Launching a 12 Card AMD RX580 Windows 10 based Mining Rig - Blog

Load up 6 of the cards with the stock bios and test just a few cards with the stock bios. Start adding cards, see where it breaks. I def prefer adding 1-2 cards at a time rather than a full rig at once.

Yep I do one at a time, flash bios, test hash, remove card and start with next one. It takes longer to set up a rig, however the time saved troubleshooting one card/riser is a blessing.
 
Sorry to hear you having problems. I have 3 rigs running 12 cards each. But the motherboards are all different and none of them are 12 boards. I am looking for an answer for you. I really think it's a bios problem but not for sure. Gotta look some more.
 
Interesting
I had that exact issue at first.
But I noticed that the fast scrolling text indicated an issue with PCI-E 1 and PCI-E 2. So I shut down, unplugged both 1&2 cards, and rebooted and it just went to a black screen instead of scrolling the errors.
I then swapped out those two card's risers and added each card back one back one at a time. I then had no more scrolling errors, but still just a black screen that never progressed.

So I chalked it up to two more bad risers (5 total bad risers out of about 35 risers) - and then gave up - because with nothing more than a black screen what do you do? I logged in at simplemining.net but never saw my rig appear.



I suspect the problem they allude to in that writeup is just a bandaid on a sword wound. Meaning the way I read it, you have something failing, and you are eliminating the error notification, rather than figuring out what is failing. The micropause issue seems VERY much relevant to what I am experiencing in Windows.

I think I still have something failing - I don't know what yet. I wish I knew a way to specifically test each individual PCI-E riser/card. Like a quick cinebench type tool that you could run on each card individually and see what card/riser has issues.

Airbrushkid, do you use auto, gen1, gen2, or gen 3 for your cards in your 12 card rigs?

I noticed in another thread that now with 1709 Windows update, supposedly Windows 10 can run 12 Nvidia cards. Sure seems like 12, 1070 cards would be less trouble than the 12 RX580 -- at least from my vantage.
 
Use gen2 for PCIe.

Gen 1 has not worked for me on my 6 cards rigs, and gen 3 used too many lanes. This was on my Z270 boards, none of which are mining specific boards.

A suggestion I saw on btc talk, was to use hot glue on the riser connections. Test each card with its bios and a riser, once it tests good, put hot glue on the riser-usb connections.Then once you have the cards situated on the board, hot glue the riser to the mobo pcie slot. It is removable, but it prevents vibration and such. I started doing it on my rigs and I thing it is helping.
 
I run PCIE Auto and Simplemining OS. I don't like Windows for mining. But for gaming yes. Mine been running non stop for about month now. And for risers I've had 20 out of 100 go bad so far but I haven't tried them all. The riser, usb/pcie x1 adapter card and usb cable can go bad.


Interesting
I had that exact issue at first.
But I noticed that the fast scrolling text indicated an issue with PCI-E 1 and PCI-E 2. So I shut down, unplugged both 1&2 cards, and rebooted and it just went to a black screen instead of scrolling the errors.
I then swapped out those two card's risers and added each card back one back one at a time. I then had no more scrolling errors, but still just a black screen that never progressed.

So I chalked it up to two more bad risers (5 total bad risers out of about 35 risers) - and then gave up - because with nothing more than a black screen what do you do? I logged in at simplemining.net but never saw my rig appear.



I suspect the problem they allude to in that writeup is just a bandaid on a sword wound. Meaning the way I read it, you have something failing, and you are eliminating the error notification, rather than figuring out what is failing. The micropause issue seems VERY much relevant to what I am experiencing in Windows.

I think I still have something failing - I don't know what yet. I wish I knew a way to specifically test each individual PCI-E riser/card. Like a quick cinebench type tool that you could run on each card individually and see what card/riser has issues.

Airbrushkid, do you use auto, gen1, gen2, or gen 3 for your cards in your 12 card rigs?

I noticed in another thread that now with 1709 Windows update, supposedly Windows 10 can run 12 Nvidia cards. Sure seems like 12, 1070 cards would be less trouble than the 12 RX580 -- at least from my vantage.
 
As a sidenote, I have a 5x card RX rig that recently went stupid when I switched from ribbon risers to usb risers. I have had all kinds of stability issues that went away when I turned off the box fan blowing on the rig. I can only assume it was related to vibrations induced from the airflow affecting the risers. Now that it is stable for the first time in weeks I am afraid to shut it down to hot glue all the cables.
 
