Decode
PC terms decoded
What the names on PC parts really mean, and why they are called that.
CPU names
AM4, AM5
AMD's desktop CPU sockets: the slot on the motherboard the processor sits in. A CPU only fits a board with the same socket. AM4 is Ryzen 1000 to 5000, AM5 is Ryzen 7000 onwards.
Why the name: AMD has used "AM" for its desktop sockets since Socket AM2 in 2006. The number is just the generation. AM4 CPUs have the pins on the chip; AM5 moved the pins onto the motherboard.
LGA (LGA1700, LGA1851)
Intel's socket names. The number is how many contact points the socket has, so LGA1700 has 1,700 tiny spring pins.
Why the name: LGA means Land Grid Array: the CPU has flat gold pads ("lands") and the pins live in the motherboard. That is why you check the board for bent pins, not the CPU.
sTR5 and Threadripper
AMD's workstation platform: much bigger CPUs with up to 64 cores (96 on the PRO versions), 4 or 8 memory channels and lots of PCIe lanes for pro work.
Why the name: The name says what it does: rips through work on many threads at once. The chip is so large it needs a special cooler mount.
Ryzen 3 / 5 / 7 / 9
AMD's tiers, like small to large. Ryzen 5 is the gaming sweet spot, 7 and 9 add cores for work.
Why the name: The tiers line up on purpose with Intel's Core i3, i5, i7 and i9, so shoppers can compare at a glance.
Ryzen model number (9800X3D)
First digit = generation (9 = Ryzen 9000). The next digit = tier inside it. The letters at the end change what the chip is good at.
Why the name: Ryzen 8000 desktop chips (8400F, 8600G) are actually the same generation as 7000, built from laptop designs. Model numbers are marketing, not maths.
X3D
AMD CPUs with a big extra block of cache memory stacked on top of the chip. Games love it: an X3D is usually the fastest gaming CPU you can buy.
Why the name: "3D" refers to 3D V-Cache: the cache is a separate layer of silicon stacked vertically, bonded directly onto the processor.
X, G, F (AMD) and K, F, KF (Intel)
Suffix letters. AMD: X = higher clocks, G = strong built-in graphics, F = no built-in graphics. Intel: K = unlocked for overclocking, F = no built-in graphics, KF = both.
Why the name: An F chip needs a graphics card or the screen stays black. It is the most common "my new PC shows nothing" mistake.
Core i5-14600K
Intel naming: i5 is the tier, 14 is the generation (14th gen), 600 is the position inside it, K means unlocked.
Why the name: Intel used this pattern from 2008 until 2024, when desktop chips became "Core Ultra".
Core Ultra 7 265K
Intel's newer naming. Ultra 5 / 7 / 9 are the tiers. The first digit (2) is the series. These sit in the LGA1851 socket.
Why the name: Intel dropped the little "i" when Core Ultra arrived, first in laptops (late 2023), then desktops (late 2024).
Cores and threads
Cores are the CPU's workers. Threads let each core juggle two jobs. 8 cores / 16 threads is plenty for gaming; editing and 3D love more.
Why the name: AMD calls it SMT, Intel calls it Hyper-Threading. Same idea. Intel's newest desktop chips actually removed it.
GHz (clock speed)
How many billions of steps a core does per second. Only useful for comparing chips of the same family: a newer 4.5 GHz chip can beat an older 5 GHz one.
Why the name: The "megahertz race" ended in the mid-2000s when chips got too hot. Since then, smarter designs matter more than raw clock speed.
Graphics card names
RTX and GTX
NVIDIA's gaming cards. RTX cards (since 2018) have ray-tracing cores and AI cores for DLSS. GTX was the older line without them.
Why the name: RT in RTX is for ray tracing, the lighting technique movies use. Gaming RTX cards arrived in 2018 with the RTX 20 series.
The numbers in RTX 5070
50 = the generation (RTX 50 series). 70 = the tier: 50 and 60 are entry level, 70 mid-range, 80 high end, 90 the top.
Why the name: So an RTX 4090 (older top card) can beat a new RTX 5070. Compare tiers within a generation, and check our Compare page across generations.
Ti
A faster version of the same tier: an RTX 5070 Ti sits between the 5070 and the 5080.
Why the name: Ti is short for Titanium. NVIDIA first used it in 2001 with the GeForce 3 Ti.
Super
A mid-generation refresh: same name, more cores, often the same price. An RTX 4070 Super beats a plain 4070.
