Categories
- 17.3K All Categories
- 9.5K Mining
- 574 Pool Discussion
- 378 Promotional
- 1.4K General Project Discussion (non-technical)
- 485 Education
- 811 Protocol and Client discussion
- 170 web3-js
- 29 Whisper
- 16 Swarm
- 3 RLP
- 303 IoT & Hardware
- 1.2K Smart Contracts and Dapps
- 28 Serpent
- 359 Solidity
- 664 Projects
- 1.2K Reference clients code and builds
- 249 Eth & AlethZero- Cpp Implementation
- 471 Geth - Go Implementation
- 242 Mist
- 15 Node.js Implementation
- 36 Python Implementation
- 49 Mix
- 36 Other Implementations
- 170 Meetups
- 40 Other Events
- 226 Jobs & Skills
- 281 Press and Articles
- 75 Audio/Video
- 296 Ether Sale
- 1.2K Other Languages
- 96 Chinese
- 255 German
- 33 Italian
- 111 French
- 3 Hebrew
- 42 Japanese
- 75 Portugese
- 46 Romanian
- 185 Russian
- 231 Spanish
- 47 Turkish
- 125 Watercooler
HDMI Dummy Problem: RX580 rig uses more power and Claymore gives memory errors
The RX580 cards in my rigs are not reporting temps or fans to Claymore when nothing is plugged into any of the display ports of the cards. My 1070 ti cards don't have this problem and correctly report fan speeds and temps regardless if there are HDMI dummies plugged into the cards.
I have tested this twice now with the same issue.
When I plug the HDMI dummy into one card and remote into it with RDP Claymore will not show fan speed or temps in the console or from another computer via Claymore Manager. Even if I restart Claymore the stats are not displayed. I then close RDP and VNC into the rig. When I start up Claymore I do get temps and fan speeds. Claymore Manager also report the speeds. But Claymore then constantly restarts because of memory errors. I do have overclocking set pretty high and the RX580 cards are mining at around 31.65 - 31.70 mh/s each.
I then close the VNC app and remote back in via RDP. When I then start the mining software my UPS errors out because the cards are drawing more power than they should. They are also mining at lower speeds, around 30.5 mh/s.
The only solution I have is to remove the HDMI Dummy and reboot the rig. When it first happened a few weeks ago I did not realize what was going on and ended up reinstalling the drivers. However it appears not to be necessary and once the dummy is removed and the rig rebooted all is fine again, except I can't see the temps / fan speeds.
I installed Awesome Miner and that does allow me to get those stats from the rig but you have to go through a bunch of steps for each rig to get the data. It would have been really great if Claymore's Manager could report these stats without the use of a dummy.
It seems the 4k dummy places a lot of stress on the card because it now thinks it should display an image. Not sure if anyone else has noticed this. I think you can go further with overclocking when the cards don't have any monitors or dummies plugged into them.
I have tested this twice now with the same issue.
When I plug the HDMI dummy into one card and remote into it with RDP Claymore will not show fan speed or temps in the console or from another computer via Claymore Manager. Even if I restart Claymore the stats are not displayed. I then close RDP and VNC into the rig. When I start up Claymore I do get temps and fan speeds. Claymore Manager also report the speeds. But Claymore then constantly restarts because of memory errors. I do have overclocking set pretty high and the RX580 cards are mining at around 31.65 - 31.70 mh/s each.
I then close the VNC app and remote back in via RDP. When I then start the mining software my UPS errors out because the cards are drawing more power than they should. They are also mining at lower speeds, around 30.5 mh/s.
The only solution I have is to remove the HDMI Dummy and reboot the rig. When it first happened a few weeks ago I did not realize what was going on and ended up reinstalling the drivers. However it appears not to be necessary and once the dummy is removed and the rig rebooted all is fine again, except I can't see the temps / fan speeds.
I installed Awesome Miner and that does allow me to get those stats from the rig but you have to go through a bunch of steps for each rig to get the data. It would have been really great if Claymore's Manager could report these stats without the use of a dummy.
It seems the 4k dummy places a lot of stress on the card because it now thinks it should display an image. Not sure if anyone else has noticed this. I think you can go further with overclocking when the cards don't have any monitors or dummies plugged into them.
