PC optimization for competitive esports.
Meech's two cents on the state of modern PC hardware and how to optimize it for frame-time delivery or any use case.
You don't need to read this guide before a consult; it's just background to help you understand what we do and why. Read it end-to-end or skim the sections below.
Contents
PC Optimization: Core Principles
Frame time consistency is our priority.
A popular consensus seems to be that PC tuning is simply about "min-maxing": pushing your CPU and GPU to the highest possible clock speeds, tightening RAM timings to the absolute limits, or chasing the exact undervolt without properly testing for stability under varied loads, and just going as far as possible until the PC is on the verge of stability, and this is how you achieve the "most frames" in your title of choice.
Performance Consistency & Frame Times
To understand how to optimize your PC to deliver the best frame-times in games, you have to understand how your PC actually delivers a game to your monitor. Your Average FPS reading is just a running average of Frame Times: the actual time (in milliseconds) it takes your system to generate each individual frame.
Below, two PCs each deliver a sequence of frames. Each frame-time, the time in milliseconds it took to generate that frame, is shown on the arrows between them, and both still average out to 240 frames per second.
Ideally, every frame is delivered close to that average frame-time. Your PC will never deliver frames with perfect consistency, but the goal is to keep any single frame from taking much longer than the average frame-time to deliver.
Consistent frame-time delivery is as important as, if not more important than, your average frame-time or average FPS. This is precisely why we prioritize achieving better frame-time stability for our clients in addition to achieving a higher average frame-time.
Average FPS in benchmarks doesn't give the full picture of how well your game is running. This introduces the concept of 1% Lows. This metric represents the average of the slowest 1% of frames generated in a specific time interval.
When benchmarking a game, a large gap—or “Delta”—between your average FPS and your 1% lows means your frame-time delivery is inconsistent. This variance typically indicates one or more underlying system bottlenecks:
- Hardware Throttling & Instability: Your components are aggressively downclocking or fluctuating under load rather than sustaining a flat, stable clock rate.
- OS Interruptions: Background Windows processes, telemetry, and unoptimized services are actively conflicting with your game engine's rendering pipeline.
- Misconfigured System Settings: Windows is not configured to properly utilize your hardware, often due to restrictive default power management plans and sub-optimal resource scheduling.
Left unoptimized, these bottlenecks directly disrupt your gameplay, manifesting mid-match as:
- Micro-stutters and frame hitches that break visual fluidity, especially during intense, rapid asset rendering.
- Inconsistent input feel that actively compromises your muscle memory.
- System instability and crashes.
Endgame: Stability & Lower Latency
Inconsistent frame times that average to a higher FPS is far worse and less reliable for competitive gameplay than running a lower average FPS with much better frame-time consistency.
While our optimizations will almost certainly increase your average frame rate, we focus much more heavily on Frame Time Stability and ensuring the gap between your 1% Lows and Average FPS is as narrow as possible.
We prioritize three specific pillars:
- Frame Time Consistency: narrowing the frame-time Delta, or simply raising your 1% and 0.1% lows, to promote frame time consistency: a smooth experience where the game feels the same regardless of your location on the map or the particular game state.
- Hardware Predictability: we ensure your components maintain a consistent clock rate under load. By eliminating thermal throttling and aggressive downclocking, we ensure your hardware never fluctuates mid-fight. Your system should respond identically every single time you move your mouse.
- Input Latency (End-to-End): stripping away "wake-up" latencies and OS interruptions to ensure your mouse and keyboard inputs are processed as fast as possible.
Hardware Tuning: The Core Concepts
When you hear "Hardware Tuning," you likely think of two main concepts: Overclocking and Undervolting. In practice, they're often the same exercise: we're asking your processor to hold a specific clock speed at a specific voltage that's a higher clock and/or a lower voltage than what the processor would request in its stock configuration.
Overclocking vs. Undervolting
- Overclocking: the process of pushing your hardware's clock speeds past the stock factory configuration.
- Undervolting: the process of lowering the voltage a component (CPU or GPU) requests to hit its clock speed.
Most modern hardware comes "overvolted" by default. Manufacturers program components to request significantly more voltage than they actually need as a safety margin. By undervolting, we reduce heat output by tuning components to request less voltage, which creates thermal headroom. This lets your hardware maintain peak frequency at a lower voltage and avoids the throttling (component downclocking) that happens when it hits a thermal, voltage, or power limit.
