Coins and algorithms·28 min·updated 13 листопада 2026

Hard-drive mining: Proof of Space, Chia and the real economics of disks

Hard-drive mining is not faster SHA-256 hashing and it cannot turn every old disk into guaranteed income. In Proof of Space networks, a farmer writes large plot files to storage in advance and then answers network challenges. Reward probability is mostly related to allocated space, but the result also depends on plotting, uptime, pools, coin price, SSD endurance and the cost of the disk array.

#HDD#SSD#Chia#Proof of Space#mining

0 views

What people mean by hard-drive mining

In classic Proof of Work, a miner continuously performs huge numbers of hashes and competes on speed and energy efficiency. In Proof of Space or Proof of Capacity, the main work moves to a different phase: a precomputed dataset is written to storage, and the farmer later proves to the network that the space is available.

The word mining is therefore approximate. During farming, a disk is not running ASIC-scale hashing. It stores plots, reads small fragments for a network challenge and returns a proof. In Chia, this is called Proof of Space and Time: space is combined with a proof that real time passed through a VDF.

PoW — continuous hash computation
Proof of Space — precomputed storage
the farmer stores plots and the harvester reads them
a connected disk alone does not guarantee reward

How Chia works: plot, farmer and harvester

First, a plotter creates a plot — a large file containing structured pseudorandom data. This requires CPU or GPU resources, RAM and fast temporary SSD storage. When the plot is complete, it is moved to an HDD or another long-term storage device.

The farmer watches the network and wallet, while the harvester handles the physical plots. On a challenge, the harvester looks for a proof quality on disk, and only if that quality is good enough does it read a fuller proof and return it to the farmer. This lets one controller manage many inexpensive disks.

plotting is the heavy preparation phase
farming is long-term storage and checking
the harvester works with plot files
the farmer coordinates challenges, wallet and rewards

HDD versus SSD: use each for the right job

HDD is usually better for final storage because it offers a low cost per terabyte and can hold a large plot library. Its weaknesses are mechanics, noise, slower random access and failure risk under long operation. Chia farming is designed so that even a slow HDD can return a proof within the allowed response window.

SSD is primarily useful as temporary plotting storage or cache. Plotting creates a large amount of writes and can quickly reduce an SSD's TBW life. Keeping finished plots on an expensive NVMe is often uneconomic: SSD costs more per terabyte while its advantage during farming is usually small.

HDD — inexpensive long-term storage
SATA SSD — a plotting and cache compromise
NVMe — fast temporary scratch storage
do not confuse plotting speed with farming reward probability

Plotting: the heaviest and riskiest phase

During plotting, the system sorts and mixes large datasets. This loads the CPU, memory, temporary SSD and PCIe controller. Running plotting without a plan can overheat an NVMe, exhaust pagefile space, destabilize power or wear a drive prematurely.

A practical layout uses a separate fast scratch drive, keeps important originals away from it, monitors temperature and writes, and moves finished plots to an HDD array. Verify each plot after creation so that weeks of storage are not spent on damaged or incomplete data.

monitor SSD TBW and temperature
do not run more parallel jobs than RAM can handle
keep free-space headroom
verify plots after creation

What happens during one network challenge

The network sends a challenge, and the harvester reads the required fragments from plots to build a quality. In most challenges the quality is not good enough, so there is no need to read the full plot. If the result is promising, the system performs additional random reads and sends the full proof for verification.

This is why disk farming is not like GPU mining: average bandwidth can be low, yet the system is sensitive to drive dropouts, USB-SATA bridges, cables, power and short latency spikes. A missing mount or filesystem error can reduce effective space without an obvious change in a profitability calculator.

quality is a fast first check
the full proof is read only for a promising quality
the response window matters more than peak disk speed
farm availability directly affects reward probability

Do you need an SSD for farming itself

Finished plots can be stored and read without an SSD. Chia documentation explicitly describes harvester operation with HDDs; correct plot distribution, a stable controller and connectivity matter more. SSD is justified where it shortens plotting or works as cache, not simply as an expensive container for the same files.

For a small farm, one good local HDD can be more reliable than many disks behind a cheap USB hub. For a large farm, use an HBA or backplane, SMART monitoring, spare cables, a UPS and alerts for temperature, read errors and response time.

SSD does not increase allocated space
HDD is sufficient for long-term farming
USB hubs add failure points
UPS and SMART can matter more than another terabyte

Economics: terabytes do not equal profit

In Proof of Space, expected reward share is roughly related to your effective space divided by the network's total space. That does not mean every terabyte pays a fixed amount every day. Netspace, pool fee, coin price, difficulty, uptime and operating costs change.

Calculate the full cost: HDDs, enclosure or HBA, power, fans, motherboard, plotting SSD, cables, electricity, internet, downtime and drive replacement. Large arrays may use less energy per terabyte, but capital cost and the risk of a controller failure affecting many drives also grow.

income depends on your share of netspace
a pool reduces variance but charges a fee
include the power draw of the whole system
calculate ROI with scenarios, not one price

Pools, solo farming and payouts

Solo farming can produce a larger single reward, but waiting time becomes highly uneven. Pool farming distributes payouts based on plot contribution and makes cash flow more predictable, while adding rules, a threshold and a fee. Before joining, verify the pool address, payout scheme, signature requirements, minimum amount and availability history.

Do not treat a pool estimate as a guarantee. A farmer outage, wrong payout address, bad port, damaged plots or a lagging node can reduce the actual payout. Keep a log of plots, effective space, last proof, partials, uptime and last payment.

solo — high reward variance
pool — lower variance and a fee
verify payout rules before connecting
partials history reveals problems early

Drive endurance, temperature and reliability

The most severe wear usually affects the SSD used for repeated plotting, not an HDD quietly farming. Check TBW, SMART attributes, temperature, interface errors, reallocated sectors and pending sectors. For HDDs, vibration, temperature, power quality and operating hours matter as much as the advertised capacity.

A multi-drive array needs cooling even if each disk uses little power. Airflow must pass between drives, SATA cables must not be kinked, and power should be distributed without overloading one connector. One failed disk reduces effective space; a controller failure can disconnect a whole row.

SMART is an early warning system
do not use an SSD without checking TBW
temperature and vibration affect HDDs
monitor every drive separately

Who should consider hard-drive mining

Disk farming can make sense if you already have inexpensive storage, cheap electricity, stable infrastructure and the willingness to maintain an array. It is not a magic way to monetize a random old disk: enclosure, controller, electricity and your time can easily consume expected rewards.

Start with a small test. Create a few plots, measure response time, lost partials, temperature, power and actual pool payout. Only then scale and compare effective cost per TiB rather than the advertised price of one HDD.