Currently everything is still showing memory tempature currently except for the 3060 and 3070 which I dont think have every shown memory temaptures and only the 3070Tis stop showing temaptures sometimes (depending on which miner and version). I didnt catch your one question earlier but yes, when shifting cards between my two boards the gpu positions do change, so it may have been reading old information if the 3070 or 3060 was in the slot previously. I really think youre focusing on the wrong aspect considering this issue doesnt occur on my other board or my other two rigs.
Try setting very conservative memory clocks on all cards and see if it’s stable. 99% of instability is caused by poor/too aggressive clocks and or outdated software. (You’re screenshot shows you’re on an image from last November) before trying to troubleshoot anything always make sure you’re on the latest software.
Sorry for long response delay, since Im a new member the site limits my reply/post amount.
Do you know where I can find the latest BIOS then? I have checked online but as mentioned earlier there is no real specific area to find the BIOS for these boards. I found a bios page on both asus and biostar but both bios versions were dated older than the one in the screenshot. My HiveOS is also the most recent version.
Again on the overclocking, that would result in visible gpu/driver errors displaying on the worker screen and this issue would occur on my other boards too if that was the case. I can look into setting lower rates but I’d rather not lower my hashrates either.
I would appreciate if you could get fellow staff or other regular posters who may have additional btc b250 board experience.
Since time has passed, I can also give a bit of an update regarding the board.
It has currently been running for 16 hours straight with only the nbminer restarting itself once.
I have kept the monitor plugged in to see if there is anything on screen when / if the error ever occurs again.
Here is a list of differences which have resulted in one of the longer runs for the rig:
I have used PCIE USB ports marked 0-5 instead of 6-11. Perhaps one of the ports is faulty or the bios/kernel has a bug when the last six are in use without the first six being in use.
I have kept a monitor plugged in the entire time, making use of the boards on board integrated graphics and HDMI port. A lack of the HDMI port and integrated graphics being in use could also result in bios/kernel errors and thus crash the board.
For the CPU power cable, I have used the right side of the cable at the split rather than the left side. (The board only takes a 2x2 CPU power insert. rather than the full 4x2). While I do not think this would do anything, and I have tried multiple power supplies, I did change this and so I will note it.
I double checked and made sure 4G Encoding/Decoding was enabled along with some other graphics settings.
I will post another update after about 48-72 hours of the board running without issues. After this time, I will add GPUs one by one to the board (6-11) and see if adding to any one specific port begins to cause the error again. If the board crashes before this period I will post about the crash itself.
I have also reached out to other forums and communities; non-board proposed solutions have included using a smart switch to be able to restart the device any time remotely by switching its power off and back on. Additional solutions included recommendations for other, known to be more reliable, 8 or 12 GPU boards types. These include the standard BTC B250 with actual PCIE ports and smaller boards such as the Rebtech 8 GPU system. I have also begun to look into which of my preexisting boards, with 4G Encoding/Decoding enabled, can handle a full 8 GPU setup too.
I’m hoping to be able to solve this mystery once and for all for anyone who has run into this exact same / similar issue for BTC B250C as there have been too many internet forum posts where this goes unsolved.
Did you flash the latest stable image? Your original screenshot shows an old kernel. You’re on kernel #110 now?
I am running the latest HiveOS version. However, This issue with this board seemingly dates back for years on Windows and HiveOS mining systems though based on the previously mentioned posts.
The issue was not occuring before while running the older HiveOS version.
If the system crashes again that should confirm it is not a HiveOS version issue.
As for the BIOS I mentioned that I have tried to find more recent versions online but only find articles regarding issues updating the BIOS on BTC boards and uploads of older BIOS versions.
Before you make any comment about the 3090 giving a fan speed error I have had this happen before on specifically this card while it was in two previous rigs/boards.
There are currently no listed gpu driver errors or other errors on the dashboard.
You are still on an older kernel, #83, vs. current #110, while attempting to run all the modern HiveOS and new release miner features.
You are also on a USB drive, have updated the drivers at least once, and frankly, this process alone is known to create issues. We troubleshoot these daily in Discord. In addition, USB is also terrible when enabling logging, so don’t.
The fan error issue points to additional troubleshooting, and I’d be surprised if you ran the nvidia-smi tools to identify the many other potential issues with your rig. “GPU Driver errors” have nothing to do the the drivers.
I’d first suggest investing in an SSD. (M2SATA would be great, but I don’t recall that board supporting it). Flashing the latest stable release with Etcher, running disk-expand, updating to latest HiveOS version, enabling logs, then trouble shooting.
