telemetry HUD (tick miss rate, jitter, ping), and clean up your side of the connection.

D
Daniel P. - Hardware expert
  • 10 min read
Share

How CS2 Sub-Tick Actually Works (And Why You Die Behind Cover)

How CS2 Sub-Tick Actually Works (And Why You Die Behind Cover)

The move from CS:GO's discrete 64-tick and 128-tick networking to Counter-Strike 2's sub-tick architecture changed how client actions reach the server. Understanding this pipeline explains why shots sometimes feel desynchronized and why players regularly "die behind cover."

In traditional tick-based engines (CS:GO, Valorant), the server advances the simulation in fixed steps. On a 64-tick server the world updates every 15.625 ms; on a 128-tick server, every 7.8125 ms. In CS:GO, an action made midway through a tick interval was effectively treated as happening at the tick boundary.

CS2 on Valve's official servers still runs a 64-tick simulation (15.625 ms per tick), but player actions such as shooting, jumping and movement key presses are stamped with an exact sub-tick time. The server applies the action at that moment inside the tick and uses lag compensation to rewind other players' hitboxes to the positions the shooter was seeing. Sub-tick fixes input timing. It does not make the server send you world updates more often, so it is not the same thing as a 128-tick server.

The Mechanics of Dying Behind Cover (illustrative, simplified numbers that ignore interpolation, buffering and tick quantization): 1. Player A (the peeker) sees Player B in the open and fires at T0. A's client stamps the shot with its exact sub-tick time. 2. Player B has already started retreating. On B's own screen, B is fully behind the wall at T0 + 45 ms. 3. A's shot reaches the server about 25 ms after T0 (A's one-way latency). The server rewinds B's hitbox to where A was seeing B when A fired (still in the open) and validates the hit with lag compensation. 4. The server confirms the kill and sends it to both clients. 5. B's one-way latency is about 45 ms, so the kill reaches B at about T0 + 25 + 45 = T0 + 70 ms. B has been behind cover on their own screen for roughly 25 ms, so the death looks like it happened through the wall.

In practice, that gap is roughly the sum of both players' one-way latencies, plus interpolation and buffering delay, plus tick quantization. Lag compensation only rewinds up to a limit (typically around 200 ms), so extreme ping is capped rather than rewarded.

Your buffering setting is part of this equation. According to the console description, cl_net_buffer_ticks adds ticks of delay to server snapshots and to your user commands. A higher value means you see enemies later, and your own movement (including ducking back behind cover) reaches the server later, so the server learns that you are safe later. Valve's November 12, 2024 update also fixed a mid-spray case where lag compensation was not aware of the player's buffering setting.

What you can control: your own ping, jitter, packet loss and buffer setting. What you cannot control: your opponent's connection and the server's lag compensation design. The practical goal is to keep your side clean and the buffer as low as your connection allows.

Buffering to Smooth Over Packet Loss: None (0) vs 1 vs 2 Packets

Counter-Strike 2 exposes this option in Settings > Game as 'Buffering to smooth over packet loss / jitter' (console command cl_net_buffer_ticks, range 0 to 2). Each step is one 64-tick interval (15.625 ms). Per the console description, it controls the delay applied to server snapshots and user commands, and it indirectly sets cl_interp_ratio, which should not be changed directly. 'None' does not mean zero buffering: Valve notes that the Source 2 engine already adds some buffering automatically to smooth over jitter. You cannot set cl_interp directly in CS2, so old CS:GO interp configs do not apply.

None (cl_net_buffer_ticks 0)

Lowest Latency (Default)

• Added delay: no extra buffer on top of the engine's own automatic jitter smoothing.

• How it operates: snapshots and user commands are not held back by an additional tick.

• Advantage: the lowest added latency and the smallest peeker's disadvantage this setting can give you.

• Compromise: there is no spare tick in reserve, so a late or dropped snapshot is more likely to become a missed tick. This shows up as choppy enemy movement, short teleports, and delayed hit registration.

• Recommended for: stable wired connections where the Network Problems readout stays at or near 0% over several full matches.

Requirements for 0 Packets:

  • Wired Cat5e / Cat6 Ethernet (no Wi-Fi)
  • Network Problems (tick miss rate) at or near 0% across several full matches on official servers
  • No recurring spikes in the Network Jitter graph
  • Low bufferbloat (A or A+ grade on Waveform)

1 Packet (cl_net_buffer_ticks 1)

Common Stability Pick

• Added delay: one extra tick (15.625 ms) of delay for snapshots and user commands.

• How it operates: the client keeps one tick of reserve, so a single late or dropped packet can be bridged instead of becoming a missed tick.