FINALLY

First time with all 12.

I think I’ve still got a bad riser in there. Maybe on the fourth card left to right in wattman because it took me a half dozen restarts to get that one to switch to “compute mode” in the AMD drivers.

BUT. They are all working at the moment

A few notes:

This project wasn’t at all worth the level of effort. I’ve literally spent probably 100 hours + on this one rig and it’s not quite right yet. I obsess about things when I get started and spend way too much energy on something like this. My wife literally started crying saying this was all I’d done for last two weeks straight when I was home from work. Not good.

Unless you can tinker with something endlessly this project isn’t worth pursuit unless you get lucky -- with the state of current drivers, support, and lack of existing documentation out there to draw from. Maybe give it six months. I’ve found that the more GPUs one engages the more headache there is. The four to six GPU rigs seem to be pretty much problem free.

Two 850 watt power supplies are insufficient for 12 RX580 cards

I ended up adding a little 450 watt power supply to run two of the cards and the motherboard and the 450 watt power supply is insufficient. If the undervolt fails and wattman resets the 450 watt PSU fails. I’m single mining Eth right now with 1130 MHz GPU clock and 2100 MHz memory clock. Single mining ETH to save on power use since my PSUs can’t handle dual mining power draw. I think you’d need two 1200 watt PSUs or three 850 watt PSUs to have the safe power to spare if the undervolt fails.

The windows UI is still buggy right now. I can’t click start button and I can’t launch IE. Remove all cards and the Windows UI issues disappear. I think this is the remaining bad riser at fault but I don’t know how to narrow down which one it is.

I’ve had five out of 30 risers fail or DOA... and probably more temperamental. I’ve got five different kinds of risers. I can’t say any are any more reliable than the next brand.

I’m still relatively new to the world of mining. Started dabbling in June, fired up more serious in late August with my first dedicated rig of eight 1080TI. Then a rig with eight 1060 and two RX 580, now a rig with 12 RX580.

More to come.

C8B6249B-7418-412C-982D-E232DB181BC2.jpeg
 
Last edited:
I have found no need to run afterburner undervolting if the bios is modded for 900 voltage (0.900), and a clean ddu/reinstall of AMD drivers is done. If the bios is modded after drivers are installed, Wattman will keep the old voltages applied even if you never open it in the driver config.

Something to consider. My measured power draw with a single card connected to a kill-o-watt was 110-115 watts on Eth going up to 145 watts dual Eth + lbry. That was with the mobo on an atx psu and the gpu being powered by a server psu + breakout board connected to the kill-o-watt.
 
I'm happy your getting somewhere with your rig. I was going to build a AMD rig but after all the post and chats on the mining forums I'll stay with Nvidia they seem to be very easy to set up.
 
I have found no need to run afterburner undervolting if the bios is modded for 900 voltage (0.900), and a clean ddu/reinstall of AMD drivers is done. If the bios is modded after drivers are installed, Wattman will keep the old voltages applied even if you never open it in the driver config.

Something to consider. My measured power draw with a single card connected to a kill-o-watt was 110-115 watts on Eth going up to 145 watts dual Eth + lbry. That was with the mobo on an atx psu and the gpu being powered by a server psu + breakout board connected to the kill-o-watt.
Thanks for the input.

What GPU clockspeed are you setting to 900v? I know where to set that in the BIOS, but didn't take it that low.
900 volts is pretty low I thought. Are you finding you can consistently get a rack of cards that low, or are you individually setting voltages that low.

Whatever GPU-Z shows for the GPU only draw, I think you can safely add 50 watts to that for the at the wall measurement. With dual mining My graphics cards are 130-140 watts in GPU-Z --- putting them at 180 to 190 watts from the wall.

That's with the clock speed limited to 1130 in wattman (which does reduce the voltages automatically) - but no voltage adjustments otherwise done manually.
 
I can consistently run at 900 with the core around 1130. Some are fine at 1155 but it makes no difference on Eth speed. GPU-z / afterburner consistently show ~880-890 when set to 900.

Yeah I don't go by what GPU-z shows for power draw. Like I mentioned, I had a single RX card plugged into a killowatt, so that was nice to test voltage changes and clockspeed.
 
I'm happy your getting somewhere with your rig. I was going to build a AMD rig but after all the post and chats on the mining forums I'll stay with Nvidia they seem to be very easy to set up.