The board you have is known to be flaky across the wider mining community. If that is not resolved with the above, grab a TB360 Pro 2 or even an older H110 Pro and call it a day.(M2SATA in both of these). The old boards and even new old stock are now showing up at firesafe prices.
Side note: Seriously review Rebtech history before spending a dime, they have no ability to run current nVidia drivers.
Thank you. Youve been the most helpful so far. For updating the kernel is there a way I can do that from the terminal or do I have to force a reflash of the usb?
Edit: I have found the command to update version and kernel, I am now on version 110.
I can look into using an nvme / m.2 drive as I have a few spare but theyre all 512GB or 1TB and that feels like it would be a waste. I can look into getting 128GB nvme ssds though. However, you would be correct that this board has no nvme slot. (ver 1.1 does but 1.0 and 1.2 do not.) I have been running USB on all of my rigs with only one instance ( and already older usb ) dying after a power outage. I can look into using a standard SATA SSD as I do have a few 128 or 256 GB instances laying around but I havent used those as they cost more power than a USB.
I can easily run disk expand. Where can I enable logging?
I’ll look more into rebtech and hold off for now then. I’ll also look into a TB 360 Pro 2 or H110. Is there anywhere specific you recommend I get one of those boards from? Do these boards all have a limitation of Nvidia GPUs needing to be only the first 6-8 slots? I recall seeing some blueprints showing that PCIe slots 6-11 do not fully fupport nvidia gpus on the btc b250 pcie version.
I wanted to downsize rig amount to mitigate both HiveOS and electrical costs but I’ve also seen more information about people sticking to 6-8 GPU systems and eating the costs for general stability.
I’ll post another update after another 48-72 hours have passed since these changes or whenever the next post occurs.
Lower the mem clock on the 3090 if it’s crashing the driver and showing 0% fan speed. That’s probably contributing to your issues.
Congrats on getting it running on latest kernel :slight_smile, HiveOS, and most of your miner versions as well. I am confused on your miner selection, but we all have a strategy.
NVME drives will burn a PCI address, so be choosy on your motherboard choices vs. using old school SATA.
You really don’t want to run logs on a USB drive, the write speed is not your friend as it is, but here are the commands (logs-on) from hiveshell, shellinabox, or PuTTY:
My 12 GPU rigs are only built with very stable GPUs or there are too many chances for misbehaving cards creating issues and taking a large rig down.
I move the diva/complainer types to 8 GPUs or even far less as needed and tend to group them with other similar overclocking performers, etc. As the trouble increases, the power increases and the clocks go down. During high temp and low profit times, I shut off the worst performers.
Your group of GPUs will dictate the strategy.
Gotcha, I can look into reusing an old hard drive or ssd over SATA and enable logging on that, especially once I have this runing a full 12 GPU system I’m probably going to want that.
I’m open to miner suggestions. I’ve been using nbminer since it was one of the first with the LHR limit remover. TRex miner was recommended by a friendd since its seemingly an Nvidia specific miner. I have been keeping LHR and non LHR cards separate to avoid any issues.
I’ve been running these GPUs each for months now since I started mining. This was my first rig and I’ve just been trying to shift it to this other board. My hope is to eventually condense to a 12 GPU systems for my stable cards in order to save on some power and HiveOS costs.
So far, no issues since updating Kernel either. As mentioned before the only other differences is using 0-5 PCIE USB lanes and keeping the on-board graphics active by having the monitor plugged in to live monitor it for whenever it crashes (if it’s going to again). Perhaps one of these was the main fix, but I won’t know until I start testing without a monitor plugged in and/or testing with more than 6 GPUs if one of the PCIE lanes is bad.
Thank you again so much for your help. I’ll keep this forum updated as it has the most documentation and troubleshooting regarding this known issue with the BTC B250C boards.
for miner recommendations i personally run lolminer on all my nvidia and teamredmier for all my amd, mining eth+ ton on all but that only has a week left or so. lolminer has the lowest dev fee of the lhr unlocker miners at .7%
Gotcha,
I already use TeamRedMiner for my AMD cards on a separate rig.
Is lolminer able to run LHR and non LHR cards simultaneously and still properly unlock LHR cards without conflict?
Why is ETH + TON only lasting a week? I used to mine ETH + TON but stopped after LHR limiters were removed to not limit my ETH rates and reduce electric / heat.
