plain-english reference

Hardware glossary.

32 essential PC terms explained without the marketing spin. Each entry has a plain definition, a short note on when it actually matters, and (where relevant) a link to a browser test that measures it. Written for someone who wants to understand their own rig, not pass a hardware exam.

CPU

central processor · 6 terms

Core

An independent execution unit inside a CPU. A modern desktop chip has 6–24 physical cores; each can work on a different task at the same time.

More cores don't always mean faster: many games only fully use 4–8 cores, and the highest single-core speed matters more than the total count. Cores really shine for parallel work — video editing, 3D rendering, compiling code, running many browser tabs. Rule of thumb: 6 cores is a solid gaming baseline in 2026; 8+ helps if you stream or multitask.

Test your CPU under load

Thread SMT / hyperthreading

A stream of instructions the CPU executes. With SMT (called Hyperthreading on Intel), one physical core can juggle two threads at once by filling idle execution slots.

An 8-core / 16-thread CPU has 8 real cores; the extra 8 threads recover about 15–30% of a core's worth of throughput on well-parallelized workloads. In games the benefit is smaller and sometimes negative — most competitive shooter players leave SMT on and don't think about it.

Run the multi-thread test

IPC instructions per cycle

How much useful work a CPU does per clock tick. A newer 4 GHz chip often beats an older 5 GHz one because its IPC is higher.

Marketing tends to emphasize clock speed because it's a single number that goes up. IPC gains from a new architecture generation are why "same GHz, way faster" is a real thing — Zen 3 to Zen 4 added ~13% IPC at similar clocks.

Cache L1 / L2 / L3

Tiny, fast memory built into the CPU. L1 is smallest and fastest (per-core), L3 largest and slowest (shared). Games and game engines rely heavily on cache to hide main-memory latency.

AMD's X3D chips add a big vertical L3 stack and dominate 1080p gaming benchmarks specifically because of it. In productivity workloads, cache matters less — a 5800X3D can lose to a plain 5800X in rendering while winning in games. When you see cache numbers cited, they usually refer to L3.

TDP thermal design power

The sustained wattage a chip's cooler is expected to handle. Real peak power draw can exceed TDP briefly during turbo boosts.

TDP is a cooling specification, not a hard power cap. A "125W TDP" i9 can pull 250W+ under all-core load. Read reviewer power tests, not the TDP number, if you're sizing a cooler or PSU.

Thermal throttling

When a chip hits its temperature limit and lowers its clock speed to cool down. Shows up as a benchmark that starts fast then slows.

Every laptop throttles. Some desktops throttle if the cooler is undersized or dust has clogged the fans. Watch the "peak vs sustained" gap in a stress test — a big gap means throttling. If a rebench after cleaning fans jumps 15%+, that was it.

Watch for throttling

GPU

graphics processor · 3 terms

VRAM video RAM

Memory soldered onto a GPU. Textures, framebuffers, and shader data all live here. Modern games at 1440p need 8–12 GB; 4K with mods can push 16 GB+.

Running out of VRAM causes brutal stutter and blurry textures because the game has to stream data from system RAM instead. A card with less VRAM than the game wants can score high in average FPS but feel terrible because of texture pop-in.

Test your GPU under load

Ray tracing

A rendering technique that simulates real light bouncing off surfaces. Produces more accurate reflections and shadows at a heavy performance cost.

RT is dramatic in games designed around it (Cyberpunk 2077, Alan Wake II) and mostly cosmetic in games that bolt it on. Native RT tanks framerate 40–60% on most cards; combining with DLSS/FSR upscaling is how most people run it. Only the top three or four consumer GPU tiers make it practical at 1440p+.

DLSS / FSR / XeSS upscaling

Upscaling technologies that render frames at a lower resolution and reconstruct a higher-resolution image. NVIDIA's DLSS is neural-net based (RTX only); AMD's FSR and Intel's XeSS have vendor-agnostic equivalents.

In its "Quality" preset, upscaling often looks indistinguishable from native and gains 30–60% FPS. In "Performance" it's more visible but doubles framerate. DLSS 3/FSR 3 also add generated intermediate frames — smoother-looking but with slightly higher input latency. Rule of thumb: use it, especially with ray tracing on.

Memory

RAM · 2 terms

CAS latency CL

How many clock cycles it takes RAM to respond to a request. Lower is better. Combine with frequency to get true latency in nanoseconds.

DDR5-6000 CL30 has actual latency of (30 × 2 / 6000) × 1000 = 10 ns. DDR5-7200 CL36 works out to the same 10 ns — the faster speed offset by looser timings. Chase the timing/frequency combo that gives the lowest real latency for your platform.

XMP / EXPO / DOCP

Memory overclocking profiles baked into RAM sticks. Enabling one in BIOS makes the RAM run at its advertised speed instead of the safe JEDEC default.

