Dieser Artikel ist auf Englisch.
Three ways to measure uptime
"Uptime" sounds like one number, but hosts and pools measure it in three different ways, and the results rarely agree to the decimal.
| Method | What it counts | Good at | Misses |
|---|---|---|---|
| Time hashing | Minutes the miner reports as running and hashing, divided by the minutes in the period | Simple, and easy to check against outages | A miner running at half speed still counts as fully up |
| Hashrate delivered | Average measured hashrate divided by the miner’s rated hashrate | Catches throttling, failed hashboards and heat | Depends on which rated figure is used |
| Pool-side accepted shares | The hashrate the pool estimates from the shares it accepted | Closest to what actually earns bitcoin | Noisy over short windows, and blind to why a miner stopped |
The gap between them can be large. A machine that runs all month with one of its three hashboards dead is close to 100% up by time, and about 67% by hashrate delivered. When a host quotes an uptime figure, ask which method it means.
Why pool data and site data differ
A pool sees only shares. It estimates hashrate statistically from how many shares arrive, so over an hour the figure swings with luck even when the miner is perfectly steady; over a day most of the swing averages out. Rejected and stale shares count in the miner’s own figure but not in the pool’s.
An agent on site reads each miner directly: hashrate per board, temperatures, fan speeds, and whether it has power at all. It can tell a dead power supply from a network fault, which the pool cannot. What it cannot see is a share lost after it leaves the site, for example when the internet link or the pool itself has a problem.
- Short windows. A one-hour pool chart is mostly luck; compare over 24 hours or more.
- Rejected and stale shares. Work the miner did that the pool did not credit.
- Internet or pool outages. The miner hashes, but its shares do not land.
- Configuration. A wrong worker name or pool address sends work somewhere you are not looking.
- Clocks and periods. A pool day in UTC and a report in Gulf time cover different hours.
That is why neither source is enough on its own. See how mining pools work for more on shares.
How Segments measures it
Segments’ ERP measures uptime for every hosted miner, around the clock, from two sources: the pool, and an agent running on a PC at the site. The pool data shows what is earning; the site agent shows what each machine is doing and why it stopped. Reading both lets a fault be traced to the miner, the network or the pool.
The 95% target is set in clause 10 of the hosting agreement; read it with the rest of the contract before you sign.
- Client portal. Every hosting client gets portal.segments.ae, in the same languages as this website, with their miners, hashrate, uptime and invoices.
- Centre status. The status page shows the state of each hosting centre.
- Your own pool. You choose the pool, so you can check its dashboard against both.
What 95% means in hours
| Uptime | Hours down in a 30-day month | Days down in a year |
|---|---|---|
| 99% | 7.2 | 3.7 |
| 98% | 14.4 | 7.3 |
| 97% | 21.6 | 11.0 |
| 95% | 36 | 18.3 |
| 90% | 72 | 36.5 |
At 95%, a miner can be offline for about 36 hours in a month and still meet the target. Those hours are rarely one block: a cleaning slot, a restart after a grid dip, a fan swap, a few hours of curtailment.
Everything else being equal, a miner at 95% earns about 95% of what it would at 100%. Segments’ monthly estimates assume the 95% target, so a month above it earns more than the estimate and a month below it earns less.
Why a target, not a guarantee
Some downtime is outside any host’s control: faults and maintenance on the utility’s side, curtailment requests, extreme heat, a hashboard that fails without warning. A host that guarantees uptime either prices that risk into the rate or writes the guarantee so narrowly that it rarely pays out. Segments states 95% as a target and explains how it is measured.
Billing on actual kWh is what makes a target fair. Power is invoiced on the kWh your miners actually use each month, so a miner that is switched off adds no kWh to the invoice. For example, an air-cooled miner drawing 3.5 kW and offline for 36 hours uses 126 kWh less; at 6.8¢ per kWh (October 2026), that is about $8.57 less, before 5% VAT. You lose the bitcoin those hours would have mined, but you do not pay for power you did not use.
Planned and unplanned downtime
| Type | Examples | What happens |
|---|---|---|
| Planned | Cleaning and maintenance, both included in the rate; firmware updates | The machine is off for a short, scheduled time |
| Curtailment | The grid or the site needs load reduced for a period | Miners are switched off or turned down, and use no kWh for the power they do not draw |
| Unplanned | Fan, PSU or control board failures, hashboard faults, network or pool outages, grid dips | Fans, PSUs and control boards are swapped on site; hashboards go to the UAE lab |
When a hosted unit goes into repair, there is no hosting fee while it is in repair. Repairs start with a diagnosis and a written quote, end with a 12-hour full-load bench test, and replaced parts carry a 7-day warranty. See common miner faults and fixes.
Questions to ask any host about uptime
- Is uptime measured by time, by hashrate delivered or from the pool, and over what period?
- Is it measured per miner, or averaged across the site?
- Do planned maintenance, cleaning and curtailment count as downtime?
- Where can I see it for each of my miners, without asking?
- Am I billed on metered kWh, or a flat fee I pay whether the miner runs or not?
- Do I keep paying hosting while a machine is in repair?
- Is uptime a target or a guarantee, and what exactly happens if it is missed?
- Can I check your figures against my own pool dashboard?
For the full checklist, read how to choose a Bitcoin mining hosting provider, or see Segments hosting plans.
Fragen
- What does 95% uptime mean for a miner?
- The miner can be offline for about 36 hours in a 30-day month, or about 18 days in a year, and still meet the target. Ask whether the figure is measured by time, by hashrate delivered or from the pool.
- Why does my pool show a different uptime from my host?
- The pool estimates hashrate from accepted shares, which swings with luck over short windows and leaves out rejected and stale shares. An agent on site reads each miner directly. Compare over 24 hours or more.
- Is uptime guaranteed at Segments?
- No. 95% is a target in clause 10 of the hosting agreement. Power is billed on the kWh your miners actually use, and there is no hosting fee while a hosted unit is in repair.
- Where can I see my miners’ uptime?
- In the client portal at portal.segments.ae, with hashrate and invoices; centre status is on the status page, and you can check your own pool’s dashboard as well.