Topic: [Cast XMR] high speed XMR/CryptoNight miner for RX Vega GPUs (2 KHash/s) - page 80. (Read 206388 times)

Update from Saturday nights mining rig play time.
Applied at 1207 and then a 1300 SOC timings Reg edits. The SOC goes to 1300 on both, according to HWINFO. So far so good,
that worked and I could mine at HBM1190 still without any issues.

Onto the higher frequencies of HBM, Slowly clocked it up and up, did get to 1250, but after 10 min at that speed, it crashed out.
Before it crashed I was pulling +2300H. Keeping the temps at 40C max.

HWINFO isnt reporting the memory errors on this rig - no idea why at this time, also one rig the same version of HWINFO doesn't even show the SOC timings. Good job the other rig does so I know the mod works.

Put both rigs back on the 1200SOC reg file and left it there for now. Vega 1 is pulling average share find speed of 2050 and vega2 is pulling 2100, (Actual Mining figs are 2100/2300) both on stock Volts, both HBM at 1190 now and both with -30% on the core. The only diff is Vega2 has way better watercooling at 40C max so can be at +50% power and Vega1 is at +10% power and is at 50C.  


SOO Not sure if its the SOC or the RAM erroring out. Would need to put one on 1300SOC / 1190 HBM and leave it running for a week to see if its stable.
Also, Anybody pushed more volts via the HBM? I know everybody wants more efficiency, but I'm free electricity and water-cooled. I want Max performance.

Also managed to get two 560s pulling in 440H each so total is now 5k Hashes on Monero, with 2 more Vegas on order

i wrote that dual claymore isnt as profitable as other cryptonight coins   but in 1 month it will be for sure so i want to start adjusting my rig for Claymore dual  thats why i asked if there is some1 who  is currently running 6vegas + or more on Dual Claymore

You do know XMR isnt the only cryptonight algo coin, right? 

Dual mining Eth + decred/silla  3 usd more on each card now than regular xmr , altho If u go on nicehash the profitability is still better on cryptonight  mining , also  other cryptonight coins seems to be doing better than ETHN + decred but in long term mining dual mining might just be more profitable So I wanna ask any1 who has  working 6 vegas on dual mining ?

I would follow circus dad to the letter, using 1100 power tables and his template for Override tool settings.
except for the detail that I do not use Stak anymore, because i do not want to fiddle with intensity, affinity and such..., Cast does not need all that.
It seems to me that Cast reach 1950/1980Hs per Vega effortlessly, where Stak does a little less.

I use Nanopools or NiceHash pool (using Cast) and I get a little more Hs average, over long periods of time, than what I get on my screen.

I think you are almost there, try to use Cast 8.0 and see what happens. Use the -fastjobswitch setting and you should get good performance at the pool.

I have red here, that some guys run 6 Vegas per rig without issues. It seems the sweet spot.
I would run 6 vegas too, even 12= 6+6, if only I could find them....  ))

Take care!

Hi There,
I would like to say that if you follow to the letter the settings from CircusDad mining guide, you may just get the whole gig going flawlessly.

All windows settings as the Daddy says, power tables 1100 and settings for Override tool as per his template. It just works.
As the "Commonorx" says, Cast does not have any issues, it is just our settings.

My story has been: since the beginning I followed Circus DadDy to the letter, but overlooked at the PCIe settings and I was getting the crashes and all the possible crap, so I began going back and forward with clock and volt settings, batch files and restarts, getting almost there but without solving the issue completely.
Then I reinstalled windows clean again and I was still getting the same problems until I realized about the last detail of the PCIe settings.
Once I put GEN1 instead of Auto, everything began to work flawlessly.
I get 850 watts at the wall with 4 Vegas + 1 RX580, 8650hs combined for any Cryptonight coin, and 265Mhs for the 5 cards combined on Ethereum, although for Ethereum I have to use different Mhz and voltage settings to stay below 950W at the wall.

I hope this helps.

Sigh..... Nice confession here. All your issues had nothing to do with CAST XMR.

Thanks for helping to drag this thread down to utter fools banter.

Thanks, guys, solved at least some of my problems.

Motherboard config switched to gen 1 helped.
Fans are working good now.

I was able to lower power consumption from 1200 to  850 for 4 card rig. Had to use custom power tables in windows registry, since wattman does not allow you to customize entire curve, just 2 last points on it. So far smoth sailing for entire night. I kinda get a lot of computation errors - around 5%. Maybe the memory is set up to high.

