well I appreciate the responses!

Both are Windows 10 Pro Creators Update x64.

Rig 1:
2x Strix 1080ti, Corsair AX1200i 1200w plat PSU. total power load is 520w, its not coming anywhere near its limit. Cards undervolted to 0.812v, pull about 215w each with zcash and 202w with lbry. both in mobo. I definitely don't doubt the overvoltage protection trip, its a known thing with a lot of AX1200i's. However, I've pulled far more wattage through with both set at 120% power limit during stress testing, with each consuming 320w simultaneously.

Rig 2:
6x Gigabyte 1080, undervolted to 0.825v, ~135w each. single 8pin, powered risers. no issue running zcash for weeks. Seasonic SSR-1200PD 1200w plat. Total load is ~930w from wall, however kitguru has tests pulling 1350 sustained from it before a shutdown.

The fact that this is happening on two separate circuits with different powersupplies, different cards, different configs (direct in slot on mobo, other rig is on powered risers) lead me to believe its really not a power issue - though I understand why it would be top of the list. Hmm.

The only common things I can think of are the same OS, same nvidia driver version, and CUDA 7.5 / 8.0 from different ccminer builds. I also wonder if the ccminer implementation of the lbry / lyra2rev2 algos contains some power spiking, which might trip over voltage protection on the PSUs.

did a quick search, didn't find much relevant to this:

Suddenly when I run ccminer 2.0 for lbry or lyra2rev2 both of my rigs exhibit a behavior where they will instantly power down. Its not power related, they are on separate circuits, have been running zcash mining 24/7 for several weeks.

Intermittently, usually within 5 seconds to a few minutes there will be instant power off. Using CUDA and gtx 1080's and gtx 1080tis in separate rigs.

There's nothing in event viewer after the fact, and the shutdown is not a blue screen - instant power cut (and subsequently reboots due to BIOS reboot on power loss).

Has anyone else experienced this before? Its crazy, and only ccminer does this. I've tried 1.85 as well, different nvidia driver versions (382.53 newest atm). Super weird.

NOTE: It really isn't power related - I'm nowhere near limits of current / temperature. One of the crashes did take longer than usual, and I was able to get a windows 10 kernel MEMORY.DMP. I'm almost certain its an nvidia driver fault.

I appreciate the insight, things are a bit clearer now.

whats the fastest eth hash most people get on a 470/570? the 480/580 is a bit different, can push the mem faster and has more CUs. I've seen about 31MH/s for the 480/580 around the top. I've got my 570's all dual mining ETH+DCR at 29MH / 667MH @ 125w stable. At this point I think I can only squeeze a few more watts by keeping the power limit and undervolt ratio the same, but the lower I go with the undervolt the more stability problems. I'll stress test 0.8250v overnight.

Hmm, I've been playing around with version 9.4.

Should my GPU (RX570 4GB) have poor ETH hashrate (19-22) and get better as I undervolt? That seems backwards. Normally you have to jack powerlimit or give +vcore to stabilize clockspeed under utilization. Is this expected behavior?

At stock, 1.00v vcore and 1246mhz core, 2000mhz mem @ 1500 straps: 24MH
At undervolted / underclocked, 0.806 vcore, 1100mhz core, 2000mhz mem @ 1500 straps: 29MH.

I've played around with the powerlimit settings, different undervolts at different pstates. I can't quite see the pattern as to why this miner performs better the more I reduce power.

Does anyone have an idea?

Hey doktor,

I just picked up a VisionTek Reference RX570 4GB and it shows up as "unknown bios" with your tool (yet works fine in others). I'd like to modify the voltages with it (essentially hardcode into the bios a VDDC offset). I've looked all over and nobody seems keen on sharing how to do it (or write a tool for one).

would you be able to help me out? I would be happy to tip.

I've uploaded my BIOS here: https://mega.nz/#!jtQFjJAR!dW8Fl_OJHPR2Q99UeatUBvoFM81LHa4Xv7_kaSHHM-4

have you fixed the issue with BitStamp where if you do an "All in" buy order, it doesn't calculate properly and the order cannot be placed unless you subtract a penny (then it goes through fine)?

Offering 40 realBTC for goxBTC at 72.5% rate.

Escrow available. PM

Pay for assembly in July, product does not come to fruition. Get confirmed for assembly refund 7 months later. Several days after refund confirmation and no payout. Feels like BFL in here.

still have not received my refund after accepting it either.

IGHOR, are you adding Huobi? Seems silly to not support the exchange with the majority volume.

http://www.huobi.com/help/index.php?a=api_help

I think I mis-interpreted what was written. I interpreted
as the USD value represented in BTC (basically peanuts.)  

I did find it odd the following sentences after about exchange rate. It seems you're right, I hope you're right. Though I find it odd that he held on to the money originally because money was short, and now he's got excess money to give us such a good (and fair) exchange rate? With that exchange rate, we'll very nearly make back the full USD balance we originally spent (so that 35% refund feels more like 100% given the bitcoin multiplier present-day.). Why not just say 100% USD refund in that case? Why complicate it?

Yeah, got the email as well.

It's sad that we'll only see a partial refund, though it's better than nothing. It's a double wince of pain that we paid in BTC and it was "held" from us instead of refunded to us as bitcoin ramped up to the $1k range. Now it will be given back to us at the ~$130 rate. Welp. Pretty sure when people were refunded for chips you should have refunded their 35% assembly right then and there. Poor form.

FATCA will forever be known as FATCAT, as it serves them well.

good on you man. This was the best approach to stick it up yifu's ass.

Have people thought of the risk of moving BTC through exchanges when they are 1,000,000 or even hundrends of thousands a piece?

What's stopping an exchange from amassing hundreds of millions and just closing down. We have no reprocussions yet. Right now the price is so low (but getting there) that with the volume, they'll be moving sooooo much more money in the future.

boom.

Has the version with the buy/sell fix for btcchina been pushed yet?

Thanks, I'm sure you're aware of the symptoms. Selling BTC is only to the 3rd decimal, the field rounds up and you attempt to sell more than you actually have. I had done that a few times (place -> confirm -> nothing shows up in open orders) before realizing it. Afterwards even placing a standard amount did not work / show up in open orders.

Eventually I logged onto the site manually and my account is locked for security. Be careful until this is fixed.

Yeah, having used the MtGox functionality I assumed it all worked. Tried selling so many times with QT BTCChina without success that it locked my account. No trading / withdrawal / anything for 8 hours until support wakes up and there's a dip coming.


Woo.

Likewise, some funds finally moved around and I pumped some more into bitcoin.