Open aleqx opened 3 years ago
An example from Nicehash (not even 1h of mining).
209/224 = 6.3% rejected. In the previous session it was close to 10%. I'm aware of how Nicehash works - there was no change of DAG or coin here during the past 20 minutes, yet there were plenty of rejects.
[...]
20201212 16:27:36 [ OK ] 208/222 - 334.94 MH/s, 110ms ... GPU #5
20201212 16:27:38 [ OK ] 209/223 - 334.95 MH/s, 175ms ... GPU #4
20201212 16:27:41 Mining at daggerhashimoto.eu.nicehash.com:3353, diff: 2.89 Gh
20201212 16:27:41 GPU #0: GTX 1080 Ti - 48.07 MH/s, 37/41 R:9.76%
20201212 16:27:41 GPU #1: GTX 1080 Ti - 47.59 MH/s, 26/29 R:10.34%
20201212 16:27:41 GPU #2: GTX 1080 Ti - 47.28 MH/s, 25/27 R:7.41%
20201212 16:27:41 GPU #3: GTX 1080 Ti - 48.71 MH/s, 40/41 R:2.44%
20201212 16:27:41 GPU #4: GTX 1080 Ti - 44.34 MH/s, 19/20 R:5%
20201212 16:27:41 GPU #5: GTX 1080 Ti - 49.35 MH/s, 32/35 R:8.57%
20201212 16:27:41 GPU #6: GTX 1080 Ti - 49.60 MH/s, 30/30 R:0%
20201212 16:27:41 Hashrate: 334.93 MH/s, Shares/min: 3.741 (Avr. 4.257) R: 6.278%
20201212 16:27:41 Uptime: 49 mins 8 secs | Algo: ethash | T-Rex v0.19.3
20201212 16:27:41 WD: 49 mins 10 secs, shares 209/223
20201212 16:27:55 [FAIL] 209/224 - Job not found., 320ms ... GPU #1
20201212 16:27:56 [ OK ] 210/225 - 334.96 MH/s, 88ms ... GPU #1
[...]
same on win 10 with Nicehash, v0.19.5, tested on gtx 1080 ti & RTX 3070
It happens to me as well when mining on NiceHash, Is that related to NiceHash itself? Or could be the card OC problem?
Still having this issue as of 01/23/21 NH will tell you it's because of your overclock, even though it still does it on stock settings.
Same issues when using a 1080 in Linux.
Ubuntu 16.04, Nvidia 450.57, Cuda 10.2, both 0.19.1 and 0.19.3 (which says it reduced rejected shares)
I see a high rejected share rate on pretty much all the GPUs, sometimes as high as 7%, all of them with
Job not found
coming from the pool. Other miners (phoenix, claymore, ethminer, gminer) have <1% and I can tell that in their case it is legit, because it happens right after getting a new job from the pool (so the pool considers it stale ... it shouldn't but still, that's another story).The worst is on Nicehash.
Basically, whatever 0-1% speed increase your miner has over other miners, it gets trumped by this high share reject rate ...
I have a hunch that something in your code submit the wrong job hash in cases where multiple shares are submitted almost at the same time (shares found very quickly in sequence) -- I tcpdump'ed it to have a look. The pool sometimes doesn't have time to ack each share in sequence. Strangely, however, it's not always the case: sometimes two shares submitted almost at the same time (few ms apart) does get two
result:true
acks from the pool, whereas other times it gets only one ack and ajob not found
. The two shares do have different job hashes, and different json-rpc message ids.Something you may want to look into. I don't know your code (I know ethminer's)