Mining and AI glossary
Hashrate, J/TH, hydro cooling, VRAM, LLM… Here we explain, in plain words, the terms you come across on our product pages: cryptocurrency mining, ASIC miners, graphics cards and artificial intelligence. Search for a word or browse the glossary by letter.
A
- Cooling
Air cooling
The most common cooling method: fans push ambient air through the miner's heatsinks to carry the heat away.
- ASIC hardware
ASIC chip
An ASIC chip is an integrated circuit designed for a single calculation, such as Bitcoin's SHA-256. Dozens are mounted on each hashboard.
- Mining
ASIC
An ASIC is a mining machine built around chips designed for a single algorithm, far faster and more efficient than a CPU or a graphics card.
- Software
Autotune
Autotune is a firmware feature that sets the frequency and voltage of each chip in a miner on its own to find the best balance.
B
- Mining
Batch
A batch is a production lot of ASIC miners. It indicates the planned manufacturing or delivery period, especially for pre-orders.
- GPUs and AI
Blackwell
NVIDIA GPU architecture unveiled in 2024, succeeding Hopper and Ada Lovelace. It powers the B200 and B300, the RTX 50 and the RTX PRO 6000.
- Algorithm
Blake2s
Blake2s is a fast hash function from the Blake2 family, used by Kadena for its proof of work and mined with dedicated ASICs.
- Algorithm
Blake3
Blake3 is a very fast hash function, used in particular by Alephium for its proof of work. It is mined with dedicated ASICs.
- Mining
Block reward
The block reward is what the miner who finds a block receives: newly created coins plus the fees of the transactions included.
- Mining
Blockchain
A blockchain is a shared public ledger made of blocks of transactions chained together, which miners extend and secure.
- Mining
Break-even point
The break-even point is when a mining machine's net earnings have paid back its purchase price. It is also called ROI.
C
- Electricity
C13 / C19 connector
C13 and C19 are standard power connectors: the C13 is rated for 10 A, the C19 for up to 16 A and is found on many powerful ASICs.
- Cooling
Cold plate
A metal plate with coolant flowing through it, placed against the chips to draw out their heat. A key part of hydro miners.
- AI
Context window
The context window is the maximum amount of text, counted in tokens, that an AI model takes into account at once: instructions, documents and answer.
- ASIC hardware
Control board
The control board is an ASIC's small computer: it runs the firmware, handles the network and drives the hashboards, fans and power supply.
- GPUs and AI
CUDA
NVIDIA's computing platform for programming its GPUs for general-purpose tasks. CUDA cores are the compute units of these cards.
D
- Cooling
Dry cooler
An outdoor heat exchanger that cools the liquid of a hydro circuit with ambient air, using a radiator and fans, without consuming water.
E
- Algorithm
Eaglesong
Eaglesong is the hash function created for Nervos CKB's proof of work. Today it is mined by specialized ASICs.
- GPUs and AI
ECC memory
Memory that can detect and correct certain data errors on its own. Used in servers, workstations and data center GPUs.
- Electricity
Electrical installation
The cables, protective devices and sockets that power your machines. An ASIC miner needs a dedicated circuit sized by an electrician.
- AI
Embedding
An embedding is a vector, a list of numbers, representing the meaning of a word, sentence or document: two similar texts have similar vectors.
- Mining
Energy efficiency
A miner's energy efficiency shows the energy used per unit of hashrate, in J/TH or J/GH. The lower the figure, the more efficient the machine.
- Algorithm
Equihash
Equihash is a memory-intensive proof-of-work algorithm used by Zcash. Its hashrate is measured in solutions per second (Sol/s).
- Algorithm
Etchash
Etchash is Ethereum Classic's mining algorithm, derived from Ethash, Ethereum's former algorithm. It relies on a large dataset called the DAG.
F
- ASIC hardware
Fan
The fan removes heat from an air-cooled miner. Its airflow is expressed in CFM and its speed follows the chip temperature.
- AI
Fine-tuning
Fine-tuning means retraining an existing AI model on your own data to adapt it to a specific task, style or vocabulary.
- Software
Firmware
Firmware is the embedded software that runs a miner: chips, fans, power supply, pool connection and the web interface for settings.
- Software
Flight sheet
In HiveOS, a flight sheet is a configuration profile that brings together a rig's coin, wallet, pool and mining software.
- GPUs and AI
FLOPS
The number of floating-point operations a chip performs per second. It is the unit of computing power for GPUs and AI chips.
- GPUs and AI
FP8 / FP4 precision
8-bit or 4-bit number formats used in AI. Fewer bits mean less memory and more operations per second, at the cost of a little precision.
- GPUs and AI
FPGA
A reprogrammable chip whose circuits can be redefined after manufacture. More flexible than an ASIC, more efficient than a GPU on a specific task.
G
- GPUs and AI
GDDR7
The graphics memory generation used on recent GPUs such as the GeForce RTX 50. It succeeds GDDR6X with much higher bandwidth.
