Claymore's Dual Ethereum AMD+NVIDIA GPU Miner v15.0 (Windows/Linux)

[miner49er2107]

I have a problem. I have a 6x rig with Sapphire Pulse RX580 8GB, using Win10 and the mining driver. GPU0 have 30,5 MH/s and the other 5 cards run at 28 MH/s. I using a start.bat, wich executed (with the highest privileges) by the task planer from win10. It works with 780W from the wall. Why does GPU0 run faster than the others?

Could it be that the clocks differ on the other cards, and the 1 GPU hits your config.

I tried to start the start.bat manually via remote desktop and run as an administrator and get all cards at 30,5 MH/s. Clocks are always the same, but the power consumption raise up to 980W from the wall (-cvddc and -mvddc doesnt work).

What can i do?  

it is imperative in Claymore that the more hashrate the more power consumption occur. so I guess this is normal

if it starts automatically at startup:
Clocks are all the same. CPU Core Clock 1130, GPU Memory Clock 2150, cpu and mem voltages are 850mv on all cards (GPU-Z) | 780W from wall
GPU0 29.990 Mh/s, GPU1 27.928 Mh/s, GPU2 27.911 Mh/s, GPU3 27.916 Mh/s, GPU4 27.930 Mh/s, GPU5 27.975 Mh/s

if i start the bat file manually via remote:
clocks are the same: 1130 core, 2150 mem but voltage now differ GPU0 830mv, GPU1 850mv, GPU2 970mv, GPU3 930mv, GPU4 930mv, GPU5 970mv | 900W from wall and cant adjust the voltage via afterburner (i can choose -70mv but GPU-Z show me no change)
GPU0 29.979 Mh/s, GPU1 30.067 Mh/s, GPU2 30.007 Mh/s, GPU3 30.055 Mh/s, GPU4 30.004 Mh/s, GPU5 30.021 Mh/s

I have seen this when using Remote Desktop, It doesn't happen using Team Viewer.

[cas33]

Hey.Mr. Claymore,we miss you. Where are you? please say hello...

[greatauror28]

@claymore Do we already have a fix for the issue where some of the cards hashes normally while the rest gives their OC hashes?

This is giving me problems right now as I have 8 Nitros and 5 of then hash 30.5 while the other three only manages 25.5.

Minspeed argument doesn’t work for me as it gives bsod to my miner and crashes it.

It would’ve been nice to have all cards hashing as expected upon startup.

[Claymore]

@claymore Do we already have a fix for the issue where some of the cards hashes normally while the rest gives their OC hashes?
This is giving me problems right now as I have 8 Nitros and 5 of then hash 30.5 while the other three only manages 25.5.
Minspeed argument doesn’t work for me as it gives bsod to my miner and crashes it.
It would’ve been nice to have all cards hashing as expected upon startup.

This issue is not related to the miner, it is related to ADL library. Can you overclock all cards correctly with AfterBurner?

[duniek]

385.41 best nvidia drivers

new one 388 (from 23 october - a lot of problems)

[voidwalker]

Both asus z270-a + 1 cheap china m.2 adapter and asus z270-p + 2x m.2 adapters work flawlessly with any 8 nvidia gpus.
So I had all the reasons to expect they will work with the VEGA gpus too. The asus z270-a is no go, but I did manage to get 8 vega gpus after 2 days of trials on a z270-p. I think the secret was not updating the motherboard bios/uefi (I used the one that came with it out of the box). On another z270-p I could only get 7 cards to work, the 8th always had problems or not being recognised at all. Very strange.

I can't return any motherboards, but I can return the ones I will get in the future to test.. like the 13 slot pro btc mobo.
Very little info out there regarding 8 VEGA gpus on a rig. Hope this helps someone. Z270-P works.

Hi. So I bought 8 vega 56 to make a rig, thinking it would be no different than a 1080ti rig.
I can only get 5 cards to work on asus z270-a, the rest are not recognised, and sometimes a 6th comes up with problems in device manager only disappear at the next reboot. very strange.

Do you know what motherboard/driver etc can get 6+ cards running on windows ?

