I have "No devices are set" on RaveOS. Adding "-d" didn't help. What should I do?
Fixed in the 1.21 version. But you need to upgrade the driver version to a version that can run cuda 11_4 code. Ascii bug is still there in 1.19, but now it can be seen for only a few seconds, not for 1 min like in previous versions.
Is it working in v1.21? Please make possible to see "normal" running time in hours and days, right now i have to use a calculator just to see how long this miner is running, it is not that convenient.
fixed in v1.21 |
|
|
|
Is it possible that with option --no-cpu i have more stales shares?
no, but you can decrease the --xintensity. Try to run with --xintensity 24 (amd)
Another problem is with --lock-mclock on windows. Can't get 8300mhz (+1500 in afterburner). Max i can get with this command is 7800mhz. Any advice?
Only use this option on the --lock-cclock for now. And msi afterburner for clocks, we need to add more code to fix this. [moderator's note: consecutive posts merged] |
|
|
|
I am using SSL on both Trex and TBM.
SSL have known issues. you should use the normal port
But pool side the rate 20-40% less after a couple 6 hour tests. Around 1% stales which are paid at 80%.
I get around +5% in tbm poolside vs T-rex, so something is wrong in your setup. Mine on high intensity.. epoc 450 can never give you 32Mhash on the gtx 1070. Normally 25-29mhash.
Is there any benefit to adding this parameter with a high xintensity value on RTX 30XX cards in 2Miners?
You can increase the speed with higher intensity. Why don't you try it out. What numbers are you getting? --xintensity 224 (default) --xintensity 256 --xintensity 288 --xintensity 320 . . . [moderator's note: consecutive posts merged] |
|
|
|
|
Others have managed to run 1.20 without any problems. 1.20 was released a few hours ago, why don't you wait until hiveos add it.
|
|
|
|
|
perhaps you need to give execute permissions.
$ chmod a+rwx tbminer
|
|
|
|
with TBM 1.20 finally i can make work at least one 1060..other one shutting down (i attach the log) and so my 9 nvidia p104..
GPU0 0GB added to miner as Cuda device Something is wrong with your cuda driver because your card has 0GB memory. Download the latest driver (800mb) and install it. Make sure you inclide all the files and not only a few dll's.
any resolution for this issue ?
There is a bug in 1.20 that if you use a worker on ethermine with high case letters, it wont show as reported hashrate. this has been fixed in v1.21. So the workaround is to rename all your workers to lowercase.
I'm not so sure. All other miners are working fine and detects RX580 and RX5700. If it was related to HiveOS, all others miners should also have the issue.
No, same problem on NVIDIA cards, our miner only works on the latest drivers and toolkit. The other miners work on older drivers, our optimized miner don't. [moderator's note: consecutive posts merged] |
|
|
|
Is it possible to disable oc reset with --dagintensity?
Probobly you don't need to set the memclock with tbminer when you use a low --dagintensity. So only lock the coreclock. You can test it.
FYI, it seems that there is still some bugs just on AMD command line logs (missing stats, wrong gpu names, etc.).
I guess you're aware of it.
yes. We are going to change the library we use for AMD stats. [moderator's note: consecutive posts merged] |
|
|
|
Thanks, I'll test these values.
What about --dagintensity ?
Since the verification code doesn't fail, you don't need to use this. But you could use to increase the memory clock and put --dagintensity 0. This option is for rtx 3060ti,3070,3080 with extreme clocks. But you seem to be happy with 102MHASH. Higher clocks higher power needed. |
|
|
|
updated to ver 1.20
which xintensity do you suggest for Crazypool now?
At the moment I'm at 160 for RTX3000 cards and 32 for AMD RX6000 (if I put -1 I get too many stales)
Default 224 for crazypool on nvidia. is 60 too much stales for the rx6000? 60 works good on the 580. 78 for the 6800 |
|
|
|
64 mhs on 1070Ti?! It's real?!
No. We had a bug where we added hashrate while the program was verifying the dag file. Validating is much faster than searching, so perhaps this is the bug. I asked him to reproduce in v1.20, since he is running an older build. Could also be photoshop? |
|
|
|
Tested 1.20 on HiveOs and still not detecting RX 580. I fingercross for the 1.21  I believe this is a hiveos problem. In windows it hash at 33-34 MHASH at max clock hynix..102 watt..
Any plans on 11.5 cuda for Linux or does it help much compared to 11.4?
The 11.5 build is more stable when applying memtweak (the pill) for gtx 1080ti, 1080. [moderator's note: consecutive posts merged] |
|
|
|
|
Why don't you try v1.20 and see if you can reproduce it.
|
|
|
|
Did you try running xintensity 224 on 2Miners or is 768 better?
Higher intensity seems to be better on 2miners, but not too high. Most of my test cards are old and with bad cooling. I would need two or more identical cards on the same clocks to do a better test.
The problem with this is that we have to factor in the pool luck rate and correlate that to your miner luck rate per block and then adjust the actual block reward to remove the MEV and Tip variability to get a Pool to Pool, miner to miner equivalence.
Yes. Some pools doesn't include the pool luck rate. When comparing two pools the luck is important to calculate a proper estimate.
We just changed to epoc 450. All my test rigs survived the epoch change on v1.20. If you got crashed rigs because of high clocks, add --dagintensity 0. a possible fix [moderator's note: consecutive posts merged] |
|
|
|
Here's one of my 3090's in a test rig with just the default Xintensity:
My results while testing --xintensity 224 crazypool: 0-1% stales nicehash: 0-1% stales ethermine: 1-2% stales flexpool: 2-3% stales I had to lower the -xxintensity down to 144 to get 0-1% stales on flexpool, but 144 lost a few Mhash'es in the rig. So different pools have different optimal intensity settings As a miner developer it's hard to set the "correct" one. It require alot of testing. |
|
|
|
|
Show Posts