You paid for DDR5-6000, but out of the box it runs at 4800 until XMP (Intel) or EXPO (AMD) is enabled. If your RAM feels slower than the sticker claims, this is the reason. Check Task Manager's memory speed after enabling.

Storage

SSD / HDD · 3 terms

NVMe

A storage protocol designed for SSDs that talks over PCIe. Sequential reads of 3–14 GB/s depending on PCIe generation. Replaces older SATA/AHCI.

All modern M.2 slots are NVMe (except a handful that also accept SATA M.2 sticks). Real-world game loading and boot times are dominated by random 4K speed, not sequential — a PCIe Gen 3 NVMe feels the same as a Gen 5 for most people. Gen 5 is future-proofing.

Test your storage speed

Sequential vs random

Sequential speed is how fast a drive reads/writes a big continuous file. Random speed is small scattered operations. Real workloads are mostly random; sequential looks impressive on spec sheets.

Sequential reads of 7000 MB/s are what the box says; random 4K reads of 60–100 MB/s are what your OS and games actually feel. If two drives have the same sequential but wildly different random, the higher-random drive will feel snappier in normal use.

TBW terabytes written

How much data an SSD is rated to survive before its endurance runs out. Typical consumer NVMes are rated for 300–2000 TBW; enterprise drives go 10× higher.

Most gamers write 5–20 TB per year total, so 600 TBW is 30+ years of use. TBW mostly matters if you're doing constant video editing, running a database, or setting up a Chia farm. For a normal desktop, ignore it.

Display

monitor · 5 terms

Refresh rate Hz

How many times per second a monitor redraws its screen, in Hertz. 60 Hz is baseline; 120/144/240 Hz smooth motion for gaming; 360/500 Hz are esports niche.

The jump from 60 to 144 Hz is huge and instantly noticeable. 144 to 240 is meaningful for competitive shooters. 240 to 360+ is mostly for pro players and marketing bragging rights. Your GPU also has to actually produce those frames — no point buying a 360 Hz panel if your game runs at 90 FPS.

Test your monitor

Response time GtG / MPRT

How long a pixel takes to change color, in milliseconds. GtG (gray-to-gray) is the industry standard measure; MPRT (moving picture response time) is a marketing number.

A "1 ms MPRT" panel is usually a 4–8 ms GtG panel with strobing. Slow response times cause visible ghosting behind moving objects. IPS monitors have caught up to VA and TN over the past few years — the days of slow-IPS complaints are mostly over.

Adaptive sync G-Sync / FreeSync / VRR

Monitor tech that syncs refresh rate to the GPU's frame rate, eliminating tearing and stutter. NVIDIA's is G-Sync, AMD's is FreeSync; VRR is the generic HDMI/DisplayPort standard used by consoles.

Turn it on and forget about it. Most modern monitors support one or both; NVIDIA GPUs work with FreeSync monitors ("G-Sync Compatible") and vice versa in most cases. Meaningless if your framerate is already capped at the refresh rate or above.

HDR high dynamic range

Displays that can render brighter highlights and deeper blacks. HDR400 is minimum; HDR1000+ is the real deal. Requires HDR content to matter.

Fake HDR is worse than SDR — a "HDR400" IPS with edge-lit backlight can't hit the peak brightness the format demands. OLED and mini-LED with 1000+ nits are where HDR actually delivers. In Windows, HDR is finicky; on consoles it's usually plug-and-play.

Panel type IPS / VA / OLED / TN

IPS (accurate colors, decent response, weak blacks), VA (deep blacks, some ghosting, wide color), OLED (perfect blacks, best motion, burn-in risk), TN (fast, cheap, washed-out).

OLED is the new hotness for gaming — instant response times and true blacks — but it costs more and has burn-in for static UI elements. IPS is the safe default. VA is great for movies. TN is basically extinct outside cheap 240 Hz+ esports panels.

Input

keyboard, mouse · 3 terms

DPI dots per inch

How much cursor distance a mouse produces per inch of physical movement. 800 DPI is the esports standard; higher isn't better, it's just faster.

Pros use 400–1600 DPI with low in-game sensitivity — you get more physical range for precise micro-adjustments. Chasing 26000 DPI marketing numbers is meaningless unless you also have a 4K/8K display and want the pointer to fly across it in one flick.

Check your mouse polling

Polling rate

How often per second a mouse or keyboard reports its state to the PC, in Hz. 1000 Hz means one report every millisecond. 4000/8000 Hz mice exist but rarely feel different.

1000 Hz has been the standard for a decade and is what almost everyone uses. Higher rates measurably reduce input latency but the effect is a couple of milliseconds — real, but well below most players' perception threshold. Worth chasing only in top-tier competitive play.

N-key rollover NKRO

How many keys can be pressed simultaneously and all register correctly. Full NKRO means every key. Cheaper keyboards ghost or block on 3+ simultaneous presses.

Matters for gaming (holding W + A + Shift + Space and pressing a fourth key), for stenography, and for anyone using key combos. Most gaming keyboards advertise 6KRO or NKRO; office keyboards often don't specify because they don't need to.

