I've got this weird problem with stale shares rate for quite a while. 5-7% of stales to be precise (up to 10% on a 180Mhs 6 GPU rig at a time). At first I thought it might have something to do with GPU firmware, or process priorities, or claymore (been trying different versions, keep installing the latest with no improvement whatsoever). Tried to get rigs work through proxy, but then it became obvious that direct stratum is fastest way anyway. Network issues are out of question - I run gigabit switches to the router, then to the fiber optic router,
and get 79ms to the ethermine pool. CPU load on rigs generally around 2-5%, and I use raisers v.006c (quite standard stuff). Here's my screen
https://imgur.com/a/Xf8IDeM Any suggestions?
If you are getting high stakes then it's most likely the location relative to the pool.
What is the average ping when a share is submitted? It needs to be less than 250ms or you will get very high stale rates since ETH has very close block times.
If your ping is low then the other issue can be your computer. If your CPU is overloaded then it might submit the shares too late and hence give you stales
Thanks for reply, I even gave claymore 'above normal' priority, and it didn't make any difference.
P.S.: will try other miners and pools, to get a better picture.
What if you try phoenix miner. The dev claims that the most recent version gives 0% stales and many users confirmed that. I have not tried myself (also suffering from 5-8% stales). I'm just waiting for more people to confirm before I switch from claymore.
I've tried several things to reduce stales. Direct IP at first was promising solution, but later in evening that rig's stales went up along with other rigs, I guess it was Internet usage peak time for my ISP, although it did show lower stales for other times. I've tried phoenix miner - same eggs, just from the other side. I've tried different DNS server, also no difference since its used only once in a while, then address is cached. Then I've tried different pool with much less latency. Unfortunately that pool doesn't show stales, and I can only count shares accepted. So far, it looks promising, and I think its the only way to reduce stales. Just use your closest large enough pool. I'm only concerned about the resulting latency for accepted shares being different from that server's ping. Ping is 22ms, shares get accepted at various numbers, from 15ms(!) to 500ms, but generally 31ms (not 22ms). Hope this info helps someone.