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

[Trimegistus]

 I did buy the powered variety, unpowered ribbon risers are known to catch fire.

I had two rigs R9 270x + 7850 running 24/7 for 6 months with unpowered ribbon risers. None caught fire 

[joaocha]

Guys,

I've seen someone claiming this riser can be used not only from 1x to 16x but also from 16x to 16x...

https://www.amazon.de/Extender-Board-Adapter-Riser-Karte-Blau/dp/B01J3CH7IS/ref=pd_sim_sbs_23_3?ie=UTF8&psc=1&refRID=D0W571TA6JRTMRJYWZMS

Is this true? Anyone doing this?

Or shall I purchase a 16x to 16x powered riser cable?

http://www.ebay.ie/itm/PCI-E-Flexible-Ribbon-HQ-Powered-Riser-card-for-Gaming-Bitcoin-Litecoin-Mining-/201649471966?var=&hash=item2ef33ec1de:m:mWpYtJVgtLoDKx03a-GwBGw

You can use the PCIe 1x USB 3.0 powered riser in a 16x slot.  The little 1x card works because of the little gap in the slot that allows it to sit properly in a 16x slot.   I've had to wiggle it a tiny bit to make sure it was seated properly but other than that they work well. I am using those risers in 4 different 16x slots. They work with no issues.

The ribbon cables are different. When I realized I could use the PCIe 1x riser cards I quit using ribbon cables entirely.


Full disclosure: I sell these risers on ebay. (Version 6, newest riser)

16X-16X RIBBON RISERS DON'T BEND WELL--

I bought a few, and could not use them because the would not bend as easily as a single USB 3.0 cable.  They would not fit my rig.  If there is any advantage to using them, I'll never know.  I did buy the powered variety, unpowered ribbon risers are known to catch fire.

QUESTION:  What is the difference, or improvement, in Version 6 of the wide USB 3.0 riser card?  The internet pics all look the same.         --scryptr

check this
http://i.ebayimg.com/images/g/OukAAOSwH3NXnBRU/s-l1600.jpg

[scryptr]

 I did buy the powered variety, unpowered ribbon risers are known to catch fire.

I had two rigs R9 270x + 7850 running 24/7 for 6 months with unpowered ribbon risers. None caught fire 

LOW POWER GPUS--

I imagine that a 750ti card would not be a problem, either, as long as the motherboard could handle the power draw.  With a 280X-, or a 390X- multi-card rig a problem would occur.  Unpowered ribbon risers will catch fire if stressed.

Your risers did not.       --scryptr

[xeridea]

Thanks.  Looks like both the ASUS and MSI are Micron Memory.  =(

I am not sure if vendors use all Samsung or all Micron.  With my Sapphires so far, the 4GB cards are Micron, the 8GB are Samsung.  I would check GPU-Z, unless you are on Linux?

[fittsy]

16X-16X RIBBON RISERS DON'T BEND WELL--

I bought a few, and could not use them because the would not bend as easily as a single USB 3.0 cable.  They would not fit my rig.  If there is any advantage to using them, I'll never know.  I did buy the powered variety, unpowered ribbon risers are known to catch fire.

QUESTION:  What is the difference, or improvement, in Version 6 of the wide USB 3.0 riser card?  The internet pics all look the same.         --scryptr

It looks like there is another new version since my latest order a month ago! yikes!

The newest as that picture @joaocha posted, has the voltage regulator tube and new capacitors. (Blue vs red/pink writing on the round capacitors.)  I just checked with my vendor and they highlight the new capacitors as well as the voltage regulator tube.

See this photo: http://imgur.com/hjWykTL

Version numbers vary by vendor.  My vendor was up to 6 last time I ordered.   Others are are 1.x but they're basically the same if you look at the board design and components. Subtle things like changing out the capacitor are hard to tell though!

We'll probably be putting in a new order soon as I've sold half my supply. I'll let you know when they arrive what "version" they're on by then. They'll probably have another new release by then!

[yun9999]

 I did buy the powered variety, unpowered ribbon risers are known to catch fire.

I had two rigs R9 270x + 7850 running 24/7 for 6 months with unpowered ribbon risers. None caught fire 

Those dont pull that much wattage.  Try 290 and 390 OC or dual mining, whole different ball game.  They burn out even cheap power risers, don't think you'll be able to do it with unpowered risers when your rig is 3 GPU+

[yun9999]

Guys,

I've seen someone claiming this riser can be used not only from 1x to 16x but also from 16x to 16x...

https://www.amazon.de/Extender-Board-Adapter-Riser-Karte-Blau/dp/B01J3CH7IS/ref=pd_sim_sbs_23_3?ie=UTF8&psc=1&refRID=D0W571TA6JRTMRJYWZMS

Is this true? Anyone doing this?

