 |
December 13, 2018, 08:09:57 PM |
|
Background on cards: Cards have all been bios modded with 1-click timing patch and pixel patcher before reboot. Card-by card and confirmed working with Blockchain drivers in Claymore mining ETH with expected Mh/s results. Your responses are incredibly appreciated as I've got myself an expensive paperweight situation on my hands as it sits in the off position until I figure something out.
Previously and Background on My Nightmare: Using Claymore 11.9 -cvddc 950 -cclock 1150 -mclock 2025-2060ish, I had the Blockchain drivers installed on an 8xRX580 8gb OC build and was getting 29.8-30.5mh/s per card [all cards working stable for over 36 hours]. I decided to try to DDU and upgrade to the latest version of Adrenalin drivers released early Dec to squeeze out a few more Mh/s. My friend and I built a pair of the exact same spec builds and he is on Adrenalin drivers with the exact same bios mod and all hardware/software installed and was getting 31-32Mh/s without memory errors [most importantly]. I was at 242Mh/s combined and stable. [hindsight is terrible. I never would have changed anything]
I crashed the computer on the fresh install of Adrenalin and had to restore, after multiple more attempts with unstable restore point and failed driver installs [blue screen and reboot repeatedly], I've reformatted windows 10 and reinstalled latest Adrenalin drivers just to give it a shot again with a clean install of windows, free of any lingering GPU driver or windows issues.
Current Nightmare Situation: Installed Windows 10 without ethernet and disabled windows updates, firewall, suspend and monitor shut off to mimic previous fresh and successful installs. Changed VRAM to 36,000+ Disclaimer: I know full and well most people recommend the blockchain drivers for stability, but I feel small next to my mining teammate who is running the Adrenalin drivers with almost 15-16 more Mh/s total than me with the same setup. All cards are set to compute mode, stock settings [nothing included in config or bat] for -cvddc -cclock -mclock. One card does not mine and is dead stopped [showing 0.0Mh/s] and the others will run at 29.5Mh/s. I've lowered the cvddc to 950 and 900 and can get it to mine at 14Mh/s, but it crashes with hang on OpenCL issues, which I know are related to OC or undervolting. nevertheless, it doesnt run on stock settings. I'm concerned I've damaged the card or done something so that now the bios are flashed inconsistently because lowering the core voltage has gotten some positive result. Potentially a bug in the drivers?
Questions: (1) Can installing new GPU drivers or windows modify the GPU bios in any way? I have never had this inconsistency before and do not have any software creating inconsistent mclock and clock settings unique to each card. The performance decrease on one card should not be this drastic [14mh/s]. I've confirmed all settings in Wattman are consistent with each other. No Afterburner installed and no custom settings between the cards yet. Simply trying to get them to mine with somewhat even Mh/s. (2) Should I attempt the Blockchain drivers or is there a resolution to get more stable performance with the latest Adrenalin drivers and if so, can you recommend good setting changes to begin OC'ing and undervolting with the goal being no memory errors [previous settings above]? (3) Should I flash the bios on each card back to stock and reflash each one using 1-click timing patch and run patcher?
I believe this is caused by BIOs settings and driver issues. The only real change I've made is going from Blockchain to Adrenalin drivers, which I see many have issues with online. Power from the wall is not the issue, based on the stable performance I had before.
I appreciate your time and thought on this. Happy hashing from the watthashmanguy.
|