Wow. Live and learn.  Now I have a combination of 1) the "85%" soft power play tables from "Phase 2" here: http://vega.miningguides.com/p/faq.html and 2) intensity of 1908 (1800 was way too slow, and 1932 killed it) ... like this:
"gpu_threads_conf" : [
    { "index" : 0, "intensity" : 1908, "worksize" : 8, "affine_to_cpu" : false, "strided_index" : true },

This gets me back to right around 1950 H/s per vega.

What a journey.  @Alexander_Dumas - I also tried modifying the BIOS PCIE settings from AUTO to GEN1 -- not sure if that helped -- I still wasn't able to do 1932 intensity, but I was able to up to 1908... honestly, I don't feel like testing.  Part of me is afraid to breathe around this thing -- it might start puking again.

Next steps:
Do I dare go back to the "100%" soft power play table?
Are there other settings I should try fiddling with? I've been looking at the instructions here -- they appear to be pretty thorough: http://vega.miningguides.com/p/tuning.html.  ** I wish I'd started here a couple of days ago!! ** @Cyan_Antares -- I am itching to fiddle with the various frequencies ... :-)

I think that since I'm drawing slightly less Wattage than before, and my fans are a little quieter, I'll leave things as they are.

Now the question is:  Should I try upping my GPU count?  *IN THEORY*, I should be able to run 5 or 6 vegas in this rig, according to some comments in this thread... 

But is success/stability with 5 or 6 vegas common?  Is the "sweet spot" for GPU count on vega64 rigs usually around 4?

Thanks again for the suggestions.

In addition to what AlexanderDumas said..., just check the frequency at which your Vegas are running. Vegas are very sensible to MHz. Cast XMR was getting random hangs (sometimes it could work for hours, other times it would hang just 2 seconds after it was launched) on a 6 Vega 56 rig I built for someone else, that made me create all kind of crazy batch files to avoid them until I realised that the issue was that I had set the frequency of the Vega 56 cards to 950MHz. I reduced it to 940MHz and so on but eventually find out that the sweet spot is 935MHz, so I have never ever had a problem afterwards, they were mining for days without touching the rig and could sleep well.

@AlexanderDumas - thanks for the suggestion - I'll try it out immediately.

One more note: I tried changing my intensity in xmr-stak to 1800 for all threads.  That "works" - the issue now is that I get 1400 H/s for each vega64.  Would I rather have that than the 0 I had before?  Yeah.  But I'm going to keep plugging until I get back to 1900+ / GPU.

In any case, I'll still try the BIOS change, as I also try to up the intensity/thread...

I have the same mobo and I resolved all my gremlins, changing PCI slot settings on the Mobo Bios.
If you did not do it already, that may be the problem.
Normally the factory setting of the PCI slots is on AUTO, I switched it to GEN 1, and since 2 weeks I did not had 1 crash or any issue.
Before that I was able to run sometimes smoothly for hours or sometimes only for few minutes and then was getting sudden freezes or crashes similar to what you are getting.
Changing from AUTO to GEN 1 changed the whole gig.
It takes only a few minutes, it is worth a try.

ETH is hardly ever more profitable of the Cryptonight coins on a Vega. Unless it goes over 2000usd.
If you mine ETN or Sumo, they are more profitable than ETH, or even better on Nicehash with Cast miner.
On Nicehash is almost twice as profitable than ETH.

I needs people advice. I have a 1070 in my pc and wanted a vega 56 or 64 because I heard they have the power of 2 or 3 cards from the previous generation.

now can I use this just 2 cards in my system and will I run into problems?

also, how much more powerful is the 64 vs the 56 is it alot or can you barley notice?

Hi Everyone,

Sorry for the long post - I'm going to try to be thorough...  What happened?  I set up my first miner, with 4x Vega 64's.  Then I blew it.  I got greedy, and tried to add 3 more.  

Initially, I was able to get the 4x running fine with the cast-xmr - version 060 and xmr-stak-amd-win64.  I had some troubles with hash drops on one card with xmr-stak, so stuck with cast-xmr and ran for weeks solid with no problems, using the standard 'hellae' install guidance from here: https://www.reddit.com/r/MoneroMining/comments/74hjqn/monero_and_vega_the_definitive_guide/.  

So when I tried adding the three new cards (with extra PSU and splitter), I decided to use the more recent guidance from here: http://vega.miningguides.com/.

So far, so good.  I set up everything from scratch following that guidance (with the exception of wiping the HD and reinstalling Windows).

I got all 7 vegas running!!  Good, high hash rates (1980-2000 each), until... cast-xmr hung, after about four hours.  And since then, it only work for about 30 seconds before hanging.