The Silicon Lottery
No two chips are identical. Due to random variations during manufacturing, some chips are more efficient than others. Your 9800X3D might hit max frequency at notably lower or higher voltage than your friend's. We tune specifically for your unique piece of silicon.
This is fundamentally why we can't guarantee a specific stable overclock or undervolt on your hardware. Even if a given undervolt is stable on 90% of samples of your CPU or GPU model, your specific unit might fall in the small subset that isn't stable there.
The reverse is just as true: your sample could be a golden bin that holds a much deeper stable undervolt than average, or it could boost higher at a given voltage. This randomness in tuning potential from chip to chip is what we call the silicon lottery.
That said, it's worth keeping in perspective: even though some chips boost higher or need less voltage than others, the actual in-game rendering difference between the best- and worst-binned samples is often negligible once both are properly tuned.
Thermal Limits
Every component — the CPU, GPU, and RAM — operates within a defined voltage and temperature range, and modern hardware actively manages itself to stay inside it. AM5 X3D chips begin to thermal throttle once they hit 60–65 °C: their boost algorithm trades clock speed for heat, so a chip running hotter holds lower sustained clocks well below its ~90 °C hard limit. The same logic applies to your GPU and motherboard VRMs.
This is why having a proper cooling solution for your CPU, along with proper airflow and ventilation throughout your system, is so important. We make sure the CPU has a sufficient cooler and that case ventilation keeps the graphics card, VRMs, and memory comfortably inside their thermal envelopes — so your hardware holds its clocks instead of quietly throttling mid-match.
Our Approach
How do we leverage tuning to improve frame time stability and latency? This is where we avoid the typical "min-maxing" seen in generic guides, which suggest pushing clocks as high as possible or lowering voltage to the absolute floor and running a narrow stress test to confirm stability. In isolation, doing either of these puts your system on the verge of instability. This leads to frequency fluctuations, micro-stutters, and even system crashes.
Our protocol is a balancing act:
- Perpetual Clocks: we want your components to run at the highest clock speed they can perpetually maintain under load without throttling.
- Sufficient Voltage: we provide sufficient (but not excessive) voltage. This ensures the component has the breathing room to handle transient spikes in game demand without generating unnecessary heat.
- Stability Over Peak: a system that is "mostly stable" is a liability. We prioritize a system that is 100% stable, ensuring that your mouse feel and frame delivery remain identical from the first minute of the match to the last.
Stability/Stress Testing
When we make changes to the underlying clock speeds, voltage curves, or other hardware-level changes to your components, we have to ensure they stay stable under varied loads, including max load. To do this, we use specialized programs called stress tests, which push your components under extreme, synthetic loads they'll likely never encounter in normal use. If they hold up under that worst case, you'll have assurance that they stay stable under real-world usage demands.
Stability and stress testing require pushing hardware through specific, high-intensity conditions to simulate and exceed the absolute worst-case demands of modern gaming engines:
- CPU (OCCT): tests the processor using both steady and variable (shifting) workloads across all cores. By rapidly cycling thread utilization up and down, it checks how the hardware handles sudden voltage and power-state changes. This ensures that custom all-core clocks and undervolt offsets stay completely stable as the CPU scales up or down during a match.
- GPU (FurMark 2 & OCCT): tests maximum thermal loads alongside aggressive power-state transitions. This validates your custom voltage/frequency curve against sudden electrical spikes and rapid load shifts, such as transitioning instantly from a low-stress loading menu into a heavily populated, render-intensive scene.
- RAM & IMC (TestMem5 & Y-Cruncher): tests heavy data cycling, memory controller voltage boundaries, and temperature endurance over time. This isolates microscopic timing errors and voltage drops that directly cause frame-time hitches, micro-stutters, or sudden game client crashes during extended competitive sessions.
Hardware Overview: Component Roles
To optimize a system, you must understand the relationship between your components. Your CPU, GPU, and RAM are all interfaced through your motherboard, and together that hardware configuration ultimately determines your final latency and frame time stability on the hardware side.