There's just more moving parts with AMD. I think once you get it squared away they can be very good. But until you update the BIOS your only going to see about 19-24MH/s with a RX580 at full tilt. Now they've got compute mode you need to enable in the driver control panel to get significantly more MH/s -- but you can only enable it one card at a time, and have to restart the AMD driver control panel again each time. I'm not convinced Wattman is reliable about retaining settings - sometimes it does sometimes not - and I don't know what triggers it to lose settings with reboots/updates. afterburner is hit or miss with AMD on whether you can even control the AMD cards. The current afterburner 440 can't even control my AMD cards with the 17.11.1 drivers.

Getting a good AMD bios is half the initial battle. I made the bios for these cards I'm using by pirating some of the timings for other cards and using a few tips off anorak.tec That's all a bit of learning curve too, but I think I have the basics down now.
Once you get the AMD rigs going they seem to be very stable. I've had two RX580 going for over a month without crashing one time on my 10 card rig - but the 1060's that were on the rig with it have crashed several times over that span. (two instances of nice hash running - one for eight 1060's, one for two RX580s -- same 1200 watt power supply shared for all 10 cards)

But all in all, Nvidia's run great right out of the gate - turn them to 65% power target, overclock memory to just shy of 600Mhz, and return clocks to factory speed using the GPU Mhz boost and your off to the races. >= ~100% stock performance at 65% power draw - overall very reliably too.
Other bonus for Nvidia is that afterburner has always worked with Nvidia (that I've seen) and it's very reliable in keeping the power draw down to that 65% target.

If I was going to repeat this project it'd be with 12 1070 cards instead now that I read Nvidia also supports 12 cards in Win 10. That wasn't a known thing just a few weeks back when I started this thread.
 
Last edited:
What got it all running? Just swapping risers?
yes.
Apparently I had at least four intermittently bad riser cards of 12 -- and I have at least one more I haven't found out.
Simplemining OS helped me root out those first two slot cards that I hadn't been able to determine with windows with that scrolling PCI-E error issue, but it can't seem to help me ferret out this last one, because it's just a black screen now - with no progression beyond.

I'm hoping to just let the eth mining run until one of the cards fails - that'll be my suspect riser replacement.
if hashrate just slows - who knows -- cause I was dealing with that before and I'd swap out the riser with the hash rate that slowed to like 6MH/s and it wouldn't fix the problem.
 
Yea I've found while they don't have the same high hash rates as modded rx580's, the 1060/1070 line sips power and is way easier to set up. All of the rigs I've been building lately have been with nvidia cards, running on nvOC. Super simple
 
anyone following along and interested in a strong PSU for a build like this - I saw this today.
https://www.newegg.com/Product/Product.aspx?item=N82E16817152059

RaidMax Vampire 1000 watt Gold Rated PSU for $60 after rebate at newegg.

That's unbeatable. I bought two more for my mining operations. (I already had one of these PSU's and was very happy with it - it held up pulling 1050 watts from the wall for days on end until I balanced the load out more with another PSU).
 
anyone following along and interested in a strong PSU for a build like this - I saw this today.
https://www.newegg.com/Product/Product.aspx?item=N82E16817152059

RaidMax Vampire 1000 watt Gold Rated PSU for $60 after rebate at newegg.

That's unbeatable. I bought two more for my mining operations. (I already had one of these PSU's and was very happy with it - it held up pulling 1050 watts from the wall for days on end until I balanced the load out more with another PSU).

Just a heads up -- my experience with RaidMax isn't that great, I had an 850W psu of theirs... the cables themselves became insanely brittle after about 1 year. To the point where if you moved the cable the PVC plastic covering individual wires cracks and breaks off, in addition, the copper wire inside started to oxidize and basically fell apart. Their customer service is useless if you just want to buy another set of cables. Because of this I'm now 100% a corsair man in terms of PSU's (their support is unbeatable) recently had one of theirs die and take a mobo out (cheap worthless board) they are sending me a free upgrade and a check for $60 to cover the damage :)
 
I swapped out the motherboard to a asus b250 mining expert for the bios readout that shows if you have a bad riser.

I had yet another bad riser. (Makes about 6 of 30)

Since I’ve got working risers the system has been stable. I’ve got an issue where just one of the RX580 won’t switch to compute mode so I’m not seeing good hashrates with that particular card, but the other 11 are good.

