Hi, claymore, i have a lot of 6850 and 6870 cards. the hashrate is too low. 6X6850 : overclock to 850, got 960 h/s 6X6870 : not overclock, got 1020 h/s is this the good h/s? help. thank you.
1GB cards show bad hashrate, 2GB of video memory is required to achieve best performance.
|
|
|
Dunno if Dev follows this thread or not but I figured I would pass my findings along...
I have 2 6XR9-290 rigs. I downloaded CryptoNote Miner 5.1 beta a week or so ago and started mining XMR. I am running 13.12 modded drivers and have my rigs dialed back to 924/1250 @ powertune -15 (voltage and fans controlled by MSI Afterburner) to keep wattage and heat down. My temps and fans average in the 70C/70% with the reference style cards and 64C/63% with the 6 "Gigabyte Windstream OC" style cards at around 1500 watts per miner. With standard v5.1 files and these settings, I am pushing around 3690H/s with the reference cards and around 3740 with the Windstream OC cards.
So I thought I would download the latest v6.1 and see if there were any new optimizations that would increase hash. What I found was that the v6.1 lowered H/s by about 100H/s per unit. Thinking maybe I needed the 14.6 drivers, I tried the trick I learned from SGMiner v5 where instead of replacing the rock solid 13.12 video card drivers with the 14.6 drivers, I just unpacked the 14.6 open cl files and copied them into the CryptoNote directory where the mining executable was. I could tell a huge difference right off the bat with one rig pushing 3880Hh/s and the other 3790H/s. But within about 5 minutes the hash rate was back down to about 75-100 less H/s each than the older v5.1 averaged. I wasn't happy with that so I rebooted the machines again and went back with the v5.1 install. On a whim, I copied the 14.6 open cl files out of the v6.1 directory into the v5.1 directory to see what that would do... Wow! Nice little 50 H/s boost out of each unit! Now the miner with the reference cards is averaging 3740 and the Windstream unit is pushing 3790.
Not sure why it works so well like that, but there it is!
Appreciate your efforts!
I can test miner only on my rigs and v6.1 is the best for me. I get completely different feedbacks for every version - some people like v4, some people like v5, some people have problems with all versions except v6.1. Miner uses video memory a lot, it causes problems or unstability on some systems. I tried many different approaches in code in different versions so you can select the best version for your configuration. Unfortunately, it seems there is no way to build one version that will work on all configurations properly, AMD drivers often work strangely if miner has to use a lot of memory, especially on multi-GPU systems.
|
|
|
Claymore, I have noticed that the program will error out very ungracefully (looks like a memory leak) if the internet connection gets dropped while the miner is attempting to either submit a share and/or reconnect after a while with no sucess. It seems to happen almost every time if the internet has been down for more than a couple hours.
You should be able to replicate this problem by starting out the miner connected and hashing, and then disconnecting the internet and waiting for a day or so.
I will test it now.
|
|
|
Claymore we need a fix for this miner getting restarted over and over please
Ok, I'll release update in a day, I'm going to add an optional mining mode, a bit slower (1-2%) but with less memory usage, also "-r" option to restart miner automatically.
|
|
|
Problems with Minergate are on their side, hope they will fix them soon. You can check the log to confirm it, right now Minergate sends "blockexpired" error for current job but does not send new job for some reason, so miner cannot update job and sends stale shares, and Minergate responds "invalid jobid" for such shares. I sent a message to them two days ago, wait for answer.
|
|
|
Missed required command line options!
64.bat NsCpuCNMiner64.exe -t 4
pools.txt stratum+tcp://mine.cryptoescrow.eu:3333 47sTZ1ovnAk2Kir1Jjit5vSnL53rHSsCEjYGmyktkuAThsX7h7QAKE4MNuBsAQAvPr9aDxg9rWLqPj3 inb3T4MuHCseCNaM x
how to use pools.txt
thanks!