I tried everything I could think of - following all steps in the mining guide again, starting with uninstalling all drivers.  I tried the same hellae guidance from scratch (using Wattman, etc.).

I tried the older (e.g. alternating 2016 & 1800 intensity) and newer (1932 [edited] intensity) settings with versions 060 and 081.  I tried new and old versions of xmr-stak-amd.

With every combination I try, I get this behavior:
1) Everything runs fine, spins up, fans at ~4000, temps reporting at 35-40C.
2) About 30-60 seconds in, I get a driver reset.
 *** When I check event viewer, I see this corresponding with the screen flashing and what comes next: "Display driver amdkmdap stopped responding and has successfully recovered."
3) Then one or more cards starts showing zero hashes, while the rest keep running fine
4) Then cast-xmr hangs (and xmr-hangs at the same point) - no more hashes, no more responses

I've used the driver uninstaller called out in the instructions.  I've even done the full Windows reinstall and started everything from scratch.  In the latest iteration, I never installed Wattman etc. - just followed the MiningGuide instruction and used the .bat/.ps script to reset GPUs from device manager, then used OverDriveNT (and the .bat instructions) to set fan speeds, did the specified regkey updates, etc.

And I get the same behavior no matter what I try - run for 30-60 seconds, driver failure, then a hang shortly after.

I did recently update the bios on my Asrock H110 BTC mobo, I have the latest Windows 10 pro version (16299.192), I'm running 8 GB of RAM, I have more than enough power to the mobo & cards (1000W PSU, never drawing more than 800W) - it's just that new driver failure that I can't seem to shake.

Since I saw earlier in the thread (before I foolishly decided to try a 7-gpu setup) that more than 4 (or 5 or 6, depending on who's talking) is folly, I tried scaling back to 4.  My total reset (with OS wipe) was with only 4 GPUs.  The setup that worked before no longer works...

Any suggestions/ideas are welcome.  I know that people have seen hangs before, and I saw several answers that suggested tweaking the power settings.  The part that's bugging me is that the same cards worked before with the "stock" hellae settings.  And now they don't... argh.

Any suggestions are welcome!  I've burned most of my weekend trying to reinstall everything multiple times... no dice.  Any ideas on what else I could try?

Use trixx software and program an custom fan setting works great for me only for fan control

Someone can give us his OC for RX580 Nitro ? Elpida and Hynix to get 900H/s ?
I get 820 max....
With this :

Core 1150/850
VRAM 2150/850

Blockchain driver
Win 10
Bios original one click patch with PBE

switching to Eth ? it's more profitable right now .  !

hi guys, trying my first Radeon build. Got 8 Nitro+ Sapphire Vega 64. Right now I am testing it out with 4 plugged in.

Installed a blockchain driver.

Only in Radeon panel
Renabled HBCC Memory on each card
, did not use Afterburner or any other tool

Used settings :

Frequency% -30%
Frequency MHz - 1050 MAX
Manual fan speed - min 1300 max 3300
Power limit -20%

I am testing out on ETN mining. Got 1930-1940 H/s on each.

The problems I have right now.

Somehow the fan speed info in every tool I have ( Radeon panel and HWiNFO ) are showing wrong numbers.

When I've installed clean windows and Vegas were not recognized all fans were spinning. As soon as I have installed blockchain drivers the fans stopped, did not move at all at no load.

I guess that's what NITRO+ Sapphire version does - from what I've gathered on the net. They are set up so the fans kick in at certain temp level. They do have huge heatsink fans so it helps.

With no moving fans, I still got the info that they spin at 688 in Radeon settings panel. Wierd.

After I've manually changed speed control on the fans in the settings - bumping the low value to 1300 RPM they started to spin.

3 cards are spinning the fans all the time. Don't see any change in the speed from the sound and look, and the info on RPM is constant around 1600-1700.

One card for some reason has problems with fans - from time to time, every 15-20sec or so they go to complete stop, and just after they stop they start to spin up again. What could cause that? They are connected in the same fashion to PSU.

Now the power consumption. I'm a bit worried that they suck that much. I was hoping to go under 1500W for 8 card rig. Can't really say how much each card itself pulls - Can't find a way to show stats like that in Cast XMR miner.

Do you guys know how to show Wattage in the command line ?

So my 4 card right now pulls around 1200W - that is way to much.

Any ideas on how to make it more efficient?
How to better control the fans ?

I second that. This is a must-have if you want to go beyond 1-2 rigs "at home". Also, this will open possibilities to cast-xmr being used in serious installations (those that can't babysit a rig manually).

I don't understand why it was not built into cast at the beginning. It's much easier to do that than implement a kernel (just saing).