Or shall I purchase a 16x to 16x powered riser cable?

http://www.ebay.ie/itm/PCI-E-Flexible-Ribbon-HQ-Powered-Riser-card-for-Gaming-Bitcoin-Litecoin-Mining-/201649471966?var=&hash=item2ef33ec1de:m:mWpYtJVgtLoDKx03a-GwBGw

You can use the PCIe 1x USB 3.0 powered riser in a 16x slot.  The little 1x card works because of the little gap in the slot that allows it to sit properly in a 16x slot.   I've had to wiggle it a tiny bit to make sure it was seated properly but other than that they work well. I am using those risers in 4 different 16x slots. They work with no issues.

The ribbon cables are different. When I realized I could use the PCIe 1x riser cards I quit using ribbon cables entirely.


Full disclosure: I sell these risers on ebay. (Version 6, newest riser)

HEhe, I actually buy quite a few risers from this guy on Ebay.  I would highly recommend using him over the Amazon ones.  Cheaper and good quality.  Avoid the skinny riser that doesn't use Molex by all means.    I actually have 20 on order with you right now under heaven_knight.  Anyway you can make the shipping a bit faster.  Got tons of RX470 waiting on risers.   =)

[yun9999]

Anyone running 5 or 6 GPU NVIDIA 1070 rigs?  I can get it to hash at proper speed now but they will start crashing if I run more than 3 per Rig. 

The 1070 needs about 150W each from the wall. So if you have 6 cards, you need at least 1100W power supply. 

I have a EVGA G2 1300w, it's not power.  I also test everything with my Killawatt on the 120V before moving it to my 240V garage setup.  I have many AMD rigs just first time dealing with EVGA and seems like there's so many caveats.

Now I get the following errors and it would crash
GPU0: Gpuminer cu_kd failed 30 (28), unknown error
NVML : cannot get current temperature , error 999
NVML : cannot get fan speed, error 999


Also what epoch level should I set for these GPU?  There was another error I had previous with the DAG file size where they suggest looking at readme.file

Any NVIDIA expert /veteran who can chimed in on this?  I can get 1 card running stable but in a multi 3+ GPU rig, I would get those fan errors, or other things that would crash the OS.  The equipment is currently running in my office at 72F so it's not heat related or anything.

[peoples]

About ADL in GPU-PRO drivers, just found a beautiful answer:

https://bugs.freedesktop.org/show_bug.cgi?id=96786

...
ADL is not supported in the amdgpu-pro driver.  The same functionality can be accessed via standard Linux interfaces (e.g., hwmon for temperature and fan control, sysfs for clock control, etc..
...

So AMD decided not to worry about ADL support in new drivers for Linux. Excellent solution.

Hey, thanks for 6.3 release. Linux version looks smooth so far, no more VM fault errors.
One of your corefeatures in my opinion is the selectiv shutdown of cards in case of overheat, maybe you can implement this also with the new AMDGPU driver ?

You can query the temperature of the cards from: /sys/class/hwmon/hwmonX/temp1_input
To identify the cards that are useing amdgpu driver you can query the /sys/class/hwmon/hwmonX/name class ?

BTW: Overclocking is done by the keys  /sys/class/hwmon/hwmon1/pp_mclk_od (Memory Clock)  and /sys/class/hwmon/hwmon1/pp_sclk_od GPU Clock. +1 changes value by 20.

Just to clarify
For temperature just get it from /sys/class/hwmon/hwmon/hwmonX/temp1_unit, the result must divide by 1000 to get the correct celcius
and we can check whether the hardware is amdgpu or not by checking /sys/class/hwmon/hwmonX/name , If it is amdgpu then it's the right one.

Above file can be access without root.

For overclocking and setup fan speed currently the user has to be root, even sudo is not working. Maybe the user need to be included in specific group first. I leave this to the linux user, since we only need to set it once. Note that when using root user the path is no longer /sys/class/hwmon but change to /sys/class/drm/cardX/device/hwmon

For memory +1 increase 20Mhz, for core clock +1 increase 12MHz. To get the highest one use -1 instead.

Maybe for now just include the temperature so the watchdog can protect the miner in case something happen.

@claymore
what do you think about this ?

@claymore any plans for temperature under linux ?

Third try.

@Claymore
Please update us on this one.

[SwedishBTC]

I am having a problem with one of my rigs it is crashing all the time (sometimes I have to disconnect the power to all GPUs, uninstall the onboard graphics, and then connect the power to all cards again before rebooting the PC to get the PC to boot, and sometimes I even get THREAD_STUCK_IN_DEVICE_DRIVER error message when trying to boot the PC), and I did find out that if I turn off GPU 5 it seems to work better and not crashing.
I am running in Windows 10 x64 with Sappire Radeon RX 480, and I did try different drivers without any luck. And I got the same clock settings on all cards (I am running 4 more RX480 on another machine and that one did not crash one single time in over 3 weeks) This is how the logfile looks right before the crash if I run with GPU 5 on...


