What changed after mining SASEUL on one graphics card for 73 days? 25% vs 50% usage, and what happened in October

What changes after 73 days on one graphics card?
At 50% usage, speed went up 3-4x. The last time it shut itself off was October 15
August, 25% usage
Daily average hashrate about 180-315 MH/s
50% from September
About 550-890 MH/s, share count about the same
What happened in October
Pool timeouts about 40-164 a day in early October, fewer from the 14th. Shut itself off 11 times, none after the 15th
There's a log of one graphics card running SASEUL from August 18 to November 1. Leave out the three days it wasn't on (September 23, 24 and 28) and you get 73 days. It picks up where the 28-day log left off.
Two things changed. Raising card usage from 25% to 50% made the hashrate jump three to four times. And in early October, timeouts with no answer from the pool went up, and the miner shut itself off a few times. Timeouts started to fall around October 14, and from October 15 it went about 17 days without shutting itself off.
So the thing to do right now is check now and then that the miner is still running.
Raising usage from 25% to 50% made the hashrate jump three to four times
Usage is the setting for how much of the graphics card the miner gets to use. In this log it was mostly 25% in August and 50% from September. There were short stretches at 100% too.
The hashrate (how many guesses per second) gets written to the miner log the whole time. Averaged by day, August at 25% was about 180-315 MH/s. From September to October 9 at 50% it was about 740-890 MH/s, mostly 800-870 MH/s. From October 10 to November 1 it was also 50%, at about 550-860 MH/s, mostly 750-820 MH/s. October 15 (about 565), 18 (about 622) and 19 (about 548) were lower, and the log doesn't say why.
840 MH/s means roughly 840 million guesses a second. Comparing the middle of each range, it went from about 230 to about 840 MH/s, so more than three times. But the two periods didn't have the same dates, card temperature or driver, so I can't say exactly how many times faster.
The setting went up two times, but the speed went up more than two times. The log alone doesn't say why. The 28-day log explains that 50% doesn't run the card at half strength. It runs at full and then rests, back and forth.
But accepted shares per day stayed about the same, around 7,000
Shares (the "proof of work" you hand to the pool) were accepted at about 6,300-7,200 a day in August. In September it was about 7,300-7,500 a day. After October 10, days when it ran all day were about 6,600-7,500 too. The hashrate went up three to four times, but the count barely moved.
Every share in the log has a line saying "difficulty adjusted to …" (the share bar was set to …). It looks like the pool changes the bar for one share person by person. My guess is that faster cards get a higher bar, so the counts come out about the same.
That means comparing "how much work got done" by share count alone is easy to get wrong. The example in the post on what shares are assumes one fixed bar. I couldn't confirm whether the pool counts the height of the bar when it splits the coins.
Most shares the pool turned down were "answers to a problem that had already moved on"
When the pool doesn't take a share, it sends back a one-line reason. Over the 73 days the most common was "Stale share", 24,970 times. It means the pool moved on to the next problem while the share was being found.
On days it ran all day, stale shares ran about 300-600 a day, roughly the same the whole time. Against the 6,600-7,500 accepted per day, that's 4 to 9 out of every 100. The 28-day log had rejects around 5% and guessed "maybe the answers arrive late". The pool's answers point the same way.
Every reason the pool turned down shares (total, August 18 to November 1)
| Pool's answer (exactly as in the log) | Meaning | Count |
|---|---|---|
| 409 Conflict "Stale share" | The pool moved on to the next problem while the share was being found | 24,970 |
| 400 Bad Request "Timestamp out of range" | The share's time was outside what the pool accepts. The log doesn't say why | 3,629 |
| "The read operation timed out" | The pool didn't answer while sending a share or getting a new problem, so time ran out | 1,451 |
| Other connection errors | e.g. getaddrinfo failed (couldn't find the internet address) | 354 |
What "Stale share" and "Timestamp out of range" each mean is covered in the stale share post.