Why the name: NVIDIA introduced Super cards in 2019 with the RTX 20 Super series. The combo name Ti Super (4070 Ti Super) sits a step above Ti.
Radeon RX 9070 XT
AMD's gaming cards. 9 = generation, 070 = tier. XT = the faster version, XTX = the top one.
Why the name: AMD changed its numbers in 2025 so they line up with NVIDIA's: an RX 9070 is meant to compete with an RTX 5070.
GRE
A special edition, like the RX 7900 GRE.
Why the name: GRE means Golden Rabbit Edition. It launched first in China for the Year of the Rabbit in 2023, then went worldwide.
Intel Arc A / B
Intel's graphics cards. A = first generation (A750, A770), B = second (B570, B580). Good value at 1080p.
Why the name: The letters follow fantasy classes in alphabetical order: Alchemist, Battlemage, then Celestial and Druid.
VRAM (video memory)
The graphics card's own memory for textures and AI models. 8GB is tight for new games at high settings; 12 to 16GB is comfortable; AI wants as much as possible.
Why the name: For local AI, VRAM matters more than speed: a model that does not fit in VRAM runs many times slower.
Founders Edition, OC, Dual, Gaming
The brand names on cards (ASUS TUF, MSI Gaming Trio) are versions of the same chip with different coolers and looks. Speed differs by only a few percent.
Why the name: Founders Edition is NVIDIA's own design. Bigger coolers mainly buy you quieter fans and lower temperatures.
Memory (RAM)
DDR4, DDR5
Generations of RAM. They are physically different: the notch is in a different place, so DDR5 will not fit a DDR4 board.
Why the name: DDR means Double Data Rate: the memory sends data twice per clock tick, on the rise and the fall of the signal.
MT/s (megatransfers) vs MHz
DDR5-6000 means 6,000 million transfers per second. That is the speed you compare.
Why the name: Many boxes say "6000 MHz", but the real clock is 3,000 MHz. Because of Double Data Rate, MT/s is the honest unit.
CL (CAS latency)
How many clock ticks the RAM waits before sending data. Lower is quicker, but only when you compare the same speed.
Why the name: Real delay = CL × 2000 ÷ speed. DDR4-3200 CL16 and DDR5-6000 CL30 both wait 10 nanoseconds, so a "bigger" CL is not always slower.
XMP and EXPO
A one-click setting in the BIOS that runs your RAM at the speed on the box. Without it, RAM runs at a slow default.
Why the name: XMP is Intel's Extreme Memory Profile; EXPO is AMD's version. Forgetting to turn it on is one of the most common "free performance" fixes.
Dual channel (2×16GB)
Two matching sticks let the CPU talk to both at once. 2×16GB is faster than 1×32GB.
Why the name: On a 4-slot board, use slots A2 and B2 (the 2nd and 4th from the CPU). The board manual has a little diagram.
ECC and RDIMM
ECC memory fixes small errors on the fly; RDIMM adds a buffer chip so you can use lots of it. Used in servers and workstations like Threadripper.
Why the name: Cosmic rays really can flip a bit in memory. ECC exists so a long render or simulation does not silently go wrong.
Storage
NVMe and M.2
M.2 is the stick shape and slot. NVMe is the fast way it talks to the PC. Most M.2 drives today are NVMe, but some older ones are slower SATA.
Why the name: NVMe means Non-Volatile Memory Express. "Non-volatile" just means it keeps data when the power is off.
PCIe 4.0 and 5.0
The lanes that connect graphics cards and SSDs. Each generation doubles the speed: about 7 GB/s for a fast PCIe 4.0 SSD, about 14 GB/s for 5.0.
Why the name: For games, a good PCIe 4.0 SSD already loads as fast as 5.0. You feel the difference mainly when copying huge video files.
Why 1TB shows as 931GB
Drive makers count 1TB as 1,000,000,000,000 bytes. Windows counts in powers of 2, so it shows about 931GB. Nothing is missing.
Why the name: Strictly, Windows should say 931 GiB (gibibytes). Almost nobody uses that word.
SSD vs HDD
SSDs use chips and have no moving parts: fast and silent. HDDs are spinning disks: slow, but cheap for big backups.
Why the name: Putting Windows on an SSD is the biggest "feels like a new PC" upgrade for an old machine.
Motherboard and power
Chipset (B650, X870E, Z790)
The letter is the class. AMD: A = basic, B = mainstream, X = high end. Intel: H/B = mainstream, Z = overclocking. The number is the generation.
Why the name: E (as in X870E) means Extreme: more PCIe 5.0 lanes. For most builds a B board is plenty.