Each of the cards in compute mode are hitting 29mhs consistently
The card that won’t flip to compute mode is doing 21mhs.

The machine is currently making about $35 a day +- $10 depending on the algorithm I’m mining and 1500 to 2400 watts depending on what algorithm as well.

It’s been stable for about 4 days now continuous with awesomeminer.

It was a headache to set up - but now that it’s up it’ll ROI pretty quickly at current profit levels.
 
Last edited:
If you had to make another 12 GPU system today, would you still go AMD or would you do NVIDIA? Not considering the systems you already have. IOW, what would you recommend to someone starting right now today to go down this road?
 
I would recommend NVidia for mining.

AMD is too much headache.

For my AMD RX580 I had to:

A) Flash bios on cards to get to 29MH/s Eth - if you don't you'll get 21-24MH/s or so. I flashed each card multiple times - which is a pain and spent many hours researching downloading and building different bios files to try. To sell the cards I’ll have to switch the bios back. This is very time consuming.

B) Change all cards to compute in the drivers to get another couple mh/s - and if they won't switch (one of mine won't) EVEN with the bios flash - then you'll get just about a 21MH/s hash rate. (stupid)

C) MSI Afterburner doesn't work with 12 AMD cards (or any AMD cards for that matter right now with the newest drivers). It's a game of cat and mouse between MSI Afterburner revisions and the newest AMD drivers. But it won't work with 12 identical cards period from either AMD or Nvidia - and so with AMD you have to program AMD settings all through Wattman. Wattman sucks. It resets if there is an error, it resets at a driver update, it resets if the host locks ups. You have to set up each and every card through Wattman individually. It takes 15 minutes and a heck of a lot of clicking to do that each time. It's unreasonable. You have to switch each card to compute mode individually and you have to set your undervolt and overclock settings individually in Wattman. It's archaic, and ridiculous - AMD knows people are using their cards for mining -- that's why they released drivers to allow for > 12 cards --- and mining drivers and ADVERTISED IT. So why in blazes do they require you set all those card settings and even compute settings individually per card??? instead of global switches. It baffles the mind, but speaks clearly to the whole difficulty and annoyance factor using AMD cards.

D) AMD cards use lots of power. My 12 card AMD rig uses as much as 10 amps at 240volt when dual mining -- even at my reduced clockspeeds of 1130 MHz core and -25% power draw in Wattman. That's 2,400 watts when dual mining. Way more than I expected going in. Once when my wattman settings were lost, and the cards reverted to stock, I noticed 14 AMPS at 240 volt. That's over 3,300 watts!!!!! Even with the undervolt settings I use, it's close to 200 watts per card while dual mining. True - if I'm doing cryptonight, which is the lowest power draw algorithm I’ve found it only is drawing ~5 amps at 240 volt so about ~1250 watts (about 100 watts per card), but each mining algorithm is different power draw, and you'll tend to mine what is most profitable, and cryptonight isn't the most profitable currently. As power hungry cards these RX580 generate a LOT of heat. They are eating up more electricity than my 1080TI 1:1 when dual mining - and are generating significantly less profit than my 1080TI. (of course they cost less than half to buy).

E) Did I mention AMD drivers suck for 12 cards? I can't even get Catalyst Control Center to open half the time without rebooting first and trying to open it right after a reboot - I don't know what that's an artifact of, but I also can't click my start bar button in Windows 10, nor can I open Edge or IE if there are 12 AMD cards. I've reformatted a half dozen times, switched motherboards 2x, and tried all the tricks I know, or could find in websearching. It's an artifact of having more than 8 or 10 AMD cards as best I can tell. My NVidia systems with 12 cards do not have this problem - I can open their driver control panel, and click the start button and open Edge and IE just fine.


True, Nvidia will cost you slightly more, but their drivers work with MSI Afterburner - ALWAYS, or NVidia inspector, and they are WAYYYYYYYYYYYYYYY less headache to use from a complication standpoint (no bios flashing, no compute mode, global settings options, less power draw), and they mine more algorithm options with higher profitability compared to AMD that only mines a few algorithms with high profitability. I think this is a big change in the sector -- if you'd asked this question a year ago - the answer would have likely been AMD, but NVidia has AMD beat for mining in today's market from my experience, and that of my few buddies that are mining.