The Pipeline: CPU vs. GPU
Frame delivery is a two-stage process:
- CPU (Preparation): the CPU calculates game logic, physics, and player positions. It generates "draw calls" (instructions telling the GPU what to render).
- GPU (Rendering): the GPU receives these instructions and renders the final frame.
Hardware Bottlenecks:
At 1080p or 1440p on Low settings, modern GPUs are often capable of rendering frames faster than the CPU can produce draw calls.
- CPU Bound: if your CPU cannot prepare frames fast enough to keep up with the GPU, your CPU is the bottleneck. This is the primary limiting factor in high-refresh esports for most modern builds.
- GPU Bound: if the CPU provides data faster than the GPU can render it, the GPU becomes the bottleneck.
The CPU: AMD & Intel
AMD (Socket AM5)
Common CPUs include the Ryzen 7 9800X3D and Ryzen 7 7800X3D.
- 3D V-Cache: AMD's Socket AM5 X3D models stack a large amount of last-level (L3) cache directly on the CPU silicon: extremely fast local memory right next to the cores. This makes X3D chips far more forgiving of loose RAM timings and lower bandwidth, since the CPU can pull data from its own cache far more often than from system memory.
- Tuning Strategy: we utilize Precision Boost Overdrive (PBO) and Curve Optimizer to find the most efficient voltage for your specific chip. This reduces thermal output and allows the CPU to maintain higher sustained boost clocks without thermal throttling.
Intel (LGA 1700 & LGA 1851)
Common CPUs include the Core i9-14900K (LGA 1700) and the Core Ultra 9 285K (LGA 1851).
- Thermal Management: Intel flagship chips are known to pull significant power and run at high temperatures under stock settings. Note: for LGA 1700 chips (13th & 14th gen), we highly recommend a contact frame paired with at least a 360 mm AIO to flatten the IHS and keep temperatures in check.
- Tuning Strategy: we mitigate this by undervolting and disabling per-core boost clocks. This prevents aggressive frequency fluctuations and ensures the CPU operates within a stable thermal envelope while maintaining consistent clock speeds across all cores.
The GPU: NVIDIA & AMD
While competitive esports titles are heavily CPU-bound, the graphics card still requires tuning to render consistently. The core goals are identical across both hardware vendors: eliminate aggressive clock downclocking and tightly manage operating thermals.
Forcing the maximum performance state locks the GPU into its highest power deployment profile. This prevents the core from aggressively downclocking during low-utilization scenes, which is critical in CPU-bound titles where the card frequently waits on the processor to prepare the next frame. Forcing this state ensures that when a sudden graphic-intensive load spike occurs, the hardware responds instantly instead of catching on a power-state transition, eliminating preventable frame-time hitches.
NVIDIA (GeForce)
- Voltage/Frequency Curve: We optimize the V/F curve to find the lowest stable voltage required to sustain the card's maximum boost frequency. Stripping away this excessive stock voltage sharply decreases power draw and operating temperatures, allowing the core to hold flat, consistent clock steps without hitting power or thermal limits.
- Memory Clock Offset: We apply a positive offset to the VRAM to expand total bandwidth. This directly reduces memory controller processing stalls, stabilizing frame delivery during sudden, heavy alpha effects or rapid screen transitions.
AMD (Radeon)
- Core Undervolting & Frequency Caps: Mirroring the NVIDIA tuning approach, we scale down the core voltage requested at the card's target boost frequency to reduce thermal output. To combat aggressive downclocking in CPU-bound scenarios, we narrow the gap between the minimum and maximum allowable frequencies, forcing the GPU to maintain a stable clock rate.
- VRAM Frequency & Memory Timings: Complementary to the NVIDIA memory offset, we increase the maximum memory frequency to widen bandwidth and enable optimized timing profiles to tighten internal memory latencies. This directly counteracts processing stalls in the rendering pipeline.
The RAM: DDR5
RAM tuning is the most underrated aspect of PC optimization. Most memory kits come with an XMP (Intel) or EXPO (AMD) profile: a one-click setting that applies default safe timings. Manual tuning leads to considerable gains in 1% Lows.
- AMD (AM5): how fast we can run the RAM is limited by keeping it coupled 1:1 with the Memory Controller clock (UCLK). The 3D V-Cache buffer helps mask AM5's Memory Controller limitations and latency.
