Weird Sapphire RX570 Nitro+ problem [bounty for a solution]

[arbenowskee]

Would love it, if someone would be able to help me with this one.

I built a mining machine (7x rx570 4gb nitro+, MSI Z170A pro carbon, 120gb ssd, 4gb ram, celeron G3900, superflower 1600W PSU, 7x riser) and put it in homemade open air case.

I installed windows on it, flashed bios on GPUs, configured all the details (4g encoding on motherboard, copied older version of drivers to new folder so windows detects more than 5x rx570 etc) and started mining. Was happy to get about 29 MH/s in dual mine mode. I quickly found out that one of the cards was faulty (got memory errors even with stock bios and stock frequencies) so I sent it back and happily mined with 6 cards. Also cranked up the fans, so temperature on each GPU was about 50 degrees celsius.

I got the card back, but this one was faulty as well (rig would not even turn on, even when switched to different PCIE slot, different riser and switched to secondary bios). So I sent it back as well.

Then the supplier told me, that they could not supply the chosen GPU if I would be OK with Gigabyte version of RX570. I said sure, as long as I got something right?

Now the weird shit started to happen. I hook up the 7th card and screen stays black. I turn the machine off, disconnect the card, turn it back on and windows started to repair the disk. I leave it for about an hour, the repair progress is still at 0% so I decided to reinstall it. I reinstall win 10, do all the configuration with drivers, start mining with original configuration and I notice 2 cards were mining with 22 MH/s. But it was late and I said, fuck it, I'll deal with this shit tommorow. The next morning I check mining status and see that the machine stopped mining at about 3 am, screen was black. I turned the machine off and than turned it on again - nothing.

So I try each card by itself - same results. I use the cards bios switch to switch to secondary (the quiet one) bios and windows starts up normally. I flash the primary bios again, thinking something has gone wrong, with the primary bios, restart the computer and again nothing - black screen. I switch the bios again to secondary and I am thinking maybe the bios chip is somehow damaged. I flash the secondary bios with original primary bios and now I get black screen even with that. At this point I stopped playing around and opened this thread.

Any ideas what has gone wrong and why the cards wont start up? I will probably do the 8+1 trick and flashed the secondary bios to the original one, but I am abit in the dark here.

I am willing to reward the person that gives me a solution =)

[Undefined31415]

Have you checked if any of the cards work in a separate system? (One that you are certain has working components.)

Have you been able to boot up the system using integrated graphics? (Trying to confirm that the problem is with the cards themselves or their interactions with the mobos)

If you can, I would revert all the cards to the stock bios version.

[arbenowskee]

Yeah, computer boots normally with integrated GPU. It boots to Windows and Linux.

The cards do not boot up with stock bios either, that is what is weird to me. I don't have any other machines at the moment, but I'll find one. =)

[bobben2]

Did you check with just one card directly plugged into mobo?
You may have a riser problem with two or more cards.   I discovered that my risers would not work until I made sure the usb plug was properly seated.  I would revert back to stock bios on all cards and have the rig run stably for a while before trying any bios mod.

[arbenowskee]

The problem is, it does not start with original BIOS . Anything I flash on primary bios, it does not work. I wont flash secondary bios, because I've bricked one card (which I will try to recover using 8 + 1 method).

I've tried plugging the card directly into motherboard, I've taken out CMOS batter, so all settings were reset. I am really at loss here. Could somekind of power surge or something else physically damage any connections card has from primary bios?

This is getting ridicules =)

[hagbase]

All how tos and vids told me to flash  silent bios with custom rom not the oc bios.

All cards flashed with https://anorak.tech/c/downloads/sapphire   custom powersave bios.

Any device in windows device manager?
Do you see your card in ATIFlash if it's connected to your mainboard without riser?
Is any of your cards able to boot up and mine with default oc bios?

[P00P135]

Did you try actually resetting CMOS instead of just taking the battery out?

[Cereberus]

I configured a few rigs with these exact specific cards and I had problems at one board, freezing, black screen, restarting and all kind of errors. I just downloaded DDU here http://www.guru3d.com/files-details/display-driver-uninstaller-download.html removed the display driver thoroughly and then did a clean install of the latest AMD driver with all 6x cards on the rig. Then everything started working smooth with Nicehash as I use that software for mining. The boards were Asrock H81 PRO BTC and until now I have not seen a better motherboard for mining and trust me I have seen a lot and configured a lot of them.

[arbenowskee]

All how tos and vids told me to flash  silent bios with custom rom not the oc bios.

All cards flashed with https://anorak.tech/c/downloads/sapphire   custom powersave bios.

Any device in windows device manager?
Do you see your card in ATIFlash if it's connected to your mainboard without riser?
Is any of your cards able to boot up and mine with default oc bios?

I can boot them up with silent bios, I tried even mining with them (at least 1 by 1) but not with OC bios.

[hagbase]

Ok so cards are not dead yet, maybe just a bad flash.

If you switch to OC Bios is your system able to boot or instant bsod/reboot?
Do you see the device with OCBios in device manager/ATIFlash? Try to restore your backup of Sapphire OC Bios with ATIFlash/CLI as described here:
https://anorak.tech/t/how-to-flash-a-amd-graphic-card-vbios/15/8


 

[arbenowskee]

I don't understand why would it be a bad flash, I did not flash them when the computer crashed.... I flashed them a while back and they were running OK for quite some time. The problems began when I reinstalled windows and computer crashed in the middle of the night.

