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

With the biostar 250-btc pro board I used a simple splitter cable.

https://minerparts.com/product/triple-psu-cable-3-power-supplies-24-pin-atx-cable/

It’s not really doing anything except turning the all three PSUs on at once.

You don’t need anything extra at all if you use the evga bundled PSU power on testers. You can just start your second and third PSUs with those —- and turn them off and on with the toggle switch on the back of the PSU.


With the ASUS b250 mining expert it has three PSU connectors on the board so I’m just using those.
 
Last edited:
Archaea said:
With the biostar 250-btc pro board I used a simple splitter cable.

https://minerparts.com/product/triple-psu-cable-3-power-supplies-24-pin-atx-cable/



With the ASUS b250 mining expert it has three PSU connectors on the board.
Click to expand...
So, let me ask:

Let's call it PSU1 PSU2 PSU3.

PSU1, 2, and 3 are all linked with your adapter cable, meaning they all turn on together. So, lets say PSU1 is with the primary PSU cable - the one that had all its pins go into the mobo - did PSU1 power ALL the risers? And then PSU 2 and PSU 3 only powered GPUs? Or did you power the risers and GPUs from PSU 2 and PSU 3?

Edit: If you did power risers and GPUs with PSU 2 and 3, did you power both risers and GPU on only the same PSU?
 
I updated my last post after you made this question. So you might reread that.

But in answer to your additional question no. Absolutely not. It’s important you don’t do that. Your riser card and associated GPU all need to be powered by the same power supply.

Don’t mix them. You’ll may otherwise cause voltage ripple which can damage components. So each individual card and it’s associated riser needs to be powered by the same PSU because the card will pull power from both the powered riser and the pci-e connections.
 
Archaea said:
I updated my last post after you made this question. So you might reread that.

But in answer to your additional question no. Absolutely not. It’s important you don’t do that. Your riser card and associated GPU all need to be powered by the same power supply.

Don’t mix them. You’ll may otherwise cause voltage ripple which can damage components. So each individual card and it’s associated riser needs to be powered by the same PSU because the card will pull power from both the powered riser and the pci-e connections.
Click to expand...

Thank you for this. To make sure I understand completely:

PSU 1 = mobo, SSD, GPU 1 (and its riser), GPU 2 (and its riser), GPU 3 (and its riser)
PSU 2 = GPU 4 (and its riser, GPU 5 (and its riser), GPU 6 (and its riser)
PSU 3 = GPU 7 (and its riser), GPU 8 (and its riser), etc.

Am I understanding this perfectly correctly? Or at least, am I understanding that this is the correct way to distribute power?
 
Last edited:
Yes that’s correct.

You’ll not need your add2PSU parts. If you do decide to use them note that sometimes the delay on power up the the the second and third PSU can possibly be too slow for the motherboard to register the cards. It’s not always a problem but could be. If the cards don’t get recognized then you’ll know why.
 
  • Like
Reactions: Joust
like this
Archaea said:
Yes that’s correct.

You’ll not need your add2PSU parts. If you do decide to use them note that sometimes the delay on power up the the the second and third PSU can possibly be too slow for the motherboard to register the cards. It’s not always a problem but could be. If the cards don’t get recognized then you’ll know why.
Click to expand...
Thank you very much for your guidance.
 
Mainboard is ASUS B250 mining expert. You can plug 3 PSU to this board. I use one 1200W and one 1000W (https://www.bequiet.com/de/powersupply) . I want to build it up to 16 GPU'S.
The Risers are connected to the PSU not over the mainboard. Not more than 2 raiser for one outlet of the PSU !
And like Archaea wrote always Riser and GPU on the same PSU
 
Last edited:
AMD drivers for a 12 card rig in Windows just suck - no 2 ways about it.

I have tried 3 different motherboards. Two Biostar 12 card boards, and one Asus 19 card motherboard.
I have tried multiple different bios files for the AMD cards.

I had it working nicely for the last couple weeks, since my last update - then today some unwanted, unrequested, automatic windows 10 update hit and installed 17.40 beta AMD drivers on my system. (I had windows 10 updates disabled). Then whole thing went to crap again.

I'm back and fighting this thing tooth and nail. Start bar won't work once again. AMD driver control center takes about 3 minutes between each screen and tab -- NOT KIDDING (want to switch to the next graphics card - sit there 3 minutes). Compute mode switched off, all settings reset. I'm thissssssssssss close to selling these 12 RX580 AMD cards. They simply aren't worth the hassle!!!!!!!!!!! (at least in my attempt at running 12 at once).

junk junk junk AMD drivers...absolutely junk compared to NVidia drivers for mining.