---------------------
I'd probably recommend just doing 8 cards personally. MSI Afterburner is like a gift from the heavens in this realm. MSI Afterburner doesn't like to work with over 8 cards. You can use NVidia inspector for 12 (or more?) - but it's just not as easy - you have to click shortcut files to change the overclock/undervolt settings on each card. So in my case with my 1060 rig - I have to click 12 shortcut files after a reboot. (I should probably rig the shortcuts up to be a batch file that kicks off after reboot, but the shortcuts can't control the fan speed, and I like to manually set my fan speeds :()

So I'd recommend eight, 1070, eight 1070TI, or eight 1080Ti, 8GB of RAM, 120GB SSD, and a couple appropriately sized power supplies with an ASUS B250 Mining Expert motherboard if I was going to start fresh. The bios health check pinpointing which risers are working or not is a absolute WIN.
You can increase from 8 to 12 NVidia cards, but know that means you'll need to use NVidia inspector -- which works - but isn't pain free. 8 NVidia cards with MSI Afterburner is about as painfree as it gets.

---------------
As a super helpful trick - if you just have 8 cards, you can use task scheduler to set a nightly automatic reboot, and have the mining software automatically resume on reboot with MSI Afterburner automatically controlling the cards at each reboot. This works wonders to keep your machines reliably mining for weeks on end with no weird errors or issues. Bonus MSI Afterburner doesn't seem to lose settings....ever. not even after a NVidia driver update.


-----------------------------
At this very moment - on mining pool hub - all three rigs happen to be mining Equihash.

My AMD 12 card RX580 System (at 1130 core Mhz) is earning $35.75 per day or $1083.10 per month mining equihash at a combined speed of 3,431H/S and using 6 AMP at 240 Volt (~1440 watts).

My Nvidia 12 card 1060 system (at 65% power target) is earning $45.69 per day or $1,384.33 per month mining equihash at a combined speed of 4,400H/S and using 4-5 AMP at 240 Volt. (~1080 watts)

My Nvidia 7 card 1080TI system (at 65% power target) is earning $46.13 per day or $1,394.14 per month mining equihash at a combined speed of 4,410H/S and using 5 AMP at 240 volt. (~1200 watts)


Quoted above is for equihash, (zencash) it's a fairly low power draw, and it favors NVidia for hashrates as I've got my miners configured with lowered power draw targets.

By comparison - Ethereum hashing numbers typically favor AMD, but equihash algorithm is more profitable than eth right now - even on the AMD rig, so it was autoselected by awesomeminer profit switching function on all three rigs. If I were mining Eth you would see 29MH/s with RX580, and only 22MH/s with the 1060. The 1080TI would be in the 34-36MH/s range for Eth --- but power draw would be very different too. The dual mining of eth on my AMD rig uses 10 AMP at 240 volt which is 2,400 watts for the 12 cards. The twelve 1060s would still be drawing 5 amp at 240 volt - dual mining, and the seven 1080TI would still be drawing 5 amp at 240 volt. With minng Eth - the AMD system is usually 1/3 more profitable than the 1060 system, and would be just about as profitable as a 1070 system - but the AMD RX580 system would use more power than a 1070 system and outside of eth or maybe cryptonight algorithm a 1070 system would generally be more profitable.
 
Last edited:
Thanks for the response. I think the 1070Ti (or 1080Ti) also have the edge as far as resell.

So it seems the B250 is only available through channels I've never heard of (i.e. not Amazon, Newegg, or BH, basically). Since I'm only going 8 right now, would an ASUS Prime Z370-A be out of the question. Granted you'd have to fuss with risers, which is why I assume you recommend the Mining Expert, but apparently, as long as you order 70x the number of risers you need (hyperbole) and don't mind the time sorting out the good from bad, that could work?

And one final thing, if you're going to be building multiple rigs anyway, what is the benefit of building a 12 GPU rig over building two 6 GPU rigs other than the obvious increase in cost of non-mining components? Don't get me wrong, I understand that more = better when it comes to the worker thing. It's just, by the sound of it, if I were going to be building a farm of say 10+ rigs, it sounds like I'd save myself a lot of headaches by building 15 8GPU rigs rather than 10 12GPU rigs, assuming I don't mind the extra expense of the non-mining parts vs the headache.

I appreciate you sharing your build log and knowledge acquired via sweat and tears, literally.
 