- AI
Generative AI
Generative AI refers to models that can create new content, such as text, images, audio, video or code, from a simple written request.
- AI
GPU rental
GPU rental means renting computing power by the hour. On the owner side, you make your cards available to other customers in exchange for payment.
- Software
GPU settings
GPU settings (power limit, core clock, memory clock) adjust a graphics card's power and frequencies to balance speed against power consumption.
H
- Mining
Halving
The halving cuts the block reward in half. On Bitcoin, it happens every 210,000 blocks, roughly every four years.
- Mining
Hash function
A hash function turns any data into a fixed-size fingerprint that cannot be reversed. It is the basic calculation behind mining.
- ASIC hardware
Hashboard
A hashboard is the circuit board in an ASIC that carries the computing chips. It is what produces the hashrate.
- Mining
Hashrate units
Hashrate units (kH/s, MH/s, GH/s, TH/s, PH/s, EH/s) count hashes per second; each step up is worth 1,000 times the one before.
- Mining
Hashrate
Hashrate is the number of hash calculations a miner or a network performs every second. It is the basic measure of mining power.
- GPUs and AI
HBM
Very high bandwidth memory, stacked in layers right next to the GPU. HBM3e is used on AI accelerators such as the NVIDIA H200.
- Energy
Heat recovery
Reusing the heat produced by miners to warm a home, water, a greenhouse or a building, instead of releasing it outside.
- Cooling
Heatsink
A finned metal part fitted on a chip to move its heat into the air or a liquid. Found on ASICs, GPUs and AI servers.
- GPUs and AI
HGX and DGX
HGX is NVIDIA's baseboard combining 4 or 8 SXM GPUs for server makers; DGX is the complete server sold by NVIDIA.
- Software
HiveOS
HiveOS is a Linux-based mining operating system. It sets up and monitors GPU and ASIC rigs from an online dashboard.
- Mining
Home miner
A home miner is a small miner designed for the home: quiet, low-power, plugged into a standard outlet and often used for solo mining.
- GPUs and AI
Hopper
NVIDIA data center GPU architecture unveiled in 2022, used in the H100 and H200. It introduced FP8 and the Transformer Engine.
- Cooling
Hydro cooling
Cooling by liquid flowing through plates mounted on the chips. Hydro miners are quieter and pack more power into less space.
I
- Cooling
Immersion cooling
A technique where the miner is submerged in a dielectric liquid, which does not conduct electricity and absorbs heat from every component at once.
- AI
Inference
The stage where an already trained AI model is used to respond: generating text, analysing a document or an image. It is the everyday use of AI.
K
- Algorithm
kHeavyHash
kHeavyHash is Kaspa's proof-of-work algorithm. It combines a Keccak-family hash with a matrix calculation and is mined with ASICs.
L
- GPU hardware
LHR
A limiter NVIDIA introduced in 2021 on some RTX 30 cards to cut their Ethereum hashrate. It no longer has any practical effect today.
- AI
LLM
Large language model: an AI trained on huge volumes of text, able to understand a request and answer it, token by token.
- AI
Local AI
Local AI runs an artificial intelligence model on your own hardware, at the office or at home, without sending your data to the cloud.
M
- GPUs and AI
Memory bandwidth
The amount of data a GPU's memory can exchange with the chip each second, in GB/s or TB/s. A key factor in AI speed.
- Mining
Merged mining
Merged mining lets you mine several cryptocurrencies that share the same algorithm at the same time, without using any extra energy.
- Mining
Miner hosting
Miner hosting means entrusting your machines to an equipped site that powers, cools and monitors them for you.
- Software
Miner web interface
A miner's web interface is its built-in settings page. You open it by typing the machine's IP address into a browser on the same local network.
- Mining
Mining algorithm
A mining algorithm is the calculation rule a cryptocurrency imposes on its miners, such as SHA-256 for Bitcoin or Scrypt for Litecoin.
- Mining
Mining difficulty
Mining difficulty sets the effort needed to find a block. The network adjusts it to keep block production at a steady pace.
- GPU hardware
Mining frame
A mining frame is an open frame that holds a GPU rig's motherboard, power supply and graphics cards, with plenty of space between them.
- Mining
Mining pool
A mining pool combines the power of many miners to find blocks more often and share the rewards according to the work each one contributes.
- Mining
Mining profitability
Mining profitability compares what a machine brings in (rewards and fees) with what it costs: electricity, hosting, maintenance and purchase.
- ASIC hardware
Mining rack
A mining rack is a shelving unit or cabinet that holds several miners with their power and cooling, such as the Antrack.
- GPU hardware
Mining rig
A mining machine built from several graphics cards connected to a motherboard on an open frame, with one or more power supplies.
- AI
Mixture of experts
A mixture of experts (MoE) is an AI model split into specialised sub-networks, only a few of which compute each token: faster for the same size.
- AI
Model parameters
Numbers adjusted during training that hold an AI model's knowledge. Their count, in billions, shows the model's size and the memory it needs.