Anyway with OC bios, system does not boot up at all. I tried to boot them up with silent BIOS, than switched to OC bios while computer was running and flashing it. Flash succeeds, but computer does not start up, no matter which OC bios I use (stock, 1500 memshift, 1750 memshift).

[h311m4n]

Have you tried to simply change the risers? One bad riser can make an entire system act funny.

I'm guessing you used DDU to remove and resinstall the driver? Patched it so it sees modded cards and all?

Have you tried changing RAM sticks?

[hagbase]

Maybe be more precise next time?

than switched to OC bios while computer was running and flashing it

Anyway good luck.  I'm out.

[Inka-mining]

Would love it, if someone would be able to help me with this one.

I built a mining machine (7x rx570 4gb nitro+, MSI Z170A pro carbon, 120gb ssd, 4gb ram, celeron G3900, superflower 1600W PSU, 7x riser) and put it in homemade open air case.

I installed windows on it, flashed bios on GPUs, configured all the details (4g encoding on motherboard, copied older version of drivers to new folder so windows detects more than 5x rx570 etc) and started mining. Was happy to get about 29 MH/s in dual mine mode. I quickly found out that one of the cards was faulty (got memory errors even with stock bios and stock frequencies) so I sent it back and happily mined with 6 cards. Also cranked up the fans, so temperature on each GPU was about 50 degrees celsius.

I got the card back, but this one was faulty as well (rig would not even turn on, even when switched to different PCIE slot, different riser and switched to secondary bios). So I sent it back as well.

Then the supplier told me, that they could not supply the chosen GPU if I would be OK with Gigabyte version of RX570. I said sure, as long as I got something right?

Now the weird shit started to happen. I hook up the 7th card and screen stays black. I turn the machine off, disconnect the card, turn it back on and windows started to repair the disk. I leave it for about an hour, the repair progress is still at 0% so I decided to reinstall it. I reinstall win 10, do all the configuration with drivers, start mining with original configuration and I notice 2 cards were mining with 22 MH/s. But it was late and I said, fuck it, I'll deal with this shit tommorow. The next morning I check mining status and see that the machine stopped mining at about 3 am, screen was black. I turned the machine off and than turned it on again - nothing.

So I try each card by itself - same results. I use the cards bios switch to switch to secondary (the quiet one) bios and windows starts up normally. I flash the primary bios again, thinking something has gone wrong, with the primary bios, restart the computer and again nothing - black screen. I switch the bios again to secondary and I am thinking maybe the bios chip is somehow damaged. I flash the secondary bios with original primary bios and now I get black screen even with that. At this point I stopped playing around and opened this thread.

Any ideas what has gone wrong and why the cards wont start up? I will probably do the 8+1 trick and flashed the secondary bios to the original one, but I am abit in the dark here.

I am willing to reward the person that gives me a solution =)

UBUNTU 16.04 Server with AMDGPUPRO and stop suffering

[Culture]

install amd driver 17.7.1 (no mixed drivers) then apply the pixelclockpatcher (renamed to only sign the driver), then reboot. 

[arbenowskee]

Have you tried to simply change the risers? One bad riser can make an entire system act funny.

I'm guessing you used DDU to remove and resinstall the driver? Patched it so it sees modded cards and all?

Have you tried changing RAM sticks?

I have tried to plug the graphic card directly into the motherboard. same result

UBUNTU 16.04 Server with AMDGPUPRO and stop suffering

tried linux, no money.

install amd driver 17.7.1 (no mixed drivers) then apply the pixelclockpatcher (renamed to only sign the driver), then reboot. 

I doubt its a driver problem, system does not boot up, meaning I don't even get past any bios/uefi screen.

[h311m4n]

Have you tried to simply change the risers? One bad riser can make an entire system act funny.

I'm guessing you used DDU to remove and resinstall the driver? Patched it so it sees modded cards and all?

Have you tried changing RAM sticks?

I have tried to plug the graphic card directly into the motherboard. same result

UBUNTU 16.04 Server with AMDGPUPRO and stop suffering

tried linux, no money.

install amd driver 17.7.1 (no mixed drivers) then apply the pixelclockpatcher (renamed to only sign the driver), then reboot.  

I doubt its a driver problem, system does not boot up, meaning I don't even get past any bios/uefi screen.

Sorry if this has been proposed before but a system that doesn't boot up usually comes down to:

-RAM
-Faulty PSU/power cable. I had some second hand pcie -> molex cables I bought off aliexpress that fried a riser. When I would start the rig up, I'd here a "click clack" in the PSU.
-Motherboard

Drivers are probably not the cause since you're not even getting to the windows loading screen.

Have you also tried to disable everything like Audio, Ethernet etc. in the bios?

I'm asking because one motherboard in particular was a giant pain in the ass, it was an asrock anniversary, one with 6 pcie slots on 1150 chipset. I got it working once with 6 GPUs, but had to disable everything in the BIOS (audio, ethernet...). Then decided to do some changes, can't remember what it was, I thing I did a DDU to update some drivers, and never again was I able to get 6 cards working on that thing.