• Advantage: reduces micro-stutter and short teleports on connections with occasional jitter, powerline adapters, or mild congestion.

• Compromise: enemies appear on your screen about one tick later, which widens the peeker's advantage against you, and your own inputs may also reach the server later.

• Recommended for: standard consumer broadband, shared household internet, or any connection that keeps showing tick misses with the setting at None.

When to Select 1 Packet:

  • Recurring Network Problems warnings at None after wired and router fixes
  • Unavoidable Wi-Fi connection with airtime contention
  • ISP routing with transient jitter bursts
  • Enemy models micro-teleporting during dry peeks

2 Packets (cl_net_buffer_ticks 2)

Emergency Stability Only

• Added delay: two extra ticks (31.25 ms) of delay for snapshots and user commands.

• How it operates: the client keeps a two-tick reserve.

• Advantage: the strongest smoothing the menu offers. It masks heavier packet loss and jitter bursts.

• Compromise: enemies are shown to you about 31 ms later than at None, and because user commands are delayed as well, the real cost in a duel can be larger than that. Enemies holding angles will see you before you see them, so this is a poor fit for ranked play.

• Recommended for: cellular hotspots (4G/5G), congested public Wi-Fi, or badly compromised routing, as a temporary workaround while you fix the root cause.

When 2 Packets Is Necessary:

  • Cellular tethering or satellite links
  • Sustained tick misses that persist at 1 packet
  • Unplayable rubberbanding on None and 1 Packet

How to Read the CS2 Telemetry HUD: Tick Miss Rate, Jitter & Ping

The Telemetry HUD is CS2's replacement for the old CS:GO net_graph. You enable it under Settings > Game > Telemetry. Its network readout has changed over time: the Armory Update (October 2, 2024) added graphs for network jitter and misdelivery and changed how packet misdelivery was measured, and many players saw a sudden jump in "loss" even though the netcode itself had not changed. The November 12, 2024 update changed the readout to measure how often a tick is missed because of network loss or jitter, and added a detailed option that shows the raw packet loss and jitter numbers. Valve keeps adjusting this area, so check the latest release notes if a readout behaves differently from what is described here.

Configuring the Telemetry HUD: Use Settings > Game > Telemetry, or the developer console: • cl_hud_telemetry_frametime_show 2 (frame time and FPS) • cl_hud_telemetry_ping_show 2 (ping) • cl_hud_telemetry_net_misdelivery_show 2 (network problems) • cl_hud_telemetry_serverrecvmargin_graph_show 2 (server receive margin graph) For the show commands, 0 means never, 1 means only when conditions are poor, and 2 means always. The detailed network quality option (raw packet loss and jitter numbers) and the Network Jitter graph are in the same Telemetry menu. The level at which the network readout is flagged as poor is set by cl_hud_telemetry_net_misdelivery_poor (accepted range 0 to 100).

Interpreting the Telemetry Metrics: 1. Frame Time (ms): client rendering time per frame. At 240 FPS the average is 4.17 ms. If frame time spikes while ping and network problems stay flat, the stutter is local (CPU/GPU load, background apps, thermal throttling), not your internet. 2. Ping (ms): round-trip time to the game server. On official servers this goes through Valve's Steam Datagram Relay (SDR) network. A stable value matters more than a low one. 3. Network Problems (tick miss rate): this is measured in two directions. Downstream is how often your client needs the next snapshot and it has not arrived. Upstream is how often the server needs your next user command and it has not arrived. It only counts network events that cause a missed tick, so it is not a pure packet loss number, and 0% does not prove the network is perfect. A low warning threshold (for example 0.5%) is a sensible starting point, so that a hitch can be traced to the network instead of your PC. 4. Packet Loss (%): packets that never arrived. Shown in the detailed display. Even small sustained loss will cause problems, because the server cannot act on a user command that never arrived. 5. Jitter (ms): variation in packet arrival times. Source 2 buffers automatically to absorb some jitter, so jitter only hurts once it exceeds what the buffer can cover. At that point it appears as missed ticks. This is exactly what cl_net_buffer_ticks adds margin for. 6. Network Jitter graph and Server Receive Margin: the jitter graph shows packet delivery anomalies over time, and the receive margin graph shows how early or late your commands reach the server relative to when it needs them. Repeated bad spikes mean the problem is on your line, router, or ISP route.

Readouts are highlighted when they cross a "poor" threshold, and you can adjust those thresholds in the Telemetry menu. Use the readings to decide where to look: frame time spikes point to your PC, ping and jitter spikes point to your network or ISP route, and sustained packet loss points to a physical or routing problem.