On linux I couldn't even boot with more than 3 cards, it would give me some pci express errors.

Also, what miner works with vega on linux with the 17.40 + rocm drivers ? or is there any way to enable old opencl for vega ?

Please advise. thanks.

I personally think that if you're building a rig of 8 or more GPUs, the sensible options are to get mining-specific mobo like BTC Pro, Asrock H110 or Asus Mining Expert. Expecting the two M.2 adapters (or one!) would work smoothly without problems is just a gamble.

Can you still return the Asus Prime mobo?

Is there any way i can set "my def fee" for the people i am building rig for?

For example every 9 minutes there is happens wallet swap for 1 min and then again to the normal wallet?

Why would you want to change dev fee address to your address when you're not the one who coded and developed the miner?

Given that you've build the rigs for your customers, wouldn't it be better if you just sell them at a profit? (which I think is what you're doing unless you're on to becoming a charity and/or a saint)

[KayspaCoin]

driver fixdag running stable?

[asdpof]

driver fixdag running stable?

Mostly. In a rx580 i have, the miner is unable to control the temperature. Other than this, its all good for me.

[TunerDude007]

v.10 sets correct clocks and voltage for wattman with the last two non-blockchain (drivers 2017.10.20 and the one from september), but then it throttles down upon full startup to 40% less GPU clock while keeping specified voltage values. Then upon exit, I can see in wattman the system maintains the correct OC values specified upon miner startup that the miner itself did not use.

All test were done by using all four specs for GPU and mem clocks+voltages due to mixed results in OCing without increased voltages being specified as noted in thread.

I have had no success in running stable the blockchain driver with and without any OC. The miner does start, I do see a north of ETH 5Mhs increases, but then it freezes my system. I've had very low li/lidag/ethi/dcri values with that driver and still the system keeps on crashing. Vega64 under windows obviously.

Thanks for all the great work, miner looks sweet!

I am using the Robin Hood drivers + Claymore 10 + Windows 7 on a mix GPU RIG - Rig consists of 4 cards
MSI ARMOR RX 570 4GB     cclock 1175 mclock 2100 cvddc 900 mvddc 900   30.1Eth 67.0LBR         Dcri 20
GiG Aurora  RX 580 4GB     cclock 1275 mclock 2050 cvddc 950 mvddc 900   29.9Eth 67.5LBR         Dcri 20
Asus Rog     RX 560 4GB    cclock 1175 mclock 1950 cvddc 850 mvddc 900    13.7Eth 34.0LBR         Dcri 20
MSI         GTX 1060 3GB    OC'd in MSI AB   PM 70   Core +125 Mem+ 900    23.5Eth 21.0LBR         Dcri 28

I switched this rig over to the Robinhood drivers from basic drivers 3 weeks ago as I wanted to switch  back to Ethereum from Music coin for a while. Since the switch 3 weeks ago I had to restart once due to the GTX 1060 freezing the system after a few days with mem set to +950.
At the wall on the meter the system is pulling a total of 595 watts dual mining.

Note: When I start up claymore for the first time, I wait till I see the first set of hash rates then I close claymore and rerun Claymore to get the proper hashrates. The very first instance of Claymore both my 570 and 580 will report 19mh Eth but closing and reopening Claymore corrects this. 

Sorry guys I can't read the whole thread but I hope you are all aware of amd new drivers... 12+ Gpu's much more stable now

http://support.amd.com/en-us/kb-articles/Pages/Radeon-Software-Crimson-ReLive-Edition-17.10.2-Release-Notes.aspx

[carlo_0000]

but the driver is slower than blockchain driver?

because i put in on a computer 1x rx570 when blockchain driver won't run, no idea why, driver install correctly but miner close at start, looks like something is missing
i tried manually install driver, and a cleanup driver but something wrong
i have the same problem on my computer with an r9390 blockchain driver won't work


i put 17.10.2 but it s very slow 22.8mh only dual mining, lost 1mh compared to previous driver i had

[Marvell2]

but the driver is slower than blockchain driver?

