RX480 gpu0 reports 4Mh/s only on linux

[piwi3910]

hi guys,

I have a weird problem i can't figure out.
i have a 5 GPU rig running on windows with claymore 7.4 nicely hashing at 30.1Mh/s
I now want to do the same but under linux using the AMDGPUPRO drivers

all seems fine running ubuntu 16.04
however without overclocking i get these values:
ETH: GPU0 4.176 Mh/s, GPU1 26.804 Mh/s, GPU2 26.776 Mh/s, GPU3 26.778 Mh/s, GPU4 26.868 Mh/s

I tried this with ubuntu 16.04, ethos, simplemining.
They all have the same issue GPU0 is super slow.

I also see the following difference when i check the clocks:

piwi@gpuminer01:~$ cat /sys/class/drm/card2/device/pp_dpm_mclk
0: 300Mhz
1: 2000Mhz *
piwi@gpuminer01:~$ cat /sys/class/drm/card0/device/pp_dpm_mclk
0: 300Mhz
1: 2000Mhz


Any clue how to fix this and gpu0 mining normally, like it does on windows.
It's also not the card, switching cards doesn't help.

[xxcsu]

try to connect your monitor or dummy plug to that card which one hashing with 4mh/s , reboot the system.

[piwi3910]

already had a monitor on it, same rig with windows 10 no issues.
tried it with a X session on the monitor and without.
no difference.
and it's always gpu0
enabled the onboard intel graphics to use with the monitor so that no screen was attached to any of the cards, same issue
if i swap cards around it's always gpu0 with the crappy hashrate, seems like it never goes to mode 7 even when putting the card in manual mode and setting mode 7, it doesn't take.
and what ever i do to over or underclock, i always get this message in dmesg:

VDDCI is larger than max VDDCI in VDDCI Voltage Table!


tried with drivers 16.40 till 16.60

[piwi3910]

anyone?

[Subw]

Try performance state

[o00o]

already had a monitor on it, same rig with windows 10 no issues.
tried it with a X session on the monitor and without.
no difference.
and it's always gpu0
enabled the onboard intel graphics to use with the monitor so that no screen was attached to any of the cards, same issue
if i swap cards around it's always gpu0 with the crappy hashrate, seems like it never goes to mode 7 even when putting the card in manual mode and setting mode 7, it doesn't take.
and what ever i do to over or underclock, i always get this message in dmesg:

VDDCI is larger than max VDDCI in VDDCI Voltage Table!


tried with drivers 16.40 till 16.60

I had the exact same problem where a GPU got stuck in a lower power state which caused that 4 MH/s bug on a 6x 470 rig operating Windows 10 just a few hours ago.  I solved the problem by simply connecting a screen to that particular GPU.  If that doesn't do the trick, try connecting to any other random GPU.

[bittawm]

solution:

Steps:

1. plug is a display port to hdmi plug to 4mhs card (display port to card hdmi to monitor)

2. take out the display port end and also put it into any other cards showing 4mhs (i had 5 cards of 6 do this on one rig, i have noticed its only saphire cards)

3. donate to bittawm, addy is in signature

youre welcome

[cashclash]

run into this, fixed via BIOS PCIe settings to AUTO, was on Gen1 settings

[tarnum]

Had same issue, it was about the damn risers, just mess with them a little, and reconnect carefully again.
It worked for me.

[lautre]

i had this problem on linux i had to switch GEN1 PCI-E to AUTO (as said in the HELP) in the bios and it fixed the problem.

i'm using onboard gpu for screen.

[dmatthewstewart]

i had this problem on linux i had to switch GEN1 PCI-E to AUTO (as said in the HELP) in the bios and it fixed the problem.

i'm using onboard gpu for screen.

you mean integrated graphics?

Is it an APU?

[coinzoid]

This issue happened to me a few times. And all the time it was resolved upon connecting a monitor to troubled gpu through hdmi port. I assume dummy plugs will work fine too but I don't have one to reproduce and test.

[justin.ramos]

I just ran into this. For me, the solution was to use:

Code:

GPU_FORCE_64BIT_PTR=1

Hope this helps someone.

[blacktux88]

have you installed correct drivers ??