RX570 got incorrect share, when using stock BIOS

[drgrog]

EthDcrMiner64.exe is showing the "GPU #1 got incorrect share" warning on an RX570.
I understand this can be from OC, but I'm still getting it after reverting to original BIOS, and can't figure out why.

Some details/history:

- It's a Sapphire RX570 4G PULSE OC, 1 of 2 in the rig.
- Card is new. Got 23.5 MHp/s with stock bios (1284/1750, ~110W), no errors in 12hrs.
- Copied 1500 straps up, undervolt to 900, 1320/2020, got 27.3 MH/s ~90W, no errors but not 100% stable.
- Undervolt to 950, 1280/2000, got 27.0 MH/s ~95W, more stable, no errors.
- Removed GPU, installed other one to flash and optimise it, it's all good at 27MH/s with no errors.
- Mining with both cards installed, I started seeing "got incorrect share" whilst playing with voltages and getting things stable
- I get about 10 errors per hour, only on that card.
- GPU does find shares, about 10/hr, so ~50% good 50% incorrect.
- There are zero (0) GPU memory errors in HWiNFO64, which I thought was the cause of incorrect shares.

I've tried:

- Re-flashing original BIOS from backup, (so back to original straps and clocks etc.), still get 'incorrect share' errors (and low hash rate with high power
- Using original bios with minimal mods: copy 1625 straps upwards, 1280/1850 clocks, still get 'incorrect share' errors.
- Tried card solo or as part of a pair in rig, still errors only on this card.
- Tried plugging a monitor into that GPU.
- Tried playing with the -dcri value, hard to tell but MAYBE more errors at lower dcri values (eg. 1)

Other info:

- Pool shows my effective hash rate much less than my reported hash rate.
- Temperatures are in 60-70 range, fan 20-30%, so all good there.
- Double checked that CrossFire is disabled.
- Claymore v9.8, Windows 10 x64, Radeon 17.5.1 (No Afterburner or TRIXX installed)
- EthDcrMiner64.exe -epool us1.ethermine.org:4444 -ewal 0xWALLET.name -epsw x -allpools 1 -dcri 10
- Only mining ETH, so don't know if it affects other coins.

I'm a bit new to this, so what am I missing?
Could I have damaged the card when lowering voltage?

[1714830461]

1714830461

[1714830461]

1714830461

[1714830461]

1714830461

[1714830461]

1714830461

[1714830461]

1714830461

[Mattthev]

And what are your clocks? That's not a bios problem probably. Your memory OC is too high I would say.
You can use my guide for bios mod: https://bitcointalk.org/index.php?topic=1954245.0

[drgrog]

And what are your clocks? That's not a bios problem probably. Your memory OC is too high I would say.
You can use my guide for bios mod: https://bitcointalk.org/index.php?topic=1954245.0

Thanks Mattthev, I was getting errors with the stock clocks at 1284/1750, so not a case of mem OC too high.

The errors have gone away after a windows update of all things. According to the update history, it included driver updates for "Intel - SYSTEM" and "Intel - Other hardware", which I'm guessing was chipset or mobo related.

Haven't looked any further into it, as I'm happy it's been running without errors for a couple of days now.

[Metroid]

It might be solo elpida memory, so yes, even if is not showing memory errors some cards with memory modules poor yielded will give incorrect shares. There is no hope with elpida solo. Even if is using stock bios with stock clocks and voltage, problem still remains, try to change memory timings, down clock it might be the only solution, my guess is poor cooled memory modules cause that thing, so the only hope is to downclock your core clock 10% and check your power, the less heat generated the better.

Thanks Mattthev, I was getting errors with the stock clocks at 1284/1750, so not a case of mem OC too high.

Try 1000/1750 and see if there are still errors, if not then you go increasing to 1030 till you find the right spot.