ATX, Micro-ATX, Mini-ITX
Motherboard sizes, large to small. The case must support the size. Smaller boards have fewer slots.
Why the name: Intel created the ATX standard in 1995 and it still fits today's cases. Mini-ITX came from VIA in 2001, for tiny PCs.
VRM
The power circuit on the motherboard that feeds the CPU. Cheap boards with weak VRMs can slow down hungry CPUs.
Why the name: Those finned metal blocks around the CPU socket are VRM heatsinks.
BIOS / UEFI
The settings screen before Windows starts. You use it to turn on XMP/EXPO, check temperatures and update support for new CPUs.
Why the name: BIOS means Basic Input/Output System and dates back to 1975. UEFI is the modern replacement, but everyone still says BIOS.
80 PLUS Gold
A power supply efficiency rating. Gold wastes about 10% as heat at half load. Bronze is a bit less efficient; Platinum and Titanium a bit more.
Why the name: The original 80 PLUS (2004) only promised 80% efficiency. Gold is the sensible standard for a new PC.
12V-2x6 (12VHPWR)
The small 16-pin power plug on new graphics cards. It must be pushed in fully until it clicks.
Why the name: Half-inserted plugs on early cards could overheat and melt, so the revised 12V-2x6 design uses shorter sensing pins.
Graphics and screens
FPS and Hz
FPS is how many frames the graphics card makes per second. Hz is how many the monitor can show. A 144Hz monitor shows up to 144 FPS.
Why the name: Movies run at 24 FPS. Games feel smooth from about 60, and competitive players use 240Hz or more.
1080p, 1440p, 4K
Screen sharpness. 1080p = 1920×1080, 1440p = 2560×1440, 4K = 3840×2160. Each step needs a much stronger graphics card.
Why the name: The "p" means progressive scan, a leftover from TV days. 4K has four times the pixels of 1080p.
Ray tracing
Lighting that follows the path of light rays for realistic reflections and shadows. Beautiful, but heavy on the graphics card.
Why the name: Film studios used ray tracing for decades, taking hours per frame. Games now do a version of it 60 times a second.
DLSS, FSR, XeSS
Upscaling: the game draws a smaller image and AI or clever maths enlarges it, for more FPS with little visible loss. DLSS is NVIDIA, FSR is AMD, XeSS is Intel.
Why the name: DLSS means Deep Learning Super Sampling. "Frame generation" goes further and creates whole in-between frames.
Bottleneck
When one part holds another back. A fast graphics card with a slow CPU waits around, so you pay for power you cannot use.
Why the name: The name comes from a bottle: however big the bottle, the narrow neck sets how fast it pours.
Cooling
TDP
Roughly how much heat the cooler must remove, in watts. Pick a cooler rated above your CPU's TDP.
Why the name: TDP means Thermal Design Power. Real power draw can go well above it when the CPU boosts.
AIO, 240mm, 360mm
An all-in-one liquid cooler: pump, tubes and radiator, sealed and ready. 240mm and 360mm are the radiator length: 2 or 3 fans of 120mm.
Why the name: Big air coolers often match a 240mm AIO and cannot leak or have a pump fail. AIOs win on looks and space around the CPU.
Thermal paste
A thin layer between the CPU and cooler that fills tiny gaps so heat moves across. A pea-sized dot is enough.
Why the name: Many coolers come with paste already applied. Check before adding more.
AI and work
CUDA, ROCm
Software that lets apps use the graphics card for AI and compute. CUDA is NVIDIA only and works with almost everything. ROCm is AMD's, improving fast.
Why the name: CUDA launched in 2007, which is a big reason NVIDIA leads in AI: most AI software was written with it first.
7B, 70B (model size)
The number of parameters in an AI model, in billions. Bigger models are smarter but need more memory.
Why the name: A 4-bit compressed model needs about 0.6GB of VRAM per billion parameters, so a 14B model fits on a 12GB card.
Tokens per second
How fast an AI model writes. A token is about three quarters of a word. 20 or more tokens per second feels faster than reading.
Why the name: Speed mostly depends on memory bandwidth, not raw compute. That is why VRAM speed matters so much for local AI.
H100, B200, B300
NVIDIA data centre GPUs for training and serving AI. H = Hopper, B = Blackwell. They come in servers of 8.
Why the name: NVIDIA names its GPU architectures after scientists: Hopper (Grace Hopper), Blackwell (David Blackwell) and Rubin (Vera Rubin).
Open the free interactive tool