I reinstalled AMD 17.2.11 drivers -- at least now, in this recent update they have a save and load profile in Wattman.

Good step AMD ---
Yet...
Why is the AMD dev team so stupid that they won't make a 'sync' across all cards button? THEY KNOW ---ABSOLUTELY KNOW people are mining with these things - they have the compute button. They now have the save and load button - why not just make a single page for all identical cards in Wattman. Instead, even with this new profile setting - I still have to save one profile and load it 12 times on 12 different pages, that take 3 minutes to switch between each page. (and that's not counting the 12 pages for the compute settings that are in different tabs). So it's going to take me an hour to fix this when it should take 2 minutes.

Stupid...Stupid...Stupid AMD.

USE YOUR BRAIN devs!
 
Last edited:
Archaea I've been debating selling my 580s too.

Now woudn't be a bad time to unload them, either. The 8GB models are pretty damn close to $400 on fleabay.

These NVidia rigs have spoiled me rotten.
 
Archaea said:
AMD drivers for a 12 card rig in Windows just suck - no 2 ways about it.

:) once the system is running for 48 hours the first thing to do is make a clone of your harddisk. A SSD 60GB cost nothing compared to the work in the setup. I have 3 clones. If something unexpcted happend like windows update. it takes 5 minutes to change the SSD and the rigg is on again.

I found this very useful to stop updates ´, just change the extention to .bat und run it.
Click to expand...
 

Attachments

  • Windows 10 Registry tweaks for mining.txt
    19.6 KB · Views: 27
windows 10 treats windows updates and driver updates as separate things.

use O&O ShutUp10 to turn off/on all updates and other spyware and crap as you want.

you could also just rolled back by uninstalling updates via update history.
 
atp1916 said:
Archaea I've been debating selling my 580s too.

Now woudn't be a bad time to unload them, either. The 8GB models are pretty damn close to $400 on fleabay.

These NVidia rigs have spoiled me rotten.
Click to expand...


Your post and a PM got me thinking. I put the 12 cards up on ebay at a hail mary price of $4800. Somebody bought them within a few hours of my listing.
https://www.ebay.com/itm/222775247561?ssPageName=STRK:MESOX:IT&_trksid=p3984.m1559.l2649

I probably could have listed them for more. They are hot right now.

Doesn't hurt that they make $4-$5 per day per card on nicehash right now.


For me....I think I'll just stick with Nvidia rigs from here on out. They are just easier to administer.

upload_2017-12-30_0-57-33.png
 
Last edited:
As an eBay Associate, HardForum may earn from qualifying purchases.
If the average EBay seller was that honest, straight forward, clear and concise - Ebay would have taken over the world!
 
100 hours+ of wtfbbq pain. Ugh. Stomach turns just thinking about that.

OTOH, I've spent all of perhaps 10 hours building and configuring my 3 8 card NVidia boxes and 0 hours toward actual troubleshooting.

They just run, straight and true.

I love it.
 
atp1916 said:
100 hours+ of wtfbbq pain. Ugh. Stomach turns just thinking about that.

OTOH, I've spent all of perhaps 10 hours building and configuring my 3 8 card NVidia boxes and 0 hours toward actual troubleshooting.

They just run, straight and true.

I love it.
Click to expand...
Loaded block chain drivers, clean install for Vega FE. After it was done the Radeon Pro drivers came up on reboot :mad:. Did it again but with a game driver installed - The radeon Pro driver came up again :banghead:. Reloaded again and did not reboot this time, block chain working. WTF - To game, you need to load game drivers, to do crypto with great hashing rates the blockchain ones - Professional stuff Radeon Pro drivers. With the Radeon Pro drivers you can load two more drivers with the pro dirvers and switch between then. Except it will not tell you the real driver you are going to unless they are the default ones. I am not sure if I reboot what driver will load (Pro, 17.12.1, 17.9.1 or blockchain ones). I'll leave it be for now.
 
odditory said:
Damn, you sold for way too cheap, my man. Nitro 8GB's are the most in-demand 580's. You gave someone a gift.

You should've listed them on this forum first to avoid the ebay tax. And if ebay, you should have subscribed to a basic ebay store subscription before listing ($24.95), which would've dropped ebay's cut on that $4800 from 10% to 4%.