because i put in on a computer 1x rx570 when blockchain driver won't run, no idea why, driver install correctly but miner close at start, looks like something is missing
i tried manually install driver, and a cleanup driver but something wrong
i have the same problem on my computer with an r9390 blockchain driver won't work


i put 17.10.2 but it s very slow 22.8mh only dual mining, lost 1mh compared to previous driver i had

i heard there is a setting on the new drivers under global where you have
to flip a toggle for compute workloads so that is probbaly why you are seinf a low hash rate

yup just tried it and one hour in the new drivers are good for now at least, only issue is u have to pick compute workloads for each and every card

, one plus note is i can now change the voltage settings now using wattman , when it would not work in AB wierd

[stannicek]

I tried to switch to old version miner v9.8 and it is stable when network problem happened, so it is related only to v10.0 miner and AMD blockchain drivers. I didn't tried other non-blockchain AMD drivers.

@Claymore
Can you please clarify, what is being done with cards, when restoring work after network outage?

Hello,

does someone facing hard reboot (without any info .. just like on/off on PSU) with Claymore v10.0 on windows when mining (in dual mode) and connection to both pools get lost?

Regards

Could be too much draw on the PSU? Really need to provide more detail on your setup though.

It's not PSU i guess. It is 100% reproducible just by pulling out ethernet cable from lan socket. It crashes immediately during restore of the work(when reconnected to pool). Rigs are running days without crash, but when cable pulled off or some problem happened on network, it goes to hard reboot.

[Vann]

but the driver is slower than blockchain driver?

because i put in on a computer 1x rx570 when blockchain driver won't run, no idea why, driver install correctly but miner close at start, looks like something is missing
i tried manually install driver, and a cleanup driver but something wrong
i have the same problem on my computer with an r9390 blockchain driver won't work


i put 17.10.2 but it s very slow 22.8mh only dual mining, lost 1mh compared to previous driver i had

No Dag fix in the 17.10.2 driver. So unless you're mining an ETH clone like UBQ or EXP, stick with the blockchain driver on v1703. Hopefully AMD will update the blockchain drivers for the Creators update.

[manatnatt]

Hi guys,

I'm totally new to this, I was able to get 26.xxx MH/s on modded Bios and OC for MSI RX580 8GB Gaming X on ETH only.

Now I want to do dual mode but once I ran the script, it hangs.

My code is below:

setx GPU_FORCE_64BIT_PTR 0
setx GPU_MAX_HEAP_SIZE 100
setx GPU_USE_SYNC_OBJECTS 1
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_SINGLE_ALLOC_PERCENT 100
EthDcrMiner64.exe -epool us1.ethermine.org:14444 -ewal 0x739b8012f896721EC6b9b9E23C389d48ea628CCC.Riku4515 -epsw x -dpool stratum+tcp://lbry.suprnova.cc:6256 -dwal username.workername -dpsw password

The CMD just hangs and froze.

Any suggestion?

Thanks.

[smparo]

but the driver is slower than blockchain driver?

because i put in on a computer 1x rx570 when blockchain driver won't run, no idea why, driver install correctly but miner close at start, looks like something is missing
i tried manually install driver, and a cleanup driver but something wrong
i have the same problem on my computer with an r9390 blockchain driver won't work


i put 17.10.2 but it s very slow 22.8mh only dual mining, lost 1mh compared to previous driver i had

No Dag fix in the 17.10.2 driver. So unless you're mining an ETH clone like UBQ or EXP, stick with the blockchain driver on v1703. Hopefully AMD will update the blockchain drivers for the Creators update.

You need to change "GPU Workload" on AMD Settings: go to Gaming -> Global Settings and then for each card (you have a Global Graphics for each one) you need to change GPU Workload to "Compute" (before of that operation disable AMD Crossfire). For me work: 31Mh/s before, and 31Mh/s after... I used DDU before install new driver

[Vann]

but the driver is slower than blockchain driver?

because i put in on a computer 1x rx570 when blockchain driver won't run, no idea why, driver install correctly but miner close at start, looks like something is missing
i tried manually install driver, and a cleanup driver but something wrong
i have the same problem on my computer with an r9390 blockchain driver won't work


i put 17.10.2 but it s very slow 22.8mh only dual mining, lost 1mh compared to previous driver i had

No Dag fix in the 17.10.2 driver. So unless you're mining an ETH clone like UBQ or EXP, stick with the blockchain driver on v1703. Hopefully AMD will update the blockchain drivers for the Creators update.

