Why are transactions so ridiculously expensive? "sweep_all" wants to take 8% (1200 ETNX), whereas ETN was only 0.000008% (0.12 ETN). Did someone get a decimal point wrong?

Where did this come from? It's not mentioned on the front page of this thread or on the G3K website.

Edit: Okay, I see he edited one post of his website, but not the most recent one. Very confusing.
Also confirmed via Twitter.

Yes, you absolutely should be reading the thread to find if other people have had your issue before making a separate, redundant comment. That goes for all forums on all topics. Shnikes even gave you other resources you could use, so I don't know why you would say 'thanks for nothing'.

I'm looking for a working Delta 2400W PSU from a forum-reputable person, preferably (but I suppose not required to be) in the US. I'll cover shipping.

Able to pay with USD, XMR, or ETH (or Shapeshift to whatever you want, if you're willing to take on the conversion fee).

Any idea if you'll ever get more of these?

In my experience, once you have it where you want, close the program and reopen it. Most of the time, it will keep your settings. If the rig crashes or you shut down the computer without explicitly closing HWInfo, it won't retain.

I see you've subsequently fixed this, but if it ever happens again (especially at higher memory speeds):

Step 1: Use HWiINFO64 to keep an eye on sensors. Adjust the window so you can see all of the cards in an organized line like this.
Step 2: When a hang happens, try to see which card(s) have failed (most commonly indicated by low Memory Current or Memory Power (I forget which, but it will be obvious)).
Step 3: Use OverdriveNTool to increase the memory voltage on that card. Most of the reg edits put memory at 900mV. Increase it to 925 or 950 if you have to. OdNT and HWiNFO use the same enumeration, so if "the second card" is exhibiting the behavior, increase memory voltage on the second entry.

This is how I was able to stabilize my 8x Vega rig.

Yes, when reference cards were made, only Samsung had HBM2 available. The aftermarket cards were late in part because Hynix wasn't ready.

In my experience, no; it depends entirely on the memory. Many (all?) aftermarket 56s are using Hynix memory, which doesn't perform anywhere near the Samsung memory present in all reference cards, and aftermarket 64s.

It certainly isn't the only way to go about tweaking, but it is quite versatile and thus hardly 'useless'. Would be interested in knowing what you use to manage a new GPU's settings.

Checking in as another 8x Vega miner, though on the ASRock H110 - 15,760 of that sweet Monero hashrate, every second.


On topic, I'm glad I found this, as I was actually expecting to be able to run 18 RX570s on this thing. Though is the limitation the motherboard or Windows?

(Edit: After seeing other threads, I see the mobo won't even POST with over 13 of the same card... very weird, I definitely don't understand how any motherboard would care which GPU manufacturer is being plugged in, though someone mentioned potential display outputs, which is an interesting thought.)

Version 0.7 added an HTTP listener that returns a JSON-formatted report.
It can be activated with --remoteaccess, and is available on port 7777

No, they are not "forcing you". That's a voluntary choice you make and agree to by choosing their one-click binary download. It's not like they're the only XMR mining software in existence. You are free to support another project that provides binaries with no fees.

If you do the work to modify and compile, you are free to make whatever personal changes you would like - they even walk you through how to do it, which makes people's arguments even more ludicrous.

To the claims of the devs not actually doing any work: you only need to see the commit logs to see the amount of effort that has gone into this software. Hundreds of man-hours so that you can click a button and gain credit on a blockchain ledger.  Why do you not see value in giving them a modest amount of support and incentive to continue to invest into the software?

Would have had to be Samsung; Hynix didn't have theirs available until the aftermarket boards which didn't start production until what, 4 months into Vega lifecycle?

PowerColor got a bad rap, but I think it was just silicon lottery for the ones doing public benchmarks. I have a family member with a reference PC Vega 56 with no issues.

Which drivers and OS are you using? This may be Vega-specific, but I'm fairly certain that 8 GPU was the limit on the August Beta Blockchain driver, as well as Windows 10 before the Fall Creator's Update. 

Also note that the Win10 "Fall Creator's Update" from October(?) did specifically increase the limit, though I don't know what the maximum is now.

The way I do it is to use HWiNFO64 to monitor voltage, and set core voltage to 900mV. If that takes under load (Cast/Stak is stable, and HWiNFO reports core at 900), then I'll try dropping it to 881.

Of course, I've only been successful in about half of my cards (7 of 14), but many of my others are in the low 900s, which will push the draw up to 180W or so which isn't bad; I could probably back that off and hash around 1900 at 150W, which may be more palatable for some.

A response to browndog's question was given in the appropriate thread.

You can always extract the bundle (using 7Zip or WinZip) and simply point Windows to the Packages\Drivers\Display folder when installing a new device, or do a Custom Install and unselect AMD Settings.

As you've already done, use OverdriveNTool for AMD cards.

Try flashing your card with the Vega 64 BIOS and re-applying the registry mod (use the 1100 HBM to start, to see if it's stable).