- GPUs and AI
Multi-GPU
Using several graphics cards in the same machine to combine their memory and computing power, for AI as well as mining.
N
- ASIC hardware
Noise level
A miner's noise level is measured in decibels (dB). An air-cooled ASIC is very loud; a home miner or a hydro model is much quieter.
- Mining
Nonce
The nonce is a number the miner changes in the block header on every attempt, to get a hash that meets the network's target.
- GPUs and AI
NVLink
NVIDIA's high-speed link that connects several GPUs directly to each other, much faster than PCIe, so they work as one.
- GPUs and AI
NVMe
A protocol that connects an SSD directly to the PCIe bus. An NVMe SSD reads and writes much faster than a SATA SSD, which speeds up data loading.
O
- AI
OCR
OCR (optical character recognition) turns an image or scanned PDF into digital text that can be copied, searched and read by an AI.
- AI
Ollama
Ollama is free, open-source software that downloads and runs language models on your computer, on Linux, Windows or macOS.
- AI
Open WebUI
Open WebUI is an open-source web interface that turns your local AI models into a chat assistant, accessible from any browser.
- Software
Overclocking
Overclocking pushes the chips of a miner or GPU beyond their stock frequency to gain hashrate, at the cost of higher power consumption.
P
- GPU hardware
PCIe riser
An adapter that connects a graphics card to a motherboard PCIe slot at a distance, so several GPUs can be connected and spaced out in a mining rig.
- GPU hardware
PCIe
The standard interface that connects expansion cards, such as GPUs and NVMe SSDs, to the motherboard. Its generation and lane count set the bandwidth.
- Electricity
PDU
A PDU (power distribution unit) splits one power feed across several machines through protected C13 or C19 outlets.
- Electricity
Power consumption
A miner's power consumption is its power in watts; multiplied by its running time, it gives the energy you are billed for in kWh.
- ASIC hardware
Power supply
The power supply (PSU) converts AC mains power into low-voltage DC for the miner. At Bitmain, it is called an APW.
- Mining
Proof of work
Proof of work requires miners to spend computing power to validate a block. This real cost is what secures the blockchain.
Q
- GPUs and AI
Quantization
A technique that stores an AI model's weights in fewer bits, 8 or 4 instead of 16, to cut the memory needed with little loss of quality.
R
- AI
RAG
RAG (retrieval-augmented generation) lets an AI answer from your documents: it finds the relevant passages, then writes its answer citing them.
S
- Mining
Satoshi
The satoshi is the smallest unit of bitcoin: 1 BTC equals 100 million satoshis. It is used for small amounts, such as mining earnings.
- Algorithm
Scrypt
Scrypt is a memory-intensive mining algorithm used by Litecoin and Dogecoin. Today it is mined with dedicated ASICs.
- Algorithm
SHA-256
SHA-256 is the hash function behind Bitcoin's proof of work. SHA-256 ASICs compute trillions of these hashes every second.
- Energy
Solar surplus
Electricity your solar panels produce but you do not use on site. A miner can use it during the day instead of feeding it into the grid.
- Mining
Solo mining
Solo mining means searching for a block on your own, without a pool: if you find one, you win the whole reward; if not, nothing. It is a lottery.
- Mining
Stratum
Stratum is the protocol a mining machine uses to talk to its pool: it receives calculation jobs from the pool and sends back its shares.
- GPUs and AI
SXM
The module form factor of NVIDIA data center GPUs: the chip sits on a module mounted on the server's baseboard rather than on a PCIe card.
T
- GPUs and AI
Tensor Cores
Specialised compute units in NVIDIA GPUs, designed to multiply matrices at very high speed. They do most of the work in AI.
- ASIC hardware
Thermal paste
Thermal paste fills the tiny gaps between a chip and its heatsink to transfer heat better. It wears out over time and must be replaced.
- Electricity
Three-phase and single-phase
Single-phase (230 V) powers most homes; three-phase spreads power across three phases (400 V) and suits several miners.
- AI
Token
The unit of text a language model works with: a word, part of a word or a symbol. An LLM reads, counts and generates text in tokens.
- GPUs and AI
TPU
An AI accelerator chip designed by Google. It is an ASIC dedicated to the matrix calculations of neural networks and is not sold as a card.
- AI
Transformer
The Transformer is the neural network architecture behind LLMs: its attention mechanism links each word in a text to the others to grasp its meaning.
U
- Software
Undervolting
Undervolting means lowering the voltage, and often the frequency, of a miner's chips to use less energy per unit of hashrate.
V
W
- Mining
Wallet
A wallet, or crypto wallet, holds the keys that give access to your cryptocurrencies and provides the address where a pool sends your earnings.
X
- Algorithm
X11
X11 is a mining algorithm that chains eleven different hash functions. Created for Dash in 2014, it is now mined with ASICs.
Z
- Algorithm
zkSNARK
A zkSNARK is a short cryptographic proof that a calculation is correct without revealing its data. Aleo has made it the basis of its mining.
No term matches your search.