You should always have a reserve rig which is up to date, 100% functional, and can be used for testing 1 or 2 GPUs the act up, require thermal pad changes, etc. You don’t want to test things in a big rig. Such test actions, can be done on Hiveon Pools for no cost and with the new adjustable payout options, no long term losses. Even when billed, they bill in less than daily increments and less than $0.10/day.
You may find the board you are working with now, goes into test bench mode and only utilizes the clearly marked ports that work.
Gotcha, sounds good.
Yeah I’ve been using that previously mentioned AMD rig to test any time I get a new GPU typically as it’s not a full rig. That rig is already running kernel 110 (which I didn’t even realize until going and checking / upgrading all rigs after yesterday) and latest nvidia driver. My plan was once I had a condensed rig that I’d repurpose one of the boards and old frames as a proper test bench.
I’d also appreciate your recommendation on nvidia miner. At this point I may switch to lolminer or use lolminer for lhr and nbminer for non lhr. Unless both can be run on one now.
I can count my nVidia GPUs on the fingers of both hands vs. a materially larger amount of AMD GPUs. With the level of bias toward AMD I am:
- 100% Team Red Miner for AMD
- 100% T-Rex for nVidia (you can see 100% unlocking and non-LHR detected above)
I’ll pay the extra dev fee to be able to log miner errors clearly between the two very different power tables, etc.
Full disclosure: I expect the future of mining to force me to rethink my approach as various miners add various new algos. Who knows what we’ll be doing in a few months 
I do expect I’ll be running multiple miners in my rigs for flexibility.
Thank you.
Yeah I’ll look into either using lolminer for cheapest and/or trex to reduce total miner count and extra logging/stats like you’ve mentioned.
I do appreciate AMD for the lower power consumption and may look into picking up another 5700XT or two for RR mode.
Next time I respond will be back on topic with either the board having crashed again or stating that Im past 72 hours. (The longest period I’ve seen recorded online of a b250c running before the error occurs.)
Still thinking that, assuming it is fixed now, it may have been that ports 0-5 need to be used first or that the on board graphics / hdmi need to be kept occupied. Could also have been the older HiveOS kernel. All good things for any future individuals facing this issue to try as solutions.
Running a very wide mix of motherboards here, Kernel #83 did not have stability issues, most of my rigs were running 30+ days until I shut them down for maintenance, fan cleaning, etc.
The OS either runs, or it doesn’t. Power, heat, hardware, miners, and settings are the issue for a running rig.
Side note: Did you see this, definitely dig into your R mode particulars, deeply:
Hey there,
The rig has now been running 24 hours without issue so far. However, people have reported the weird issue of this motherboard losing connection to mining systems but remaining in an on state (ethernet lights flash, gpus run, etc) can take up to 70 hours before it occurs. One day down and two to go essentially.
I do not currently have the amd gpus running on a BTC B250 based board and do not plan to, I do not have enough to warrant running them on a 12 GPU board. However, I will keep that setting in mind.
I also looked into Rebtech and you’re correct, they are now only recommended for AMD rigs as they do not support the latest Nvidia drivers.
The rig ran for 2-3 days without issue.
I have since proceeded to combine my second rig with this rig. I am currently running 10-11 GPUs on the board. I encountered some minor, unrelated issues (Nvidia Driver error 45 and 62, these being “not enough power” and “overheating” respectfully). The combined rig is coming up on 24 hours of running and has not had issues (besides initial setup ones mentioned).
I have also learned more about HiveOS watchdog and properly enabled that for when LA gets above 5 (instead of the default 25) or when hashrate drops signaicantly below standard.
Back to the main topic, here is a list of all major differences between when the board would encounter the blue screen / connection loss issue and it now being in a seemingly stable state. I am hoping that anyone who encounters this known issue witht he board will be able to fix it after trying these.
- Updating to HiveOS Kernel 110 (try latest version and kernel for any system if this connection loss / blanking out error is occurring)
- Keeping a cable or dongle in the HDMI port thus keeping the on-board graphics active
- Ensuring that the first six slots labelled 0-5 (or labelled PCIE 0-2, USB groups of 2) are used first.
- Re-seeding the motherboard and CPU power connectors (using the right-side half of the CPU connector this time)
As an additional note my overclocks are the same however I did reduce some of their core clocks as I was able to get the same hashrates with lower clock and a small decrease in temaptrues.
Special thanks to Grea as you have been the largest help here and allowed me to learn more about both HiveOS and mining than the original problem entailed.
If the problem happens again I’ll make a post but for now, Im signing out.
2 Likes