Seriously, if you haven't shipped them yet, cancel that sale. :)
Click to expand...
That's not really helpful at this point, I think. He got a return on his investment and moved on. Painless.
 
Joust said:
That's not really helpful at this point, I think. He got a return on his investment and moved on. Painless.
Click to expand...

It's throwing money away. If he hasn't shipped yet, he has nothing to lose by cancelling the sale and re-doing it the way I recommended.
 
Odditory,

I thank you for the pm and the message here. Unfortunately for me, I didn't know about the storefront discount/opportunity. At this point my buyers already paid me, (it was an instant payment) and I don't want to make him nervous that something shady is going on --- so I'm just going to roll with what already transpired and keep the storefront idea in the forefront for future sales.

I did call eBay and ask if as a customer courtesy, if they would let me buy a subscription to the storefront today and then potentially limit this transaction to 4% - they said no dice, since the transaction had already occurred.

Nobody likes the 13% ebay/paypal fee, but I knew those fees were there going in - I sell quite a bit on ebay (why haven't I ever used storefront? --- gah - didn't know about the monthly option being so easy!!!)

Since I acknowledged the expected 13% fees, expected it, and doing something different now could cause confusion for the buyer - I'll just let it go...
I think the price I got was more than fair, so I'm going to lose sleep over it.

Thanks for looking out for me though. I appreciate the heads up, and will very likely use the monthly storefront subscription in the future when I need to sell another round of expensive equipment!
 
well the buyer can leave negative feedback and nothing you can do about it. however it will go away in a year or so. I'd do as suggested.

edit: actually I think buyer could only leave neutral feedback and make a stink in the comments. which won't effect 100% buyer rating. but if you give buyer heads up and why and telling them you are listing it again buyer may understand and not do anything but buy relistings.
 
Canceling the sale means rolling the dice.
What if that pisses the buyer off and they pass when the auction is put up again?
How many other buyers will be interested, and how many have the means to pay?
If it's me, I get the cards shipped and move on. I sleep well tonight knowing that money is in my account now.
 
pgaster said:
Canceling the sale means rolling the dice.
What if that pisses the buyer off and they pass when the auction is put up again?
How many other buyers will be interested, and how many have the means to pay?
If it's me, I get the cards shipped and move on. I sleep well tonight knowing that money is in my account now.
Click to expand...
I've made a number of deals. Sometimes it wasn't optimised for me. In my unsolicited opinion, once your deal has been made and money has changed hands, it's done. I'm not going to get preachy about integrity or anything, and I can see that opinions differ here, but it's how I conduct business personally.

Once again in my opinion, the guy got a return and a quick sale. He can feel good about that and move on.
 
Joust, I agree.

The money changing hands part is really pretty immutable IMO. My auction terms said I would ship them out by Tuesday Jan 2.

But before any of this discussion about canceling and relisting came up I had messaged the buyer with the following when I woke up this morning.

Thanks for your quick payment.

I'm currently mining with these cards as noted in my ad. Because today is Saturday --- and Sunday and Monday the cards won't ship (Sunday FedEx doesn't typically ship, and New Years day is Monday) they will likely sit in a shipping warehouse this weekend. Noting that -- but given that I want to be ethical, I'll give you your choice. Currently the 12 cards are making $42 per day on nicehash. I can ship them out today at your request via FedEx ground, they'll probably sit in a warehouse tonight through Tuesday morning. Or I can ship them to you on Tuesday, and we'll split my mining proceeds with you over this weekend. The cards are using about 1440 watts - at my electric price of .06 kw/h which is ~$2 per day. So lets just say ~$40 per day profit. 40 x 3 days (Saturday, Sunday, Monday) = $120 bucks. $120 bucks divide by 2 people = $60 each.


So your call.

Option 1:
Do you want me to ship them today? They are yours after all - you paid for them already...

Option 2:
Do you prefer I ship them Tuesday, and refund you $60 of your paypal payment.

let me know.

P.S. I'll need to know reasonably soon so I can get them boxed up and out the door if you want them shipped out today!


-archaeantrader


The buyer wrote back and said he chose option 2, and would like to split mining proceeds and have me ship Tuesday. So I refunded him $60 today and I'll ship Tuesday.


A man's Integrity is all he has in this world. I won't trade my integrity for profit on something I don't feel right about - even if it's subjective or gray area to some folk.