CS2 Sub-Tick vs Valorant 128-Tick: Structural Network Comparison

Many competitive players find that hit registration and angle holding feel crisper in Valorant than in CS2, despite both games demanding millisecond aim precision. Several structural differences contribute.

1. Server Tick Simulation Rate: • CS2 on Valve's official servers runs at 64 ticks (15.625 ms interval) with sub-tick input timestamps. The server simulates and sends world state 64 times per second. • Valorant runs 128-tick servers (7.8125 ms interval), so world state is simulated and sent twice as often. Impact: sub-tick removes the input timing error of a low tick rate, but it does not change how often snapshots reach you. With 128 ticks, new positions arrive in smaller steps and each buffering step costs less time. Lag compensation depends on latency and interpolation delay, not directly on tick rate, so a higher tick rate does not by itself shrink the rewind window, but it does reduce the tick quantization part of the delay.

2. Routing Infrastructure (SDR vs Riot Direct): • CS2 official matches route through Valve's Steam Datagram Relay (SDR). SDR protects game servers from DDoS attacks by ingesting traffic at edge relays, but your packets still cross the public internet between your home and the relay, so stability depends on your ISP's peering. • Valorant uses Riot Direct, Riot's own network of peering agreements and routes with ISPs, designed to shorten and stabilize the path between players and Riot's servers. How much it helps depends on your region and ISP.

3. In-Game Network Buffering Settings: • CS2 offers None, 1 Packet and 2 Packets. Each step is one tick (15.625 ms), and the delay applies to server snapshots and your user commands. • Valorant offers Minimum, Moderate and Maximum. A Riot developer described them when the setting was introduced: Minimum is 1 frame of delay (7.8125 ms) with your movement data sent to the server at the full 128 per second; Moderate is 3 frames (23.4375 ms) with your movement data sent to the server in chunks at 64 per second; Maximum is 5 frames (39.0625 ms) with your movement data sent in chunks at 32 per second. Note that the 64 and 32 figures describe how often your client sends its data to the server, not how often you receive world updates. These numbers date from the setting's introduction, so check the current in-game tooltip, since Riot can change them. • In both games the lowest setting gives the least delay, and stepping up is a legitimate trade-off when your connection cannot keep the lowest setting stable.

Step-by-Step Network Tuning Pipeline for Competitive CS2

Follow this sequence to find out whether your connection can run with no extra buffering, set up telemetry, and fix problems at the source before you add buffer.

  1. 1Verify Wired Infrastructure & Baseline Jitter

    Disconnect Wi-Fi and connect your gaming PC with a Cat5e or Cat6 Ethernet cable directly to your router. Wi-Fi adds radio contention, retransmissions and airtime slicing that inflate jitter. If you have no choice but Wi-Fi, use the 5 GHz or 6 GHz band and stay close to the router with a clear line of sight.

    Test the local link first: open Command Prompt and run ping -n 100 followed by your router's address (often 192.168.0.1 or 192.168.1.1). Windows reports minimum, maximum and average rather than variance, so look at the gap between minimum and maximum. You want 0% loss and a very small gap. Then repeat the test against a public host such as 1.1.1.1 to include your ISP's leg. A clean router test with a bad public test points to the ISP, and a bad router test points to your local setup.

    Keep in mind that a public host is only a rough proxy: the route to Valve's relay can differ, so the in-game Telemetry HUD remains the real test of the path CS2 uses.

    Powerline adapters (HomePlug) can drop packets when household appliances cycle on. If a cable run is impossible, MoCA (over coaxial cable) is usually more stable than Powerline or Wi-Fi.
  2. 2Enable the Telemetry HUD in CS2

    Go to Settings > Game > Telemetry and set frame time/FPS, ping and network problems to Always. Turn on the detailed network quality display and the Network Jitter graph, and set the network warning threshold low (for example 0.5%).

    Or use the developer console:

    cl_hud_telemetry_frametime_show 2

    cl_hud_telemetry_ping_show 2

    cl_hud_telemetry_net_misdelivery_show 2

    cl_hud_telemetry_serverrecvmargin_graph_show 2

  3. 3Dial in cl_net_buffer_ticks for Your Connection

    Go to Settings > Game > Buffering to smooth over packet loss / jitter and start with 'None' (cl_net_buffer_ticks 0).

    Play several full matches on official servers in the same region while watching the telemetry. If Network Problems stays at or near 0% and the jitter graph is quiet, keep 'None'. If you see recurring tick misses and enemy models stutter or teleport, switch to '1 Packet' (cl_net_buffer_ticks 1) and compare again over the same number of matches. Use '2 Packets' only if misses persist at 1, and treat it as a temporary workaround. After you fix a network problem (cable, router, ISP route), test 'None' again.

