Quote from: scryptr on October 30, 2017, 06:13:33 PM

Quote from: greatauror28 on October 23, 2017, 03:43:22 PM

Quote from: Wolf0 on October 22, 2017, 01:07:09 PM

Quote from: Kumic on October 22, 2017, 09:37:06 AM

Quote from: Call_Me_Bambi on October 21, 2017, 10:43:11 PM

Quote from: greatauror28 on October 21, 2017, 02:42:26 PM

Okay, after 3 months of mining during the last couple weeks rig has been freezing and the last line of the log file indicates:

GPU got incorrect temperature 511 ignore

Thoughts?

I got this error when one of the fans on an RX580 failed. I don't know how long it was running with one failed fan but when I noticed the failed fan the card was very very hot !

In that case, the temperature was so high that the card itself couldn't measure it and that's why you got GPU got incorrect temperature 511.
That's my guess.

No, the card didn't respond - so it returned an code meaning invalid, which is what you see.

Hey Wolf0 do you think this is a failing riser issue? I kinda ruled that out already because it takes days before this happen. I believe if it's a failing riser it would kinda happen right away?

I hate to think this is a GPU issue as they're barely 4 months old.

Quote from: Vann on October 22, 2017, 08:58:07 PM

Make sure Windows didn't update the drivers automatically. On Windows 10 Pro you can set Windows Update to not include drivers. Type 'gpedit' in the search box and in the Group Policy Editor navigate to Administrative Templates ==> Windows Components ==> Windows Update and set 'Do not include drivers with Windows Update' to enabled.

Good man, thanks for sharing this! Appreciated.

IT DEFINITELY COULD BE A RISER--

Risers do not always fail immediately. The riser can overheat, or short within a connection, and fail to work properly. Have you tried the card in a motherboard without a riser? By itself? If you then get the same result, your card is bad. However, it could have failed because of a bad riser, a riser with a short, or with a component that failed due to heat.

Have you swapped a fresh riser into the system? Have you swapped risers in the system to see if the riser produces the same result on another card?

The "511" readout is not good news, but it does not automatically mean the card is dead. And, failing risers can damage a card. Other things to look for is cable friction against a fan, spacing between cards, or placing the card in a hot point within the system (not good). --scryptr

I just disconnected/reconnected the cables, reseat the GPUs onto the riser ports and the problem hasn't resurfaced in more than 8 days now.

Fingers crossed.

Quote from: Jup1t3rR on October 30, 2017, 07:08:06 PM

Quote from: greatauror28 on October 30, 2017, 01:13:59 PM

Quote from: staman on October 30, 2017, 01:08:15 PM

Quote from: Ursul0 on October 30, 2017, 11:58:09 AM

ok gyus some good news:
crimson-relive-17.10.2-oct23 driver is
1) more stable than the blockchain drv (has the DAG fix that can be enabled with "compute" workload)
2) has proper API (no issues for setting up the volts with overdriven tool)
3) supports more then 8 AMD gpus on the same platform

all good news indeed.

Well, i strongly disagree. My personal experience (and plenty of others) is completely opposite to yours

1) how can you even say that out loud? that driver is not even 7 days old

2) OverdriveNTool has been working just fine with blockchain drv (23aug)
3) yes, it let you post it in device manager and yes, it let you mine with it.. but do try enable compute workload and see that mess if you are running 12 cards! And loosing 4mh's per card with NO-dag-driver * 12 cards.. hell no !

Please stay away from such a comments, you are fooling others!

Actually what Ursul0 cited is pretty factual, as I've experienced it myself.
This driver, together with the Win10 FCU fixed my OpenCL issues - and i've tried everything under the sun to fix it. I used to use aug 23 blockchain driver and a recently-installed win10 pro without any updates.

I only run 8 cards but here's some proof that's it's stable.

Previously with the setup above, miner doesn't get past 3 hours without openCL issues.
It has the dag fix just like the blockchain drivers - and this version is fully-supported by AMD and not in Beta status.

Note that you have to enable *compute* settings in each card in the Radeon Settings for this to work.

whats your power usage mate?

i got rx580s 8gb hynix and would love you know what your power usage is

currently im running 6gpu @ 170MH - 790W

I pull ~1250w at the wall with 240Mh eth and 6.1Gh dcr.

I think ~123w per card doing eth-only is pretty high. I dual-mine so ~140w is really what the Nitros consume if properly undervolted.