I still made profit. I knew the terms I signed up for. So did the buyer. It's unfair in a purchase arrangement to change terms without buyers agreement or participation once money has changed hands and if terms haven't changed. My opinion anyway. I sell a lot of stuff on ebay. The store function will be useful at a later date. I do appreciate you brining it up. But I won't be taking advantage of it here. It's water under the bridge.
 
I have now tried three different 12 card mining rigs on two different motherboards models, (4 different individual boards)

12 identical EVGA 1060 (6GB) in both a Biostar B250-BTC Pro and a Asus B250 Mining Expert
12 identical Sapphire RX580 (8GB) in both a B250-BTC Pro and a B250 Mining Expert
12 mixed 1080TI (8 PNY XLR8, 4 MSI Founders Editions) in a B250 Mining Expert
All via Windows 10, 1709 patch update.

My conclusion is that 12 card mining rigs are not worth the effort with Windows 10. Last week I consolidated 12 of my very stable 1080TI from two different rigs into a single 12 card rig to try it. I didn't meet success.

Using NiceHash my overclocks (while undervolting) suddenly weren't as stable on the same settings for the 1080TI cards I'd used in 8 card systems setups for a long time. Various algorithms would bug out the NiceHash client and it would report 0 profit, while still submitting valid shares --- leading me to confusion if I was actually getting paid or not???(neoscript in particular was terrible at this), and the Windows user interface was MUCH slower and nearly unusable with 12 cards plugged in. Moving from 8 to 12 cards, the Windows start button stopped working again so I had to result to command line navigation or windows + run bar use. I reloaded Windows, same problem. Afterburner can't see the last four cards, so it can only apply universal settings to all, you can't customize each of the 12 cards - and it seemed like temps went up and auto overclock bonuses went down for the 8 cards I could still see. And with 12 cards, making changes in Afterburner (for NVidia) is untolerably slow (and for AMD-- Afterburner rarely even worked). I dropped back down to 8 cards, and all the problems went away, system is more responsive, Windows start bar works again, NiceHash Gui always shows incoming profit, and finally temps and overclocks are back to normal etc.

My best success for 12 card rigs was actually with Awesome Miner and Mining Pool hub. Eliminate all the lesser known algorithms (like 3/4 of the algorithms), and set it to work. Awesome miner seems to be a bit more reliable with 12 cards than NiceHash (both with NVidia and AMD), but still not as reliable as 8 cards. With 8 cards I prefer to use NiceHash.


My simplest setup of these three rigs above was my most reliable. My 1060 setup was just a single 1200 watt PSU for everything, and each 1060 was set to 66% power draw, and it was really quite reliable with Awesome Miner. With NiceHash on the 1060 rig I noticed the algorithms were constantly changing and I think changing too often to be as profitable as I expected. I tried changing the interval in the advanced config, but it didn't take. With Awesome Miner I could successfully change the interval and I made it 30 minutes, or 10% as the change deltas. That seemed to work well with the 12 1060. The only issue I had with the 1060 rig was driver updates sometimes knocking out O/C and undervolt settings, and one instance in there where the MSI afterburner tool stopped working with 12 NVidia cards for a short while. Without the afterburner undervolt - the 1200 watt PSU wasn't up to the task of driving 12 cards, so I had to temporarily add another PSU, (until MSI updated the afterburner client once again to support 12 cards) but no other issues day to day.

------------

Ultimately - Windows 10 and tools like NiceHash and AwesomeMiner, and MSI Afterburner or even AMD's WattMan are NOT yet ready to run 12 cards - neither for AMD, nor Nvidia.

So this is my current conclusion for this 12 card blog (at least using auto switching profit tools like NiceHash and Awesome Miner).

Yes it works, but it doesn't work hands off, and it's more administrative effort at this time than it's worth.
Maybe it's worth checking back mid this year.
With this opinion cemented - I'll be limiting my rigs to 8 identical cards for the immediate future.
 
Last edited:
Archaea said:
I have now tried three different 12 card mining rigs on two different motherboards models, (4 different individual boards)

12 identical EVGA 1060 (6GB) in both a Biostar B250-BTC Pro and a B250 Mining Expert
12 identical Sapphire RX580 (8GB) in both a Biostar B250-BTC Pro and a B250 Mining Expert
12 mixed 1080TI (8 PNY XLR8, 4 MSI Founders Editions) in a B250 Mining Expert
All via Windows 10, 1709 patch update.

