Hashrate drop on multigpu rig 12 x Asus strix rx 580 8GB samsung

[Daniel0303]

Motherboard Biostar b250 pro btc
amd 17.11.1 driver

Hashrate drop from 30 to 25 sometimes on eth

Any idea why this happens?

[wacko]

There's a few reasons for this to happen. Gotta monitor the rig to figure this out. What's going on with the clocks and temperatures when this drop occurs? Is it all 12 GPUs or only a few particular ones? What's the CPU load at these moments? Any HW memory errors (HWInfo)?

[Metroid]

It's normal.

[Unknown Suspect]

Yeah I have one rig with eight RX580's that runs very smooth at 29.x MH/s for ETH. Every once and a while GPU #6 will drop to 25 MH/s for a few minutes then goes back to 29.x MH/s.

[Metroid]

That is an issue concerning memory timings, rock stable memory timings will not do that. So get used to it if you modified memory timings, nvidia cards are not like that cause you can't change on nvidia cards.

[steve8lily]

There's a few reasons for this to happen. Gotta monitor the rig to figure this out. What's going on with the clocks and temperatures when this drop occurs? Is it all 12 GPUs or only a few particular ones? What's the CPU load at these moments? Any HW memory errors (HWInfo)?

How many memory errors must fix? I have6GPU rig, 1 GPU memory error 1.7mils which run for 10hours...

[wacko]

How many memory errors must fix? I have6GPU rig, 1 GPU memory error 1.7mils which run for 10hours...

You're not supposed to have any memory errors at all. GPUs might keep running and mining for days even when there are some HW errors, but that will eventually lead to the system instability, it'll crash/reboot at some point. And it might also affect your actual sharerate. So normally you should aim at 0 errors.

[steve8lily]

[/quote]
You're not supposed to have any memory errors at all. GPUs might keep running and mining for days even when there are some HW errors, but that will eventually lead to the system instability, it'll crash/reboot at some point. And it might also affect your actual sharerate. So normally you should aim at 0 errors.
[/quote]

Can anyone share with me how to resolve these memory errors via amd WattMan as I dunknow how to use it?

[wacko]

Can anyone share with me how to resolve these memory errors via amd WattMan as I dunknow how to use it?

These memory errors are due to memory overclocking and/or due to using too aggressive memory timings. The easiest way to resolve this is to start decreasing the memory clock until the errors are gone.

[Daniel0303]

There's a few reasons for this to happen. Gotta monitor the rig to figure this out. What's going on with the clocks and temperatures when this drop occurs? Is it all 12 GPUs or only a few particular ones? What's the CPU load at these moments? Any HW memory errors (HWInfo)?

Hi, i haven't check hwinfo, temp are ok, monitor core is hard for me to do, gpuz took too long to load
The problem now is i got 10 xfx with samsung memory, run with stock bios also drop in hashrate
a few gpu drop hashrate once in awhile in the end all of them ever dropped at least once in 1mins

Would be grateful if you could give some advise