  4. 4Check the Bandwidth Setting

    In Settings > Game, make sure 'Max Acceptable Game Traffic Bandwidth' is set to 'Unrestricted'. This setting corresponds to the rate console command, and the Unrestricted value (and the default) is 786432:

    rate 786432

    There is no higher valid value to chase, and a lowered cap can delay updates during busy moments, so the goal is simply to make sure it has not been reduced (for example by an old autoexec).

  5. 5Reduce Bufferbloat and Competing Traffic

    Run a bufferbloat test (for example Waveform) while the connection is busy. If latency climbs sharply under load, enable smart queue management (SQM, Cake or fq_codel) on your router if it supports it, or cap upload and download speed slightly below your real line speed. Pause large downloads, cloud backups and streaming on other devices while you play, and limit Steam's own download bandwidth or schedule game updates outside play time.

    Also check the ping to the server region you are matched with. Playing on a distant region adds latency that no buffer setting can remove, and a VPN or a poorly peered ISP route can add jitter, so test with the VPN off.

  6. 6Optional: Windows Network Adapter (NIC) Settings

    These tweaks are experiments, not guaranteed fixes. Open Device Manager > Network Adapters > right-click your Ethernet adapter > Properties > Advanced. Option names differ between Intel and Realtek, so change one setting at a time and re-test with the Telemetry HUD:

    • Energy Efficient Ethernet (EEE) / Green Ethernet / Power Saving: Disabled. These can put the link into a low-power state between bursts of traffic.

    • Flow Control: Disabled if you suspect pause frames from your router are causing stalls.

    • Interrupt Moderation: leave at the driver default unless you have measured an improvement.

    Update the adapter driver from the chip vendor or your motherboard maker, and replace the Ethernet cable if loss appears only on that link.

Frequently Asked Questions

  • It gives the lowest latency the menu allows, provided your connection is clean. Setting cl_net_buffer_ticks 0 avoids adding an extra tick (15.625 ms) of delay to server snapshots and your user commands, so enemies appear on your screen sooner. If your connection regularly misses ticks at 0, the stutter and delayed hit registration usually cost more than the saved latency, and 1 Packet is the better setting.

  • Dying behind walls mostly comes from lag compensation and your opponent's connection, not only your own. If an opponent peeks you with 60 ms ping, their client sees you in the open before you finish ducking behind cover. The server honors their shot at the sub-tick timestamp, rewinds your hitbox to what they saw, and confirms the kill. Buffering adds to the delay on your side, so keeping jitter low and the buffer as low as your connection allows helps, but a low ping cannot remove the effect completely.

  • It is the tick miss rate: how often the client or the server was ready for the next tick but the data it needed had not arrived. Downstream misses are snapshots arriving late or not at all, and upstream misses are your user commands arriving late or not at all. It only counts network events that cause a missed tick, so it is not a pure packet loss figure. Occasional small values are normal. Persistent values mean the problem should be fixed or the buffer raised.

  • Start with a wired Ethernet connection and a quick ping test to your router to rule out your local link. Then check bufferbloat under load, close background traffic, and test cl_net_buffer_ticks 1 to let the engine bridge late packets. If the local link is clean and the problem appears only in CS2, the cause may be the route to the server, so test at different times of day and consider contacting your ISP with your telemetry readings.

  • On Valve's official servers, no. The simulation runs at 64 ticks, and sub-tick timestamps make the timing of your actions precise within each tick. Sub-tick is not a 128-tick server: snapshots still arrive every 15.625 ms. Servers run by third-party platforms can be configured differently, so check the platform you play on.

  • Valorant runs 128-tick servers, so world updates arrive twice as often as CS2's 64-tick simulation, and Riot's own routing network can give more consistent paths in many regions. CS2 uses sub-tick timestamps to fix input timing, but snapshots still arrive every 15.625 ms and packets travel over the public internet to Valve's relays. These differences affect how the game feels, but they are not the only factor and vary by region and ISP.

  • Use 786432. That is the default and corresponds to 'Unrestricted' under Max Acceptable Game Traffic Bandwidth. There is no benefit in setting a higher number. Just make sure the value has not been lowered, for example by an old config.

  • No, not as a tuning tool. You cannot set cl_interp directly in CS2, and cl_interp_ratio is managed indirectly by cl_net_buffer_ticks, which you should change instead. The old CS:GO interp config bundles do not give you a hidden advantage in CS2.

You May Also Like