My conclusion is that 12 card mining rigs are not worth the effort. Last week I consolidated 12 of my very stable 1080TI from two different rigs into a single 12 card rig to try it. I didn't meet success.

Using NiceHash my overclocks (while undervolting) suddenly weren't as stable on the same settings for the 1080TI cards I'd used in 8 card systems setups for a long time. Various algorithms would bug out the NiceHash client and it would report 0 profit, while still submitting valid shares --- leading me to confusion if I was actually getting paid or not???(neoscript in particular was terrible at this), and the Windows user interface was MUCH slower and nearly unusable with 12 cards plugged in. Moving from 8 to 12 cards, the Windows start button stopped working again so I had to result to command line navigation or windows + run bar use. I reloaded Windows, same problem. Afterburner can't see the last four cards, so it can only apply universal settings to all, you can't customize each of the 12 cards - and it seemed like temps went up and auto overclock bonuses when down for the 8 cards I could still see. And with 12 cards, making changes in Afterburner (for NVidia) is untolerably slow (and for AMD-- Afterburner rarely even worked). I dropped back down to 8 cards, and all the problems went away, system is more responsive, Windows start bar works again, NiceHash Gui always shows incoming profit, and finally temps and overclocks are back to normal etc.

My best success for 12 card rigs was actually with Awesome Miner and Mining Pool hub. Eliminate all the lesser known algorithms (like 3/4 of the algorithms), and set it to work. Awesome miner seems to be a bit more reliable with 12 cards than NiceHash (both with NVidia and AMD), but still not as reliable as 8 cards. With 8 cards I prefer to use NiceHash.


My simplest setup of these three rigs above was my most reliable. My 1060 setup was just a single 1200 watt PSU for everything, and each 1060 was set to 66% power draw, and it was really quite reliable with Awesome Miner. With NiceHash on the 1060 rig I noticed the algorithms were constantly changing and I think changing too often to be as profitable as I expected. I tried changing the interval in the advanced config, but it didn't take. With Awesome Miner I could successfully change the interval and I made it 30 minutes, or 10% as the change deltas. That seemed to work well with the 12 1060. The only issue I had with the 1060 rig was driver updates sometimes knocking out O/C and undervolt settings, and one instance in there where the MSI afterburner tool stopped working with 12 NVidia cards for a short while. Without the afterburner undervolt - the 1200 watt PSU wasn't up to the task of driving 12 cards, so I had to temporarily add another PSU, (until MSI updated the afterburner client once again to support 12 cards) but no other issues day to day.

------------

Ultimately - Windows 10 and tools like NiceHash and AwesomeMiner, and MSI Afterburner or even AMD's WattMan are NOT yet ready to run 12 cards - neither for AMD, nor Nvidia.

So this is my current conclusion for this 12 card blog.

Yes it works, but it sure doesn't work hands off, and it's more administrative effort at this time than it's worth.
Maybe it's worth checking back mid this year.
With this opinion cemented - I'll be limiting my rigs to 8 identical cards for the immediate future.
Click to expand...


Good to know, I'm going to stay with my 6 card boards :)
 
Archaea said:
I have now tried three different 12 card mining rigs on two different motherboards models, (4 different individual boards)

12 identical EVGA 1060 (6GB) in both a Biostar B250-BTC Pro and a Asus B250 Mining Expert
12 identical Sapphire RX580 (8GB) in both a B250-BTC Pro and a B250 Mining Expert
12 mixed 1080TI (8 PNY XLR8, 4 MSI Founders Editions) in a B250 Mining Expert
All via Windows 10, 1709 patch update.

My conclusion is that 12 card mining rigs are not worth the effort. Last week I consolidated 12 of my very stable 1080TI from two different rigs into a single 12 card rig to try it. I didn't meet success.

Using NiceHash my overclocks (while undervolting) suddenly weren't as stable on the same settings for the 1080TI cards I'd used in 8 card systems setups for a long time. Various algorithms would bug out the NiceHash client and it would report 0 profit, while still submitting valid shares --- leading me to confusion if I was actually getting paid or not???(neoscript in particular was terrible at this), and the Windows user interface was MUCH slower and nearly unusable with 12 cards plugged in. Moving from 8 to 12 cards, the Windows start button stopped working again so I had to result to command line navigation or windows + run bar use. I reloaded Windows, same problem. Afterburner can't see the last four cards, so it can only apply universal settings to all, you can't customize each of the 12 cards - and it seemed like temps went up and auto overclock bonuses went down for the 8 cards I could still see. And with 12 cards, making changes in Afterburner (for NVidia) is untolerably slow (and for AMD-- Afterburner rarely even worked). I dropped back down to 8 cards, and all the problems went away, system is more responsive, Windows start bar works again, NiceHash Gui always shows incoming profit, and finally temps and overclocks are back to normal etc.

