Quote from: greatauror28 on January 19, 2018, 05:51:52 PM

Quote from: Und3r74k3r on January 18, 2018, 03:12:41 AM

Hi guys. I have this RIG:
6x 580 Nitro+ 8GB Hynix
Asrock H100 Pro BTC+
Using a custom BIOS. Concrete modded my original "silent" bios with Polaris 1.6.6 with "1 click" metod. For OC and undervolt using "OverdriveNTool"

Using Adrenaline drivers.

Have 180 mh/s

1- 1200/900 y 2100/900 - ASIC 66.8% - 0 memory errors

2- 1200/900 y 2150/900 - ASIC 72.8% - 0 memory errors

3- 1200/900 y 2100/900 - ASIC 74.6% - 0 memory errors

4- 1200/900 y 2150/900 - ASIC 79.4% - 0 memory errors but today have "Hangs in OpenCL call" the other days i can tested no errors in some hours

5- 1200/900 y 2100/925 - ASIC 68.6% - 0 memory errors but today have "Hangs in OpenCL call" the other days i can tested no errors in some hours

6- 1200/900 y 2100/950 - ASIC 79.5% - 0 memory errors. This GPU always have "Hangs in OpenCL call". In few days i will recieved a new riser to try. Yesterday and today no Hangs.... I am very confused

Max time mining without errors is 10 hours... I try for 2 months config this rig, i desperate..

1. Update Win10 fully to FCU.
2. Trash any other OC software.
3. Upgrade (or downgrade) to AMD 17.10.3
4. Use Claymore v10 and set all clocks and voltages there.
5. Profit

Note:
-this assumes you know how to set GPUs to compute
-dont forget to use PBE v1.6 to mod
-dont forget to use ati patcher

I install 17.10.3, no OC Software and Claymore 10, get more errors with same OC and undervolt settings.

Quote from: Piskeante on January 19, 2018, 07:30:36 PM

hello.

actually i have 2 Sapphire RX 580 Nitro + 8gb running (a third MSI RX 580 Gaming X 8gb is coming on monday (it's was a bargain considering prices nowadays). Just a few tips for those that read this.

Afaik 850mv on the core is just about "right" in many scenarios. both of my cards cannot stay up with that voltage for a long time. Moreover, it happens to me than whenever my internet fails, the cards hang. Why? Because the voltage offset actually afects iddle voltage so, if the card uses let's say 100 on idle, if the internet hangs, the miner stops, the cards tries to go to 300 mhz on the core, and than the offset is applied so 100 minus offset = crash.

According to my findings, my cards can run for more than 50 hours (damm temperature hang error) with 880mv on the core at 1240mhz. The memory (Hynix and Micron) at just fine at 2250mhz without any issues at all. My micron card is perfectly fine at 2250mhz on the memory at 865mv (i modded that on the bios, no issues). For the hynix, i have it at 900mv (also modded). Stock setting for that card was 1000mv.

With DCRI 10 those 2 cards are actually hashing at 32.1mh/s. Both are using like 88W on the core and 60 for the other components, so around 150W total GPU power consumption. Solo mining. it's possible to reduce power consumption by undervolting (also reducing core speed), but my excel tells me that it's better to use a bit more Watts and squeeze the best out of it, because usually more hash rate is better than too much undervolting.

Just to clarify the numbers i'm talking about. For a reduction from 880mv to 856mv on the core, in case you are worried about electricity, you just save around 6W every hour. also ASIC quality is important, both of them are at 75% but if you win the silicon lottery maybe you can go way down undervolting. The best thing of this, is that once you know exactly what to modify in the bios , it's easy to get what you want.

My best advice, (if the cards are going to go for mining only), is to test them with certain settings, make sure they are stable, and then put that config into the bios, so that you don't have to apply manually the settings using MSI Afterburner.

I use original "Silent" BIOS editing by Polaris 1.6.4. Only editing the timmings by "1 click" method.

To OC and undervolt i use OverdriveNTool