Timeouts went up a lot in early October, then fell from October 14
A timeout means the pool didn't answer in time when a share was sent or a new problem was fetched. Through September it was 0-26 a day. From October 1 to 12 it rose to about 40-164 a day (164 on the 3rd, 151 on the 12th). It began to fall on the 14th: 36 on the 14th, 31 on the 15th, 14 on the 16th and 7 on the 17th. From October 22 to November 1 it's 0-35 a day (0 on the 27th, 2 on the 31st, 4 on November 1). That's clearly fewer than early October.
Timeouts per day (pool didn't answer)
Shares sent with an out-of-range time peaked at 229 on October 12, then ran about 10-64 a day after October 22. Only stale shares stayed about the same the whole time.
Around the same time, the pool grew. In this blog's network log, the number of mining machines on the pool's front page was 6 on September 27 and 9 on October 4. I don't know whether that's connected to the timeouts. How mining changes as the pool grows is in the more home miners post.
The miner shut itself off 11 times, and not once after October 15
Over the 73 days the miner stopped 26 times. 15 of those were someone pressing the stop button. 11 times it shut off by itself. This kind of stopping by itself is also covered separately in the miner stops by itself post.
The days it shut itself off were September 22, September 27, October 3 (twice), October 4, October 6 (three times), October 11, October 12 and October 15. The 8 times from September 22 to October 6 all looked the same. After a line saying a share was being sent to the pool, no answer came back, and 13-20 seconds later "GPU Miner stopped" appeared.
It didn't turn itself back on afterward. So mining is empty for that stretch. Like on October 3, when it shut off at 11:49 pm and was started again at 1:13 am the next day, it sat empty for close to an hour and a half.
The last time was 11:59 am on October 15. After that, for about 17 days until 8:33 pm on November 1, it never shut itself off. The 4 stops in that stretch were me pressing the stop button on October 25 and 26.
The shutoffs disappearing overlaps with timeouts starting to fall around October 14. That doesn't mean the cause is the same. What changed (the pool side or the internet) can't be told from the log.
The cause is inside the program, so I couldn't confirm it. My guess for the first 8 was that the miner hit a connection problem while sending a share to the pool and couldn't handle it, so it quit. Those times overlapped with when timeouts went up.
When I checked in early October, it didn't look like the computer went to sleep. It was set not to sleep while plugged in, and there was no sleep record for the last two days either.
So what should you do? Check about once a day that it's on, and compare by hashrate, not share count
- Open the miner window about once a day. There were no shutoffs lately, but I don't know why, so they could come back. If the window is closed or the hashrate box is frozen, start it again. If you don't notice it's off, that whole time is empty
- Run at the default 25% usage for a few days before raising it. After raising it, compare by hashrate (MH/s) and the coins that actually came in, not share count
- A few hundred rejected shares a day was normal in this log. Watch instead for timeouts showing up dozens of times a day or more. In this log, the miner shut itself off at the same time when timeouts were high in early October
- If you haven't started yet, begin with the Windows install screens walkthrough. The basics about SASEUL coin are on the SASEUL basics page
Checking it's running and restarting it from the black window (Command Prompt)
These commands are in the miner's help. Use them in the folder the miner is in. Your receiving address is saved after you enter it the first time.
miner.exe status
miner.exe start
miner.exe log
miner.exe stop
status shows whether it's running, start turns it back on, log shows the log, and stop turns it off. If you search the log for "timed out" or "GPU Miner stopped", you can see whether the same thing happened to you.
Sources
- Miner log: miner.log from the SASEUL Miner window version 0.9.2.18, pool saseulmine.com · log from 2026-08-18 to 2026-11-01, counted on 2026-11-01
- 28 days on one graphics card (how the 50% setting works, rejects around 5%, the guess that answers arrive late) · 2026-09-16
- What is a share in SASEUL mining? · 2026-10-09
- This blog's network log (mining machines on the saseulmine.com front page, 6 and 9) · recorded 2026-09-27 09:46 and 2026-10-04 07:12
I hold SL. The numbers come from one graphics card's miner log (August 18 to November 1, 2026), counted and written down on November 1, 2026.