Quote from: Ursul0 on December 31, 2016, 08:27:50 PM

Quote from: maxmad_x on December 31, 2016, 08:11:38 PM

Quote from: Ursul0 on December 31, 2016, 07:41:10 PM

Quote from: Prelude on December 31, 2016, 05:45:43 AM

Glad to hear you're working on an update, claymore! Any chance you can add auto detect in ethman so it knows what client we're running? You could have 3 speed ratings available in the rig properties, ETH/ZEC/XMR so the manager keeps up when we switch algorithms.

Quote from: citronick

While non-RX cards was doing great on v9.6, unfortunately very unstable for RX cards.

I've got a few RX 480/470 rigs running perfectly on 9.6, think it might be something to do with your setup. Here's an example of a 6 GPU rig running for almost 3 days straight:

Rig is running W10 4GB RAM 16.11.4 with 4x 480 and 2x 470, all have modded timings.

Excellent cards btw: refs470 with samsung (1625 into 2000 and 850mv all)
It can be very stable (it never restarts or hangs, there were some app crashes and unsuccessful miner restarts that I got fixed upping the voltage to 880mv) 4x470 on asrock btc r2:

EDIT: there's asus rx470 in slot with hdmi dummy

However 6 cards 2*480 and 4*470 (mostly hynix) on asrock latest (and discontinued) anniversary gave me hell. Among other shit I got banned from dwarfpool(didn't check the banned IP yet, so dunno for how long) because of lowdiffculty ... I guess... Btw the same system does ZEC and ETH solid

HYNIX SUCKS BALLS !!
literally I hate all of my RX480 hynix..!!

Just a word of caution.. Cryptonight is more correlated with core clock. So first the highest core clock at which your 1900 mhz memory maxes out. Thats your core clock for 1900Mhz.. Note the wattage (which usually changes with the core much more than memory). So basically now edge core clock +20MHz and find memory that maxes out with that core clock. Then nudge it 10Mhz again and calibrate memory.

Alternate approach is use 2000 memory clock and which cards fail first. Note those and bring those all down with a factor of -25Mhz. So basically if you decide 1950Mhz for all cards.. those bad cards will be at 1925MHz. So thats the thing for you..
Either calibrate from high to low or low to high.. Once you find stable values that don't crash at least for 1Hr.
And then Believe me.. Decrease your memory clocks by 20Mhz more !!! and then leave your cars run overnight. Dont tweak the power settings after that.. ! So you will be at higher power settings for lower clocks..
I am hating hynixes Angry

.. using 20W more than all my reference cards for 1 MH/s lower hashrate.
But with above strategy atleast its stable for a week!!

I 'm aware of the core speed being the key.
Also I did noticed that lowering mem clock to stock values (which works just fine with ZEC, as it's even more computational dependent) on XMR reduces the hashrate, as well as the stock relaxed mem timings. However I didn't have time to lower the mem to it's limit (the coers I run at 1180 to 1040).
Maybe will try sometime, thanks for the tip!
All my 470s make 700H+-@105W , ZEC 205H@125W and dual eth 28@144W

Hey one question.. Do you think dual mining is still any profitable? I think faster ETH mining with cool cards will keep longevity of the cards. And 144W is pretty high. with ETH only it takes 120W so basically additional 25W for max 8 Cents per day per card. Which almost neutralizes the benefits