Test rollover here

Network

internet · 3 terms

Bandwidth vs throughput

Bandwidth is theoretical maximum; throughput is what you actually achieve. A 1 Gbps link rarely delivers a full 125 MB/s in practice.

Your ISP advertises bandwidth; your download client measures throughput. The gap comes from protocol overhead, packet loss, distant servers, and Wi-Fi. If a speed test shows 700 Mbps on a 1 Gbps plan, that's normal. If it shows 50 Mbps, something's wrong.

Run the network test

Ping latency

Round-trip time for a packet, in milliseconds. Under 20 ms is excellent for gaming; over 100 ms starts to feel bad.

Bandwidth doesn't help ping. A 10 Gbps connection to a server 300 ms away is still 300 ms away. For competitive gaming, low ping to the game's server matters more than raw download speed. Wi-Fi typically adds 5–15 ms vs wired.

Jitter

Variation in ping over time. Steady 40 ms is fine; wildly bouncing 10–200 ms feels awful even if the average is low.

High jitter causes voice call warble, rubber-banding in games, and video buffering. It's usually a sign of a congested Wi-Fi channel, overloaded ISP, or bufferbloat on the router. Fixable with QoS/SQM on decent routers.

Power

PSU · 1 term

80 Plus certification

Power supply efficiency ratings — Bronze / Silver / Gold / Platinum / Titanium. Gold at 50% load is a sensible sweet spot for most builds.

A "Gold" PSU is 90% efficient at 50% load — of every 100 W it pulls from the wall, 90 W becomes usable DC power and 10 W becomes heat. Titanium is 94%. The higher tiers pay for themselves only on heavy 24/7 workloads. For gaming, Gold is the right target; anything cheaper on the tier often skimps on other components too.

Connectivity

USB / PCIe / interfaces · 2 terms

PCIe generations

PCI Express versions — Gen 3, 4, 5. Each roughly doubles the previous generation's bandwidth per lane. Matters for NVMe SSDs and GPU-to-motherboard throughput.

A Gen 5 x16 slot delivers 63 GB/s theoretical bandwidth. No consumer GPU comes close to using that; even an RTX 4090 barely uses Gen 4 x16. Where PCIe generation is genuinely limiting: NVMe SSDs (Gen 5 doubles Gen 4 sequential speeds) and multi-GPU compute rigs.

USB versions 2.0 / 3.x / 4

USB 2.0 is 480 Mbps, 3.0/3.1 Gen 1 is 5 Gbps, 3.2 Gen 2 is 10 Gbps, USB4 is 20–40 Gbps. USB-C is a connector shape, not a speed rating.

A USB-C port can be USB 2.0 or Thunderbolt 4 or anything between. If you're wondering whether a port is fast, check the motherboard/laptop spec — the physical port shape tells you nothing. For external SSDs, USB 3.2 Gen 2 (10 Gbps) or better is the modern baseline.

Performance

general terms · 4 terms

FPS frames per second · 1% low

Frames per second — how many images the GPU renders each second. Averages can hide stutter; the 1% low is the worst 1% of frames, and often determines whether a game feels smooth.

A game running at 90 FPS average with a 40 FPS 1% low feels worse than one running at 70 FPS average with a 60 FPS 1% low. When comparing benchmarks, look at 1% lows — they're the honest metric.

Measure your FPS

Frame time

How long a single frame took to render, in milliseconds. Even FPS with consistent frame times feels better than higher FPS with erratic ones.

60 FPS = 16.67 ms per frame; 144 FPS = 6.94 ms; 240 FPS = 4.17 ms. When someone talks about "frame time spikes" they mean a graph where some bars are 2–5× taller than the average — those show up as visible micro-stutter.

Bottleneck

The slowest link in the chain. A powerful GPU paired with a weak CPU spends time waiting on the CPU; the CPU is the bottleneck. Identifying yours tells you what to upgrade next.

In-game, if lowering graphics settings barely raises FPS, you're CPU-bottlenecked. If raising resolution tanks FPS proportionally, you're GPU-bottlenecked. Task Manager showing 100% GPU utilization means the GPU is the limit; 100% on a single CPU core means that game is single-thread limited.

Stress test vs benchmark

A benchmark measures peak performance. A stress test holds sustained load to check stability, thermals, and throttling. A great benchmark score is meaningless if the system crashes 20 minutes into a stress run.

Run both. A benchmark tells you what your rig can do; a stress test tells you what it can do reliably. Undervolted or overclocked chips especially need long stress runs — they can pass a 30-second benchmark and hard-crash after an hour.

Full sweep benchmark
⌕
No terms match that search. Try shorter keywords, or clear the search to see all terms.

Now measure it yourself

Every definition here is theory. The tests take about 3 minutes each and everything runs locally in your browser. Or browse the hardware reference for verdicts on specific GPUs and CPUs.

Run the full benchmark →