You need to change "GPU Workload" on AMD Settings: go to Gaming -> Global Settings and then for each card (you have a Global Graphics for each one) you need to change GPU Workload to "Compute" (before of that operation disable AMD Crossfire). For me work: 31Mh/s before, and 31Mh/s after... I used DDU before install new driver

BBT did a clean install of the v17.10.2 driver using DDU on the 21 card Asus B250 livestream tonight and he only got 28.5 on RX 570's MH/s with UBQ. When he switched to ETH the hashrate droped to 25.5 MH/s on the AMD cards. I didn't see him try the Compute switch though.

https://youtu.be/miUoqnUqk9M?t=3285

[Victorio]

because i put in on a computer 1x rx570 when blockchain driver won't run, no idea why, driver install correctly but miner close at start, looks like something is missing

AMD Blockchain Workloads driver does not work on Windows 7, it crashes the miner immediately.
Only Windows 10 for AMD Blockchain Workloads driver.

[smparo]

but the driver is slower than blockchain driver?

because i put in on a computer 1x rx570 when blockchain driver won't run, no idea why, driver install correctly but miner close at start, looks like something is missing
i tried manually install driver, and a cleanup driver but something wrong
i have the same problem on my computer with an r9390 blockchain driver won't work


i put 17.10.2 but it s very slow 22.8mh only dual mining, lost 1mh compared to previous driver i had

No Dag fix in the 17.10.2 driver. So unless you're mining an ETH clone like UBQ or EXP, stick with the blockchain driver on v1703. Hopefully AMD will update the blockchain drivers for the Creators update.

You need to change "GPU Workload" on AMD Settings: go to Gaming -> Global Settings and then for each card (you have a Global Graphics for each one) you need to change GPU Workload to "Compute" (before of that operation disable AMD Crossfire). For me work: 31Mh/s before, and 31Mh/s after... I used DDU before install new driver

BBT did a clean install of the v17.10.2 driver using DDU on the 21 card Asus B250 livestream tonight and he only got 28.5 on RX 570's MH/s with UBQ. When he switched to ETH the hashrate droped to 25.5 MH/s on the AMD cards. I didn't see him try the Compute switch though.

https://youtu.be/miUoqnUqk9M?t=3285

I have two rx580, GPU 1150/memory 2175
Before the upgrade i just view this video (go to  2:52 for GPU Workload):
https://www.youtube.com/watch?v=BbCi5QLhwEU

[faanigee]

His implementation does a little better than the ethminer (genoil or otherwise) ones when it comes to speed on the compute side - i.e. the computations. He also benefits from a technique I also use - somewhat poorly implemented on his side, yet likely somewhat effective - that is, executing some operations "for free" because the memory read is being done at the same time. This also increases the speed relative to the other public miners, and the side effect is (of course) the ASIC is idling less of the time... resulting in your raised power consumption. Appears to be touched up output from the AMD compiler ( well, almost certainly, seeing as it left the little love letter "@(#) OpenCL 2.0 AMD-APP (2442.12).  Driver version: 2442.12" as well as another noting it was used with the compile option "-legacy" )

Everything goes outside of my head 😇 any way thanks for your reply..!!!

[Ursul0]

well... apparently out of 8 cards (over 50*rx570 connected at the same time into almost identical environment & platforms) I have one or a neighboring pair of GPUs making issues [gpuz asic quality, nor mem errors, nor display/dummy all are no factors] the best way to stabilize it seems to keep cards at the same core(I do 1180) and preferably same voltage, lowering only the mem. Some volts may be required to add from default 875. [Maybe will switch to proper setups via overdrive tool... watttool was better I think..., still using it for some equihash workers with 16.11.4 drivers - that have no 100% fan issues ]

Looks like AMD api in blockchain drivers and claymore's use of it makes really weird tricks. changing the voltage may require couple of reboots to "sink in" for instance.