Stale Share
A stale is a share sent against an outdated job. The pool does not pay for it, because the answer came to a question that had already been withdrawn.
What a stale share is
The pool hands the asic a job. It computes and sends back a solution. If a block turned up on the network in the meantime, that job stopped being current, and the share sent against it gets marked stale by the pool. The miner did honest work, but the answer came too late.
Rejects and stales are not the same thing, though statistics often add them together. A reject failed a check: wrong difficulty, a duplicate, somebody else's job. A stale would have passed the check, only the job had already changed. We went through the reasons for refusals on a separate page.
Stales cannot be removed completely. Time passes while the job flies to the asic and the solution flies back. So a normal level is measured in tenths of a percent, and the thing worth fighting is latency to the pool and instability of the miner rather than the phenomenon itself.
Overclocking adds stales indirectly, through faults: a board starts computing with errors, a chain goes into restart, the job gets lost. If both stales and rejects went up after you raised frequencies, that is a reason to roll back rather than to look for trouble in the network.
Quick reference table
| What it is | a share against an outdated job |
| Paid | no |
| Normal level | tenths of a percent |
| Main cause | latency to the pool |
| Not to be confused with | shares rejected by a check |
How to bring them down
Pick a pool server that is closer geographically. A European port instead of an American one usually takes away most of the stales right away.
Check the path to the pool. An overloaded Wi-Fi bridge, a cheap switch and a broken cable give latency and losses where nobody expects them. Twisted pair to the router beats any bridge.
Look at stales together with rejects and temperature. If all three figures grow, the asic is at fault rather than the network.
An example in numbers
An asic at 200 TH/s holds 0.2 percent stales, which is about 0.4 TH/s of work thrown away, and noticing that is impossible. Three percent already takes 6 TH/s: the miner computes as if it were 194 TH/s. That is the sort of figure you cure by changing the pool server rather than by patience.
Related terms
Where to go next on the site
Software
- Software firmware
- HashCore Toolkit firmware
Hardware
- ASIC Miner knowledge base
- Antminer S21 tool
Questions and answers
What count of stales is normal
Tenths of a percent. Above one percent it is worth looking at where the latency comes from.
Are stales a way for the pool to shave the payout
They show up in the pool statistics and in the miner log alike, and the numbers usually agree. Taking it up with the pool makes sense only for a large gap that holds for days.
Does changing firmware help
Sometimes: custom firmware can hold several pool servers and switch between them faster. But if the latency sits inside your own network, firmware will not remove it.
How the terms connect
Every link in the chain is clickable. Orange marks where you are now.
Looking for an ASIC miner
The catalog holds 212 models. You can compare them by hashrate and by joules per terahash, then plug your own rate into the calculator and see what stays in your pocket.
Page written and checked by Denys Klimchuk. Updated .