My best success for 12 card rigs was actually with Awesome Miner and Mining Pool hub. Eliminate all the lesser known algorithms (like 3/4 of the algorithms), and set it to work. Awesome miner seems to be a bit more reliable with 12 cards than NiceHash (both with NVidia and AMD), but still not as reliable as 8 cards. With 8 cards I prefer to use NiceHash.


My simplest setup of these three rigs above was my most reliable. My 1060 setup was just a single 1200 watt PSU for everything, and each 1060 was set to 66% power draw, and it was really quite reliable with Awesome Miner. With NiceHash on the 1060 rig I noticed the algorithms were constantly changing and I think changing too often to be as profitable as I expected. I tried changing the interval in the advanced config, but it didn't take. With Awesome Miner I could successfully change the interval and I made it 30 minutes, or 10% as the change deltas. That seemed to work well with the 12 1060. The only issue I had with the 1060 rig was driver updates sometimes knocking out O/C and undervolt settings, and one instance in there where the MSI afterburner tool stopped working with 12 NVidia cards for a short while. Without the afterburner undervolt - the 1200 watt PSU wasn't up to the task of driving 12 cards, so I had to temporarily add another PSU, (until MSI updated the afterburner client once again to support 12 cards) but no other issues day to day.

------------

Ultimately - Windows 10 and tools like NiceHash and AwesomeMiner, and MSI Afterburner or even AMD's WattMan are NOT yet ready to run 12 cards - neither for AMD, nor Nvidia.

So this is my current conclusion for this 12 card blog (at least using auto switching profit tools like NiceHash and Awesome Miner).

Yes it works, but it doesn't work hands off, and it's more administrative effort at this time than it's worth.
Maybe it's worth checking back mid this year.
With this opinion cemented - I'll be limiting my rigs to 8 identical cards for the immediate future.
Click to expand...

Which CPU, and which version of Nicehash?
 
Last edited:
odditory

Pentium G4560 in all instances - because it's the cheapest dual core CPU with hyperthreading. 3.5GHz. (Microcenter used to sell them for $57 - Wal-Mart has them through 3rd party ANTONLINE for $61.) In all instances while mining the CPU would be at about 40-50% with 12 cards. The CPU always spikes to 99% for about 30 seconds when you open up task manager and try to look around, but under normal operation it's less than 50% With 8 cards the G4560 is more like 30-40% use range.


Other system stats:
8GB (2x4GB) of 2400Mhz DDR4 RAM on the 12 card 1060 system using 1200 watt PSU (66% power target on cards) (4 AMP load at 240 volt)
8GB (2x4GB) of 2400Mhz DDR4 RAM on the 12 card RX580 system using one 1000 watt PSU as primary, and two 850 watt PSUs as second and third PSU (9 AMP load at 240 volt)
16GB (2x8GB) of 2400Mhz DDR4 RAM on the 1080TI system using one 1200 watt PSU as primary, and two 1000 watt PSUs as second and third PSU. (70% power target on cards) (10 AMP load at 240 volt)

Various SSD's in all instances with the swap file (page file) set unique to each System at a 1:1 ratio with the total GPU RAM and System Memory combined
IE - 12 1080TI = 11GB each x 12 cards = 132GB GPU RAM + 16GB System RAM = 148GB Total RAM x 1024 (to get MB size) = 151,552MB Windows swap file manually set.


Always the newest instance of NiceHash. I typically update same day availability.

Always the newest version of AMD and Nvidia drivers - update regularly there too.
 
Gotcha. I asked Nicehash version in case you were still running legacy (at the moment there's no point w/ Nvidia GPU's) - there are some issues with certain versions/forks of ccminer used by NHML + 12 GPU's + low end CPU, but that's another topic. I would lay money that with a more powerful CPU, all your problems would've gone away.

My experience has been this: With 12 x 1080 Ti, Windows 10 build 1709 (I did de-bloat it with MSMG Toolkit), DDR4-2133 4GB memory, 250GB SATA SSD w/ 136GB Swapfile, and with an i7-7700k that I installed temporarily, both the Asus B250 Mining Expert and more recently the Gigabyte GA-B250-FinTech gave me smooth, perfect operation - totally bulletproof, no crashing, nothing. Afterburner would load minimized at start-up and send Power/Clock/Mem/Fan settings to all 12 cards, then Nicehash 2.0.10.0 would load and start mining.