I updated my previous post a bit more since you read it probably -

but to your next question:

yes...

with the current driver state and limitations on 12+ GPU rigs - it's definitely more stable and pain free to go with 6 or 8 card systems.

it's just a matter of wanting to buy less CPUs, RAM, PowerSupplies and hard-drives that is enticing about these 12, 13, or 19 card motherboards.

If that doesn't bother you, then six or eight card rigs are the way to go - and probably worth the added expense to reduce the complication.


I will never recommend any other motherboard than the ASUS B250 from this point forward until any other competitor comes up with a BIOS screen to tell you if you have a bad riser. period. .... never! ever! ever!


I literally spent probably 100+ hours troubleshooting that Biostar TB250 BTC PRO 12 GPU motherboard with that half dozen bad risers in the mix. The trick is - there is NO reasonable way to figure out which riser is failing. The event log doesn't tell you, the numbers in any logs don't match the PCI-E numbers silkscreened on the motherboard, and the GPU numbers in windows can and do change when you boot up. So your left blindly swapping things in aggravation, or hoping that MSI afterburner or wattman will be able to control all the fans for the cards after your system locks up or is in a funk so you can hopefully figure out which card isn't responding to your reqeust to ramp up the FAN speed to determine if somethings' wrong with that card. It might be that the card just was unstable at the undervolt overclock, it might be a bad riser, it might be a random algorithm bug. It could be a LOT of things -- and trying any one particular path has like a 5% chance of success IMO. Worthless!!!
I'd pay $500 for a b250 motherboard rather than spend 100 hours ferreting through intermittently failing risers again if it came to that type of tradeoff.


When I booted up the asus b250 mining extreme and was met with this screen immediately and CLEARLY calling out a intermitted failing riser that I was having many hours of trouble identifying and dreaded chasing down any further. It was like a holy angel chorus engaged and I rejoiced. If you have a microcenter nearby they sell them for $150. Otherwise, pony up the money and be thankful you'll never have to deal with chasing down an intermittently failing riser - it's a yugggee pain!

35D9D038-700D-4755-861A-5BAD2AEDE603.jpeg
 
Last edited:
From the main MPH page, mining etash (ETH) on MPH is more profitable than equihash for AMD cards. it's the opposite for nvidia cards. There is a tab in the table to swtich between the two.

What are you dual mining on MPH? Only thing for AMD cards that is dual mineable is ETH and SIA but SIA is either broken or not worth mining.

edit: nvm seems my MPH page wasn't the latest. refreshed and it does show zenclassic as being more profitable than Eth. Guess zen is getting pumped from the laundering of the stolen NH BTC.
 
Last edited:
I am not sure how you are having 200 watt dual mining.

I had a single RX570 powered by an HP server PSU, connected to a killowatt. Running at .9V, Eth mining was 110watts, Eth+Lbry was ~145 watts.

I never touched the power limit on the cards
 
Hell, it sounds like the B250 Mining Expert is worth the money even if you just use it as a riser tester :)

Sadly, there are no Microcenters where I live (Tennessee) but I just checked about 10 of them and they all say SOLD OUT for that board. Hell, I wouldn't mind buying from a buyer I don't know but I've not found a source that doesn't sound like it's simply a drop shipper who may or may not actually be able to cause it to ship to me. More than likely I place an order and get a call from them tomorrow trying to upsell me or side sell me or...

Hard to believe that some other board hasn't thought of adding this capability. I haven't even been able to find a stand alone tester...

I updated my previous post a bit more since you read it probably -

but to your next question:

yes...

with the current driver state and limitations on 12+ GPU rigs - it's definitely more stable and pain free to go with 6 or 8 card systems.

it's just a matter of wanting to buy less CPUs, RAM, PowerSupplies and hard-drives that is enticing about these 12, 13, or 19 card motherboards.

If that doesn't bother you, then six or eight card rigs are the way to go - and probably worth the added expense to reduce the complication.


I will never recommend any other motherboard than the ASUS B250 from this point forward until any other competitor comes up with a BIOS screen to tell you if you have a bad riser. period. .... never! ever! ever!