- Intel: performance scales significantly with higher memory clock speeds. On modern Intel platforms, we prioritize pushing the memory clock speed to maximize data throughput.
By tightening memory timings and maximizing bandwidth, we reduce CPU stalls — the cycles the CPU spends waiting on data from RAM — which narrows the Delta between your average FPS and your 1% Lows.
How we approach memory tuning, and how much effort is worth putting into it, depends heavily on your CPU platform: as noted above, the gains scale very differently on AMD and Intel. AMD's X3D chips are also far less sensitive to memory tuning, since their stacked cache already masks much of the latency that memory tuning would otherwise address.
Windows-Level Optimization
While the bulk of our gains come from hardware tuning, we still need to ensure the OS underneath isn't working against them. Even well-tuned hardware is held back if Windows is wasting cycles on telemetry, background services, and aggressive power-saving, or letting stray processes interrupt your game's threads.
The Windows pass strips the system down to what's actually serving your gameplay, and makes sure nothing is interrupting the game or stealing CPU cycles from it.
To set expectations: guides promising massive FPS gains from Windows tweaks alone, with no hardware tuning, are generally snake oil. Besides GPU tuning, OS-level optimization doesn't create performance; it removes interference. Beyond keeping the game's threads uninterrupted, that means stopping background processes from competing for CPU time and cutting latency introduced by power-management states and device wake-up behavior, so the hardware we've tuned can perform consistently.
- Clean Install: For most consults we recommend starting from a fresh Windows install, eliminating OEM bloatware and the accumulated registry cruft that can quietly cause stutters.
- Debloat & Telemetry: We strip out Windows telemetry, advertising, and the long tail of background services.
- Power Plan: We move you onto a power plan tuned for your specific platform and CPU architecture, because the defaults aggressively throttle clocks for power savings.
- Drivers: Update to latest motherboard, chipset, and GPU drivers.
- Scheduling & Process Priority: We audit running processes for anything known to interrupt the game process and reconfigure or disable the offenders. We also adjust process priority and core affinity so your game runs at the highest CPU priority and is pinned to the right CCD in multi-chiplet AMD CPUs.
Client Intake: Diagnostic & Baseline
Before we begin, we need a complete map of your system's current state: your exact specs, along with the parts we wouldn't otherwise know about, like your case, fan configuration, and CPU cooler. This confirms your hardware is healthy and that there are no massive hardware bottlenecks or cooling insufficiencies.
1. Submit Your Intake
Everything starts with the intake form on our site. It takes a couple of minutes and includes:
- HWiNFO64 summary: launch HWiNFO64, select "Summary-only" mode from the startup dropdown, and screenshot the System Summary window (it shows your CPU, GPU, RAM, and motherboard BIOS version).
- Internal photo: a clear photo of the inside of your PC so we can verify fan orientation (intake vs. exhaust) and CPU cooler mounting.
Meech or a staff member will contact you after reviewing your form, flag any cooling issues or hardware bottlenecks, share an initial diagnostic and plan, and schedule your live session.
2. Required Software
Download these industry-standard tools ahead of your session. We use them specifically because they have the lowest "observer effect" — they don't lag the game while they measure it.
- HWiNFO64: sensor monitoring and the hardware summary you'll submit with your intake.
- CapFrameX: our primary benchmarking tool. We use it to measure in-game frame time performance and stability. Uses Event Tracing for Windows (ETW) to avoid polling stutters.
- FurMark 2: a brutal GPU stress test used to quickly validate GPU undervolts and thermals.
- OCCT: a more thorough stability test that runs varied load types (CPU, AVX, memory, combined) and pinpoints which individual core is unstable.
Depending on the depth of your RAM tune, we may also have you install Y-Cruncher and TestMem5 for memory stability validation. These come into play when we push aggressive RAM timings or overclocks and need to confirm long-term stability under load.
3. The Session
We run the full benchmark together to capture your "before" numbers, then work through BIOS tuning and Windows-level optimizations, validating stability the entire way. You watch every change and we explain the reasoning as we go.
Note: we perform a deep CPU stability and stress test together during your live session to ensure 100% stability. Don't apply any BIOS or Windows tweaks yourself before the session.