GPU gave incorrect result regardless of the OC

[whotheff]

90% of the times my GPU 4 (5) submits shares, the calculated result is incorrect.
It's 1060 6GB Gigabyte Aorus.

I tried with no difference:

1. GPU/mem running with lower clock, higher clock, no clock
different power limits, overvolt, undervolt.
2. Running directly with no riser.
3. Running it on a seperate PC.
4. Replacing the thermal grease.
5. running on Claymore

Any advice please! I'm out of options!

[jillscarbrough]

You mean incorrect share? If it yes, mostly caused by OC, and or GPU memory error. Try to check your GPU memory error using HWINFO. Normally under-clock will solve this, but as you said it's isn't. so, memory error may most possible.

[whotheff]

You mean incorrect share? If it yes, mostly caused by OC, and or GPU memory error. Try to check your GPU memory error using HWINFO. Normally under-clock will solve this, but as you said it's isn't. so, memory error may most possible.

HWinfo shos only AMD memory errors. It cannot read Nvidia mem errors.
But even if there are memory errors, why even at stock clock?

*tried also flashing the VBios - no change

[trung1806]

Thanks for info

[jillscarbrough]

You mean incorrect share? If it yes, mostly caused by OC, and or GPU memory error. Try to check your GPU memory error using HWINFO. Normally under-clock will solve this, but as you said it's isn't. so, memory error may most possible.

HWinfo shos only AMD memory errors. It cannot read Nvidia mem errors.
But even if there are memory errors, why even at stock clock?

*tried also flashing the VBios - no change

I'm shy, Sorry didn't see you have 1060. Try to Reduce memory frequency, under a stock, if it solves you prob, improve it back by stepping.

[JaredKaragen]

You mean incorrect share? If it yes, mostly caused by OC, and or GPU memory error. Try to check your GPU memory error using HWINFO. Normally under-clock will solve this, but as you said it's isn't. so, memory error may most possible.

HWinfo shos only AMD memory errors. It cannot read Nvidia mem errors.
But even if there are memory errors, why even at stock clock?

*tried also flashing the VBios - no change

I'm shy, Sorry didn't see you have 1060. Try to Reduce memory frequency, under a stock, if it solves you prob, improve it back by stepping.

does this card produce errors in a separate machine running by itself?

[whotheff]

You mean incorrect share? If it yes, mostly caused by OC, and or GPU memory error. Try to check your GPU memory error using HWINFO. Normally under-clock will solve this, but as you said it's isn't. so, memory error may most possible.

HWinfo shos only AMD memory errors. It cannot read Nvidia mem errors.
But even if there are memory errors, why even at stock clock?

*tried also flashing the VBios - no change

I'm shy, Sorry didn't see you have 1060. Try to Reduce memory frequency, under a stock, if it solves you prob, improve it back by stepping.

does this card produce errors in a separate machine running by itself?

Yes, tried it on a seperate mb, directly in pci-e slot with claymore ( I use ethminer) and it made 77 shares for 6 hours.
There were a lot of red lines warning for incorrect shares. Clock was devreased from 700 to +500mem.

[JaredKaragen]

has the BIOS been modified at all?  if so, put the original one back on it to test again.


If not, or yes... either way; 
it sounds like it could be an issue in the bios with a bad flash or a flash that is too much for your degraded hardware; or the GPU/ram may have suffered some damage either due to simple failure over time, or due to overclocks/temps...

Not saying what it is,
but the symptoms are that of a card with a bad bios or its been overclocked too much too long and its circuits degraded in the die;  circuits in either RAM or the GPU die.... especially if all at stock settings, fresh drivers, and different miner apps/algos produce the same results.

I have come across a few open box GPU that were overstressed and threw nothing but errors and games were unreliable or glitchy.... and have also ran across second hand GPU in CL purchases that behaved as you mentioned..... tis why I bring a test system when I buy GPU on craigslist.  I always make sure I can properly test it at stock settings before I buy.

[whotheff]

Thanks for the toughts Jared, if nobody else has same problem, either they never check how much their rigs should
make or I'm the only one. Now my crappy mixed pile of 6 different brands 1060s (hynix included) beats
all brand name parts rig with 7 cards

Out of desperation i updated the bios yesterday with the official beta bios 006 from Gigabyte.
Before that it was original. Now i get the same hashrate as before by adding 200Mhz instead of 700 (probably a MSI AB bug).
However GPU 4 still produces incorrect calculations.

The last thing i can try with this card is use it for equihash/other script or just.. play.. games?

Anyway, if anyone has ever had such issue and solved it, please share your experience.

[JaredKaragen]

Thanks for the toughts Jared, if nobody else has same problem, either they never check how much their rigs should
make or I'm the only one. Now my crappy mixed pile of 6 different brands 1060s (hynix included) beats
all brand name parts rig with 7 cards

Out of desperation i updated the bios yesterday with the official beta bios 006 from Gigabyte.
Before that it was original. Now i get the same hashrate as before by adding 200Mhz instead of 700 (probably a MSI AB bug).
However GPU 4 still produces incorrect calculations.

The last thing i can try with this card is use it for equihash/other script or just.. play.. games?

Anyway, if anyone has ever had such issue and solved it, please share your experience.

play some games... see if you get artifacts.....

[whotheff]

Finally had a chance to test it in 3D. Used Furmark.

With the stock clock it worked for more than 6-7 minutes with no issues.
However, when i pushed the clocks to the one i use on all other cards
it needed about 10 seconds before it showed artifacts and the driver restarted.
I tried other clocks then and if the core clock is above 100mhz it eventually crashes.

Then tried all possible combinations of TDP, mem and core clocks
while mining. It always displays errors for wrong calculations.

For one hour about 80% of the results were wrong.

Definately I'll try to RMA it.

[JaredKaragen]

Finally had a chance to test it in 3D. Used Furmark.

With the stock clock it worked for more than 6-7 minutes with no issues.
However, when i pushed the clocks to the one i use on all other cards
it needed about 10 seconds before it showed artifacts and the driver restarted.
I tried other clocks then and if the core clock is above 100mhz it eventually crashes.

Then tried all possible combinations of TDP, mem and core clocks
while mining. It always displays errors for wrong calculations.

For one hour about 80% of the results were wrong.

Definately I'll try to RMA it.

See if you can RMA it;  but I am under the impression that if it performs fine while not overclocked;  its technically "in spec" although the ASIC has definitely seen batter days or just has never been of that good of quality.  Hopefully they let you get it replaced!

If not you will probable need to make a custom clock setting for it separate from the others.

[whotheff]

The problem is it submits wrong shares regardless of the OC.
Otherwise i would have just slowed it down with 100Mhz

[JaredKaragen]

The problem is it submits wrong shares regardless of the OC.
Otherwise i would have just slowed it down with 100Mhz

ok, that was lost in translation when I read your reply earlier.

Definitely a damaged ASIC;  sadly, its usually from applying an overclock too skookem for the ASIC's quality.  Fingers crossed on the RMA.