I literally spent probably 100+ hours troubleshooting that Biostar TB250 BTC PRO 12 GPU motherboard with that half dozen bad risers in the mix. The trick is - there is NO reasonable way to figure out which riser is failing. The event log doesn't tell you, the numbers in any logs don't match the PCI-E numbers silkscreened on the motherboard, and the GPU numbers in windows can and do change when you boot up. So your left blindly swapping things in aggravation, or hoping that MSI afterburner or wattman will be able to control all the fans for the cards after your system locks up or is in a funk so you can hopefully figure out which card isn't responding to your reqeust to ramp up the FAN speed to determine if somethings' wrong with that card. It might be that the card just was unstable at the undervolt overclock, it might be a bad riser, it might be a random algorithm bug. It could be a LOT of things -- and trying any one particular path has like a 5% chance of success IMO. Worthless!!!
I'd pay $500 for a b250 motherboard rather than spend 100 hours ferreting through intermittently failing risers again if it came to that type of tradeoff.


When I booted up the asus b250 mining extreme and was met with this screen immediately and CLEARLY calling out a intermitted failing riser that I was having many hours of trouble identifying and dreaded chasing down any further. It was like a holy angel chorus engaged and I rejoiced. If you have a microcenter nearby they sell them for $150. Otherwise, pony up the money and be thankful you'll never have to deal with chasing down an intermittently failing riser - it's a yugggee pain!

View attachment 46556
 
there is no DCR LBRY PASCAL on MPH


I've used these rigs on more than MPH. MPH is just what I'm using the last couple days.

I've tried
AutoSwitch Clients:
Nicehash
Minergate (is minergate even legit? I have $70 ETH in there I can't seem to cash out - service temporarily unavailable when I try)
AwesomeMiner
Multipoolminer

Pools:
Nicehash
MPH
Hash Refinery
Zpool
 
I am not sure how you are having 200 watt dual mining.

I had a single RX570 powered by an HP server PSU, connected to a killowatt. Running at .9V, Eth mining was 110watts, Eth+Lbry was ~145 watts.

I never touched the power limit on the cards


Here are the settings I use on my RX580

CoreClock 1130Mhz
Core Voltage 875mv
RAM 2100 - 2200mhz (I put the cards at 2200 and when one crashes I lower it by 25Mhz each time it crashes) - 2100 seems to work on all of them, but anything more is not guaranteed.
Fan Speed - 1800-2700RPM
Target Temp - 70*
Power Target - 25% (which I'm not sure even does anything since I specify the core voltage??? - but if my wattman crashes - it seems to retain the -25% power target even if it retains nothing else - so that at least keeps my power usage in check somewhat)


Sometimes they sip power, sometimes they suck power -- it all depends on the algorithm. The highest I ever see them at is 10 amps at 240 volt with these power settings working. When wattman resets - I've seen the server PDU say these 12 cards are pulling as high as 14 amps at 240 volt for 12 cards. That's more than my three PSU's are even rated for. I have a 1000 watt gold rated PSU on four cards and the motherboard, and two 850 watt gold rated PSUs - each with four cards.

The Asus Mining expert motherboard is nice in that all three PSU's plug directly into the motherboard and all turn off and turn on as normal -- as compared to the little daisy chain adapters - where usually the secondary PSU's dont' actually turn off if you reboot or power down the motherboard. With this Asus motherboard they do behave as they should in regards to power cycling.

I know the motherboard and g4560 CPU only pulls about 55 watts in my testing -- so the math and my LCD readout says these cards can really drain the power if they are called upon.

The bios file I'm currently using is from here:
http://1stminingrig.com/best-bios-r...80-8gb-limited-edition-hynix-memory-31-5-mhs/
 
Last edited:
yea, I know you have. Just seemed you were dual mining on MPH from the post so was wondering what coins. I can point to another pool for the second coin but then it becomes an issue of taking forever to get cashed out since I"m only a 1 gpu miner.

minergate yea, issue with them is getting paid out.
zpool seems to have issues as well
never heard of Hash refinery
not going back to NH
so only thing left is direct mining pools and MPH.

edit: looks like etash (ETH) is back to being more profitable then zenclassic on MPH for AMD. Guess the latest 1K BTC stolen from NH is done being laundered. Hopefully nobody helped verify transfer of their stolen BTC by mining zen.
 
Last edited:
Seems like sourcing the parts for these rigs is the main problem... I can't find 1070Ti's for less than about $480-500 per. Hell, I can find GTX 1080's (non-Ti o/c) for $500 each pretty easy...

The mining craze seems to have hit long and hard
 
Hell, it sounds like the B250 Mining Expert is worth the money even if you just use it as a riser tester :)

Sadly, there are no Microcenters where I live (Tennessee) but I just checked about 10 of them and they all say SOLD OUT for that board. Hell, I wouldn't mind buying from a buyer I don't know but I've not found a source that doesn't sound like it's simply a drop shipper who may or may not actually be able to cause it to ship to me. More than likely I place an order and get a call from them tomorrow trying to upsell me or side sell me or...