Once I was done with burn-in testing a few days, I swapped the 7700k for G3930 - and it was a total lagfest in terms of UI responsiveness and programs choking and loading slowly, because the CPU simply couldn't keep up. Afterburner would take 2 extra minutes to load at startup and send settings to the cards, excavator would lag.

I still have the G3930 installed on that rig with the Gigabyte GA-B250-Fintech, and Nicehash runs fine with the same algo performance as when the 7700k was installed, but the OS responsiveness and program loading behavior remains super slow.

So I'm trying to determine now the cheapest Skylake/Kaby I can get away with in combination with 12 GPU's that gets me most of the responsiveness back. FWIW, the CPU issue could all be moot when mining with Linux/Ubuntu instead of Windows 10. I may experiment with that next.
 
Last edited:
odditory

Thanks for sharing that experience. I sure didn't consider that could be a problem, since typical CPU use was still less than 50%, but I didn't test anything faster either. Makes me wonder if that infuriating 3 minute delay between switching through AMD Catalyst Driver tabs with the RX580 setup trying to manage Wattman overclock/undervolt settings was related to my G4560 CPU??? I now wish I would have had a chance to test a better CPU against the advice you note.

But at the end of the day - it's not only the increasingly sluggish Windows UI that bothered me. It's all the weird little bugs in total, the random crap that happens, that doesn't seem to happen with less cards. Two 1080TI on my gaming machine I don't think have ever crashed - doesn't matter the cards, two cards is just solid. 4, 6 or 8 cards are pretty darn stable, but seemingly less stable each time you add cards. With up to eight cards you can address everything individually in Afterburner, I've never seen any start button issue nonsense, Edge launches on command, Explorer launches fairly immediately, fewer WHEA logger errors with the Biostar boards in the System event log, etc. By the time you get to 12 cards on any of these boards - it's like the errors exponentially compound, and you get little wayward bugginess from every angle. If that's all CPU related then I've been bested by an issue I didn't even know I had...

My cousin has used three different boards and up to 11 cards in his. Adding the ASRock 110BTC Pro (13 GPU board). He shares my opinion that 6 or 8 is the magic number to keep it mostly headache free.

However - he uses Celeron's in all his systems -- and I know his CPU is 70-80% routinely and his UI experience is even more rubbish. So maybe you're on to something.

---


I use nicehash 2.x now. I went back to legacy for a short while, after the NiceHash hack when 2.x didn't have the most profitable algorithms, but in the last couple months it's been mostly 2.x
 
Last edited:
Archaea said:
I have now tried three different 12 card mining rigs on two different motherboards models, (4 different individual boards)



My conclusion is that 12 card mining rigs are not worth the effort with Windows 10. Last week I consolidated 12 of my very stable 1080TI from two different rigs into a single 12 card rig to try it. I didn't meet success.

.
Click to expand...


What do you consider stable? Over what time period would you see problems?

Im running 12x 1060s and more recently 11x1060s and 1x 1080 on an ASRock H110 Pro BTC+, I have taken it offline at least once a day to fiddle. I havent seen any problems, but maybe it hasnt been up at long enough stretches.


edit: Hah 10mins after posting this, it looks like it crashed. Now that is a skilled jinx.
 
Last edited:
On power: I had three 120v circuits fully subscribed. Dropped a 240v 50a subpanel. That was a pain in the ass, I'm not joking. Works great. All my PSUs are EVGA G- OR P-.
 
Joust said:
On power: I had three 120v circuits fully subscribed. Dropped a 240v 50a subpanel. That was a pain in the ass, I'm not joking. Works great. All my PSUs are EVGA G- OR P-.
Click to expand...
Keep in mind: you're going to need some kind of PDU if you go this route.
 
urbsnspices said:
What do you consider stable? Over what time period would you see problems?

Im running 12x 1060s and more recently 11x1060s and 1x 1080 on an ASRock H110 Pro BTC+, I have taken it offline at least once a day to fiddle. I havent seen any problems, but maybe it hasnt been up at long enough stretches.
Click to expand...
By stable, I mean hands completely off. I had a run of about 3 weeks as my most stable period with 12 GPUs on those 1060, and that was my most reliable 12 card config.

