AMD RX 6800 blocks at 57 mhs / low hashrate

[stefjoe]

Hi i am absolutely new to mining and gpus, so please be patience

i bought 4 RX 6800. Last week i connected the Gigabyte RX 6800 and i get 62 mhs / 102 W, that was pretty perfect for me! Ok yesterday i get 3 new RX 6800, i put the same settings like the Gigabyte, but now ALL, also the Gigabyte, doesn’t go over 57 mhs. The wattages are around 97-100 W and they are loud and hot.
I tried different settings, but no chance to get over 57-58 mhs. First i thought that the PSU was the problem, but i have the Seasonic TX 1000 80 Plus Platinium, i think it is a good choice.

What coudl be the problem? Thx for help

[costis]

Heloo…Leave only one of the 4 video cards on the platform. Try one of the 3 new books you bought!

[stefjoe]

Hello, thx for your reply Yes i tried, same problem. Also i changed the Riser, cables PSU … always the same. My equipment is absolutely new, no used hardware

[Grea]

If you search for 6800 configurations and optimizations and eventually you’ll arrive here:

RX6800 - Efficient Overclocking

Hello again. What i found today after i installed a new rx6600xt yesterday. Then again the SOC frequency dropped at rx6800. That’s why I started experimenting with frequencies and found that one of the important ones is F state. With the help of “UPP: Uplift Power Play” I increased the minimum frequency to 1551Mhz. Now again I have about 62MH / s. The main thing is that after each restart we have to apply this setting again, also after applying via upp we have to apply the OC…

In my experience, until you run the scripting above, you are better off
Core 1275-1300 with core voltage 650
Mem 1060-1074 with mem controller voltage 750 and memory voltage 1250
fan 66%
SoC will either be right or rarely stick until you run the scripting
Ditch the DPM setting entirely

Net: 1275 Core / 1060 Mem can produce over 61.5Mhs with TRM on Hive, so they are not your hard limiters.

[costis]

You can use my setting

[stefjoe]

@RX6800 - Efficient Overclocking

1551

thx for your help. i also tried yours, but the hashrate goes down to 39 mhs and 54 W. I deleted the SOC, but after saving the setting, there is a 0. Is the 0 for default?

[stefjoe]

thx again, with your settings the hashrate goes down to 12 mhs and 54 w. What could be the problem?

[Grea]

The graphical user interface of HiveOS does not show all the data.

Post a picture of both the miner interface and the HiveOS settings, like this:

[stefjoe]

it took some time for me to find out, how to do it, but here we go

[Grea]

One last thing you’ll want to access for tweaking: launch Hive Shell and execute “amd-info” to get this sort of output. Note: This is a sub 62MH/s performer, with F state below 1551:

Out of curiosity, did you work up to 1350 from a lower Core speed? Some of mine absolutely refuse to run higher than 1300, as depicted, prior to the scripts linked in the other thread.

fwiw: I am not sure where throttling starts, but have to guess AMDs running at 80C are there.

[stefjoe]

Yes i tried a lot! I checked the first 10 popular presets, but no one is working for me. Always the same, however here is my screenshot.

[Grea]

Miner and the amd-info both say the Memory state is 1000. Does the HiveOS GUI still stay you have it at “1150”?

Core and amd-info of GPU3 says 1250. Does HiveOS GUI still say “1350”?

I have been avoiding presets after getting spurious entries like those “DPM” states in your screens. Not sure if that is also causing issues, but they are not helping you.

I will share, I found it too time consuming to trouble shoot my mixed rig attempts. Not sure if that is contributing to your challenges or not, but your GPUs appear to be running lower settings then you are specifying in the GUI.

[stefjoe]

Thank you so much for your time and help!

GPU 3 is my playground and i set 1250, that is ok. But Memory is really strange, hope to find a solution. Will also search for it

[stefjoe]

When i change memory to 1075, the memory in amd-info is changed to 1069, BUT the hashrate is round about 13 mhs. Also i tried to change the driver version, but also no success.

That is really strange, because the GPU-0 run with 62mhs before i connected the other 3, when i disconnect them nothing is changing, also a new PSU or connecting to another PSU.

[Grea]

I believe 1075 is absolute max Mem value for a 6800 in Linux, but it appears your card is staying under at 1069. Just be cautious as you move it up.

Would the low performer be powering a local monitor output? Is it being driven by the same power supply this whole troubleshooting time? Should be easy enough to remove another GPU from the power supply in question to see if it is power related.

I have few poor performers, most were helped with the F state script linked before. <what is the F-state on the poor performer?> Some still require a tweak up above the value they “like” and then a tweak back down to their sweet spot. Lost an hour messing around with (2) of (8) in a rig last night. Took a couple of reboots to get the right performance to stick. Thankfully they have historically proven fine until the next change I make.

fwiw: My good ones work with anything, some are touchy, and a couple are a down right pain in the rear. If you are still running in a mixed rig, you are outside my success area

[stefjoe]

thank you very much, my local monitor is connected to an nvidia 3070, it shouldn’t be a problem to switch it.
as psu i use the Seasonic TX1000 prime 80 plus titanium for 3 amd cards, I think they are very good. i have no problems with the nvidia, they all run very well.

I’ve already tried a new PSU and only 1 card on the PSU, also with the nvidia, but nothing works. Riser and cables reconnected and and and.

I found this command in a ticket, but I still don’t know what it will bring and where to carry it out, I still have to find out more

RX6800 - Efficient Overclocking

Reboot, dont shutdown, only reboot [final] I am using this commands in teamredminer: –eth_dag_slowdown=9 --eth_stagger --eth_ramp_up

–Eth_dag_slowdown = 9 --eth_stagger --eth_ramp_up

The strange thing about the whole thing is that the gigabyte ran at 62 mhs and I didn’t change it. I order another mainboard and split up the GPUs.

I hid one thing, I upgraded to the new HiveOs

[Grea]

@stefjoe

my local monitor is connected to an nvidia 3070, it shouldn’t be a problem to switch it.

I wouldn’t, some of my OS’s definitely back off when a monitor is detected on a GPU. Not sure about Hive. If a monitor is NOT on that 6800, ignore the question for now. Unless it turns out to help

@stefjoe

I found this command in a ticket, but I still don’t know what it will bring and where to carry it out

I run similar delays with rigs with VII’s, but I don’t have it in my 6800 only rigs. Not sure it is necessary.

Have you been running the F-state scripts from the other script and it is locking at 1551?

Have the latest HiveOS running on one of my 6800 only rigs, it was touchy to reboot. But it came back up >97% of it’s highest MHs. Don’t have data to say it has any issues.

[stefjoe]

i almost did it! I run the script and 3 of my 4 gpus are running better! They were at 64 mhs

Only the last gpu makes trouble! I run the script, after reboot, the F settings were deleted, now the mhs ar erunning without the script. How i said, only the last gpu makes trouble right now :-/

[Grea]

Depending on your goals, I’d consider yanking it from the rig and running it under windows, in a single GPU rig, etc.

Time to ETH is a bit of a race right now and extended downtime is much worse than 5% increases.

[stefjoe]

received my new card today after connecting 61mhs / 105 W and no problems, also i need not to run the script now