Hard to believe that some other board hasn't thought of adding this capability. I haven't even been able to find a stand alone tester...

Hello,

may I can help you with the B250 Mining Expert. I have acces to this boards and can provide you with some. I'm living in austria, so shipping and customs have to be cleard from your side.

I myself running 2 of this boards with each 12 RX580 sapphire nitro+ 4GB without problem.
Idon't use Afterburner Wattman and all this stuff. After configuration (Bios flash Core 1200 950 mV Mem 1950 900)
of the drivers, I prevented via registry Wattman and AMI Settings. (you have to delete also the executable files in the AMD directory). Then reboot.

I use OverdriveNTool.exe not Wattman MSI Afterburner or Trixx.

No AMD Task is running.

Since then my rigg ist constatly running with total of 175 Mh/s claymore-miner on ETH only for 3 weeks without any crash.

claymore-miner (-tt 80 -fanmin 70 70 70 70 70 70 -r 0).
No Wattman no Catalyst stuff only the AMD driver . Temp for each GPU between 50° - 54° C

It took me 1 week to figure this out.
 
Last edited:
I'm very thankful for all the information in this thread. I'm about to pull the trigger on a 8x 1080Ti rig and didn't really know where to start as far as motherboard, PSU, risers. Luckily I have a dedicated 240v 60A circuit in my house that was for a hot tub that no longer exists. I'll be making good use of that as I build this rig and expand. I think that should be good for at least 24x 1080ti cards eventually.

Again, thanks for the info!
 
Seems like sourcing the parts for these rigs is the main problem... I can't find 1070Ti's for less than about $480-500 per. Hell, I can find GTX 1080's (non-Ti o/c) for $500 each pretty easy...

The mining craze seems to have hit long and hard
Take a look at superbiiz. Reasonable prices and no limit on quantities.
 
Hello,

may I can help you with the B250 Mining Expert. I have acces to this boards and can provide you with some. I'm living in austria, so shipping and customs have to be cleard from your side.

I myself running 2 of this boards with each 12 RX580 sapphire nitro+ 4GB without problem.
Idon't use Afterburner Wattman and all this stuff. After configuration (Bios flash Core 1200 950 mV Mem 1950 900)
of the drivers, I prevented via registry Wattman and AMI Settings. (you have to delete also the executable files in the AMD directory). Then reboot.

I use OverdriveNTool.exe not Wattman MSI Afterburner or Trixx.

No AMD Task is running.

Since then my rigg ist constatly running with total of 175 Mh/s claymore-miner on ETH only for 3 weeks without any crash.

claymore-miner (-tt 80 -fanmin 70 70 70 70 70 70 -r 0).
No Wattman no Catalyst stuff only the AMD driver . Temp for each GPU between 50° - 54° C

It took me 1 week to figure this out.


Could you take s screenshot of your bios power settings through PBE? Or send me a copy of one of your custom bios files so I could see how you are doing it in PBE? I’ve tried doing similar custom power settings myself in a custom bios i reverse engineered but was having issues. I’d like to compare what I tried to what’s working for you.
 
Could you take s screenshot of your bios power settings through PBE? Or send me a copy of one of your custom bios files so I could see how you are doing it in PBE? I’ve tried doing similar custom power settings myself in a custom bios i reverse engineered but was having issues. I’d like to compare what I tried to what’s working for you.

After the Biosflash, I finetuned every card with OverdriveNtool. Some like mem clock 2000 some only 1950 .
 

Attachments

  • screenshot2.png
    screenshot2.png
    35.2 KB · Views: 35
  • screenshot.png
    screenshot.png
    35.8 KB · Views: 32
  • Sapphire RX580 NITRO(plus) 4GB Elpida.exe
    360.8 KB · Views: 12
Last edited:
How did you hook up your power supplies as they relate to the mobo/risers/gpus? I'm attempting to build something similar to your rig, and power supplies are on the way (3x EVGA 750s, 1x EVGA 1000 - all G2). Planning on 8x 580s. There seems to be a heck of a lot of misinformation about multi psu set-ups, and your rig seems like a really good analogy - so I thought I'd ask. I have Add2psus on the way as well.
 
Back
Top