Comments
If you running on Windows just keep it simple and use MSI Afterburner to control Fans (set fan control and adjust graph there).
Is first time I heard about using HDMI dummy plugs on a AMD RX580. If it works perfectly without them why you need complicate it?
With Awesome Miner I discovered that two RX580 cards on one rig had fan speeds of 65% and the rest were at 50%. I could have logged in remotely via RDP / VNC and checked it with HWinfo, Trixx etc. but that is time consuming. It is much easier to keep tabs on fan speeds and temps in a small interface like Claymore Manager that reports them all in one line. Have you ever used Claymore Manager? Without the dummy the RX580 cards do not report fan speeds and temperatures to Claymore's miner.
Note i'm using Windows 10 and the newest AMD drivers (18.x) Also using Claymore 10.5.
I'll try Teamviewer to see if that changes it. I don't like the latest AMD drivers because you have to set compute mode individually (six times in my case) for each card in the rig. And even then I have found that the cards are drawing more power and maxing out my UPS. The best driver in my experience has been the block chain beta driver from AMD. But maybe that is why temps are not reported as yours.
I have found another workaround. HWinfo64 will allow you to connect networked computers so you can monitor all their sensors. And you can turn of sensors that you are not interested in. I'll be able to monitor all rigs with that software from my personal computer and get fan speeds, temps, power usage etc. Pretty cool software.
You'll have to let me know if team viewer works any better. Starting on another rig today, picked up a RX 480 8gb card locally for $220!
What's really cool about hwinfo is you can have multiple columns side by side so it would be possible to view a ton of GPUs on one screen. I use a 40" 4k which helps to monitor things.
Claymore Manager is still the best though to get a quick overall view of the cards in a compact interface. There is some room for improvement but it is very handy. I wish the Zcash miners had something comparable.
Awesome Miner is also very cool and can work with most miners but not free so I'll pass on that.
Not a bad deal with that 480! Pretty soon you'll be pushing 2 GH/s
"HKLM\SYSTEM\ControlSet001\Control\Class\{4d36e968-e325-11ce-bfc1-08002be10318}\XXXX"
where "XXXX" is the card number.
for example:
HKLM\SYSTEM\ControlSet001\Control\Class\{4d36e968-e325-11ce-bfc1-08002be10318}\0001
HKLM\SYSTEM\ControlSet001\Control\Class\{4d36e968-e325-11ce-bfc1-08002be10318}\0002
But the problem is I get memory error messages from Claymore and the hashing keeps resetting. This is because I am running an extreme memory setting in Claymore's command line. When I remove the higher memory speed setting it works fine and report temperatures.
When I connect via RDP and start Claymore the cards hash fine with the extreme settings but no temperatures or fan speeds are reported.
I have two rigs running with extreme settings (190 mh/s for 6 cards) and 1 rig with the reduced memory setting (186 mh/s for 6 cards) but with temps / fans speed reported. Will be comparing how they perform on Nanopool over a week and see if the extreme setting really translates into more profit.
Tried this tool in link below which sets the registry settings automatically.
https://bitcointalk.org/index.php?topic=2815803.0
The registry is set but it makes no difference. Does the computer have to be rebooted or shutdown for it to work? We rebooted also and even then the cards were in Graphics mode after the setting. Only solution is to set manually to compute mode and I am not going through that mess.
The trusty Beta Blockchain driver that I prefer could not be used in this instance because Windows crashes after logging on. I am thinking maybe that driver does not support more than 8 GPUs but Windows would crash even with 5 Sapphire and 3 Asus cards.
Last night after I made this post I decided to try 18.2.1. Installed without issues and no error messages, ran the automated hack to fix the registry. Tried Claymore and cards were still hashing at 20 mh/s except for one or two that were around 31 mh/s. Ran the automated tool again to reset them all to Graphics Mode. And then set them all back to Compute Mode. Rebooted the rig.
Ran Claymore again and one by one the cards started to hash at over 31 mh/s. Some would be at 20-26 for a minute or two but eventually they all settled in at 31.4 - 31.7.
It was 18.2.2. That driver gave us errors when Windows first started (missing dll files etc.). I removed the driver and tried 18.2.1 and got it working finally after running the automated registry hack!