64.bat NsCpuCNMiner64.exe -t 4 -u 47sTZ1ovnAk2Kir1Jjit5vSnL53rHSsCEjYGmyktkuAThsX7h7QAKE4MNuBsAQAvPr9aDxg9rWLqPj3 inb3T4MuHCseCNaM -p x pools.txt stratum+tcp://mine.cryptoescrow.eu:3333
|
|
|
Been running 11 for weeks no problems. Installed 12, it crashes my drivers immediately. Running 13.12 3 x r9 280x saphire toxic. windows 7 64bit
Claymore any ideas how to fix?
v12 uses more video memory. Have you set GPU_MAX_ALLOC_PERCENT 100?
|
|
|
v3.3 is available:
- Added Dashcoin, OEC, MCN and ORION support. - Added "-r" option. - Added "-nofee" option. - Fixed some bugs.
|
|
|
just to let you know 4 cards run fine on 4gb ram on windows 8.1 but 5-6 cards need 8gb...
at least for me;)
I've been running 5 290s on some of my rigs with 4GB ram with no real problems, so i'm not entirely sure that's true. It's got more to do with available virtual memory, i surmise. Perhaps Claymore can comment on physical ram requirements, i would be curious to know. I cannot find exact requirements. In most configurations it works with 4GB RAM with any number of cards. But sometimes 4GB is not enough. Virtual memory definitely helps, but if you have only 2GB of RAM it won't work in most configurations anyway. It is something related to AMD drivers and I cannot find any public documents about it.
|
|
|
any chance that 5xxx cards can be optimized even more? Is the problem the 1 GB vram for 5xxx cards?
Yes, video card must have 2GB for good performance.
|
|
|
My fault for "miner" I thought the Minner 6.0 provides more hash than in version 4.0 or 5.0, while the pool shows less power than the 4.0 or 5.0 although in version 6.0 acceleration 2-3%
You can use v4 or v5 if you think that they are faster, but it's not true. Even when I use same version I see 10-20% speed deviation in hashrate that pools show, it looks like pool-related issue, latest versions of the miner are not slower. If you know some pool that shows exact hashrate let me know and I will test all recent versions on it.
|
|
|
Important update v6.1:
- Fixed memory leak.
|
|
|
Clay did you notice that the miner 6.0 gives more hash in the mineru however the pool gives 10 to 15% less real hash. Many users have this problem. After 12 digging there has been a lot fewer coins excavated??
I don't know what is "mineru", sorry. Miner shows "raw" hashrate, it is always more than hashrate that pools show, it was already discussed here a few times. Also some pools send new jobs with big delays often, it cause a lot of stale shares and pools don't indicate them at all (except Minergate).
|
|
|
Upgraded to v 6.0 but still got a lot of these: Error in server response : {"id":1,"jsonrpc":"2.0","error":{"code":-1,"message":"Lowdifficultyshare"}} Stratum - connecting to 'xmr.crypto-pool.fr' <212.129.27.50> port 6666 Stratum - Connected Pool Diff 7500 Don't overclock your cards too much.
|
|
|
Claymore GPU 5.1 on merged mcn-mro.pool.minergate.com:45660 not working:
Fixed in v6.0, it still shows some errors but it works properly. CPU miner will be updated soon too.
|
|
|
v6.0 is available:
- Improved mining speed by 2-3% for all cards. - Reduced number of outdated shares dramatically. - Improved "-a" option, now you can specify mining mode for every card separately. - Fixed bug with Stratum support for Minergate pool.
Link Mega is still down for about a month now & google drive link doesn't contain v6.0 ! Uploaded to google disk too, check it now.
|
|
|
v6.0 is available:
- Improved mining speed by 2-3% for all cards. - Reduced number of outdated shares dramatically. - Improved "-a" option, now you can specify mining mode for every card separately. - Fixed bug with Stratum support for Minergate pool.
|
|
|
does minergate still have stratum issue?
They fixed their bug but then also found bug on my side, I will release an update today.
|
|
|
is it possible to run this miner on a 2gb ram on the pc?
i dont believe it is. whats the solution then?should i again buy 16gb ram? I use 4GB RAM in my rigs. 2GB is too small, it will fail on most configurations, there is no way to fix it.
|
|
|
|