What would happen through all these 12 card rigs is occasionally a card would drop and so just 11 would be mining, or the mining client would just be idle - having stopped for some reason, and require me to manually kick it off again. I didn't have any software to tell me when it stopped, so it was just when I happenstance checked the stats and it annoyed me that my 12 card rig wasn't doing anything. For a time it seemed some algorithms in nicehash and awesome miner didn't actually support 12 cards, and so you had to disable that algorithm on 4 of the cards.

I set up a Scheduled Task to reboot the PC once daily and that seemed to help a lot, but 6 or 8 GPUs just seemed/seems more hands off.
 
One of the reasons I use SMOS. My 12 card rigs don't go down. I once a month do maintenance on my rigs so I myself shut them down one at a time. I blow the dust out and check connectors. Your using Windows so have you tried Watchdog? From what I read if a gpu stops Watchdog knows and reboots the rig.

I am going to add 3 more gpu's to one of my rigs to make it 15 cards. SMOS supports up to 15 gpu's. I also use DSTM miner. And they updated it to v0.60.
 
Last edited:
I own the Asus Mining Expert and if you want to make it responsive you will need to enable the iGPU. Mine runs a Pentium G4400 and I don't see it spiking pass 10 % with 12 cards. The only thing is that you will need to enable iGPU and connect the HDMI to the mobo every time you restart that system otherwise it will become sluggish even though the iGPU is enabled. You can disconnect the cable after the restart. I cant see the difference in responsiveness between that system and my 10 core Xeon on the Asus Rampage V which is my gaming pc. I also run the Biostar BTC_Pro with 10 cards so far as the latest build. Its a great board and the latest bios will give you the same riser monitoring screen after post as the Asus Mining Expert. :D I don't find the NVidia any easier to set up or run 10 x EVGA 1070's +2 x 1080 TI between 2 other systems except the fact that AMD 580 need to have the mem straps modified. The 17.12.1 amd drivers introduced wattman profiles so setting it up once works flawlessly even on all the Fury X's that I mine with.
 
DId you guys notice that the new Gigabyte Finfet mining board is a direct copy of the Biostar BTC-Pro. The only difference is 4 mem slots On the Gigabyte if I'm not mistaken. I just wish Gigabyte got their bios in order and it could be another great board.
 
Imhotep said:
I own the Asus Mining Expert and if you want to make it responsive you will need to enable the iGPU. Mine runs a Pentium G4400 and I don't see it spiking pass 10 % with 12 cards. The only thing is that you will need to enable iGPU and connect the HDMI to the mobo every time you restart that system otherwise it will become sluggish even though the iGPU is enabled. You can disconnect the cable after the restart. I cant see the difference in responsiveness between that system and my 10 core Xeon on the Asus Rampage V which is my gaming pc. I also run the Biostar BTC_Pro with 10 cards so far as the latest build. Its a great board and the latest bios will give you the same riser monitoring screen after post as the Asus Mining Expert. :D I don't find the NVidia any easier to set up or run 10 x EVGA 1070's +2 x 1080 TI between 2 other systems except the fact that AMD 580 need to have the mem straps modified. The 17.12.1 amd drivers introduced wattman profiles so setting it up once works flawlessly even on all the Fury X's that I mine with.
Click to expand...
Mine are all using the g4560 integrated GPU permanently connected via HDMI to DVI cable to a DVI KVM switch. Still 40-50% CPU usage - depending on algorithm with 12 cards.

Interesting that the Biostar BTC_Pro now supports the bios check screen! I have one of those boards still I'll have to check that out! I noticed on all of three of my Biostar boards that I seemed to get WHEA Logger errors at intervals in the Windows System Even Log. That's not something I see on the Asus B250 motherboard. Do you see the WHEA logger errors (timeouts on the PCI-E bus) on the Biostar TB250-BTC Pro board? I never quite figured out if it was a riser problem, or expected with those motherboards? But the more cards I had plugged in, the more I saw those errors in the System event log.
 
None of mine get these errors. Is it amd or NVidia rigs, how much memory are you running. All of my Nvidia rigs run higher CPU utilization which is normal.
 
Imhotep said:
None of mine get these errors. Is it amd or NVidia rigs, how much memory are you running. All of my Nvidia rigs run higher CPU utilization which is normal.
Click to expand...
I got Whea logger errors with both AMD and Nvidia
I’m now running only Nvidia.
 
You must log in or register to reply here.
Back
Top