02:27:22:161   11f0   ETH: GPU0 27.335 Mh/s, GPU1 27.361 Mh/s, GPU2 27.348 Mh/s, GPU3 27.288 Mh/s, GPU4 27.351 Mh/s, GPU5 0.000 Mh/s
02:27:22:166   11f0    DCR - Total Speed: 2733.663 Mh/s, Total Shares: 319, Rejected: 4
02:27:22:171   11f0    DCR: GPU0 546.709 Mh/s, GPU1 547.214 Mh/s, GPU2 546.966 Mh/s, GPU3 545.762 Mh/s, GPU4 547.012 Mh/s, GPU5 0.000 Mh/s
02:27:22:446   ca0   DCR: weak share, skip
02:27:23:291   6ac   DCR: weak share, skip
02:27:23:761   408   GPU0 t=59C fan=44%, GPU1 t=84C fan=61%, GPU2 t=79C fan=53%, GPU3 t=79C fan=58%, GPU4 t=85C fan=61%, GPU5 t=85C fan=60%
02:27:23:766   408   em hbt: 0, dm hbt: 0, fm hbt: 0,
02:27:23:766   408   watchdog - thread 0, hb time 203
02:27:23:771   408   watchdog - thread 1, hb time 94
02:27:23:771   408   watchdog - thread 2, hb time 0
02:27:23:771   408   watchdog - thread 3, hb time 125
02:27:23:776   408   watchdog - thread 4, hb time 31
02:27:23:776   408   watchdog - thread 5, hb time 141
02:27:23:781   408   watchdog - thread 6, hb time 219
02:27:23:781   408   watchdog - thread 7, hb time 110
02:27:23:786   408   watchdog - thread 8, hb time 94
02:27:23:786   408   watchdog - thread 9, hb time 203
02:27:23:791   408   watchdog - thread 10, hb time 77656
02:27:23:791   408   WATCHDOG: GPU 5 hangs in OpenCL call, exit
02:27:23:796   408   watchdog - thread 11, hb time 77547
02:27:23:801   408   WATCHDOG: GPU 5 hangs in OpenCL call, exit
02:27:24:806   408   Restarting OK, exit...





This is how my *.bat file looks

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 eth-eu.dwarfpool.com:8008 -ewal 0xMyEtherWalletAddress -epsw x -eworker miner2 -dpool stratum+tcp://dcr.coinmine.pl:2222 -dwal MyUsername.miner2 -dpsw MyPassword -dcri 40



What can be wrong, any advice?

Try not to push your GPU5 too hard... you must be running at 2250mhz

solutions for you:

1. Find out which one is GPU5
2. back to stock setting and slowly work your way up (2100 or 2125 mem is a good start.. )
2. Try ramp up the Fan speed (I saw your temp very high 85C)
3. Continue monitoring

ok, I will try to reset that card, any advice what voltage settings I should have?
If this not solve it I will change raiser and see if it works better with another.

I am running at 2200mhz memory @ 940mV, and 1100 core clock @ 950mV. I am running the same setting for all cards. I have fan speed max set to 3200 (I dont really understand WattMan's fan settings??), can I make the fan always go on 70% speed? Because the noice is not a problem.

I did try run without even clocking the cards, I still get the same error message on GPU 5 after some time mining... Is this card "broken"?

[Claymore]

Third try.
@Claymore
Please update us on this one.

I noticed that I miss some messages sometimes, sorry.
I'm not linux dev and reading/parsing some files to get temps looks strange for me and will take some time to implement and test it properly. Therefore I will not do it right now, though I will do it in future anyway because it seems there is no "normal" way now, again thanks to AMD devs, they don't care about such things as backward compatibility. Also, this way allows monitoring only, fans management won't work.
Right now I add new coin and I think it's more interesting feature than getting temps for Linux for 4xx cards.

[kenhai]

Thanks.  Looks like both the ASUS and MSI are Micron Memory.  =(

I am not sure if vendors use all Samsung or all Micron.  With my Sapphires so far, the 4GB cards are Micron, the 8GB are Samsung.  I would check GPU-Z, unless you are on Linux?

yeah, I actually have 2 Gigabyte G1 gaming gtx 1070s and 1 is Samsung and 1 is Micron.

[peoples]

Third try.
@Claymore
Please update us on this one.

I noticed that I miss some messages sometimes, sorry.
I'm not linux dev and reading/parsing some files to get temps looks strange for me and will take some time to implement and test it properly. Therefore I will not do it right now, though I will do it in future anyway because it seems there is no "normal" way now, again thanks to AMD devs, they don't care about such things as backward compatibility. Also, this way allows monitoring only, fans management won't work.
Right now I add new coin and I think it's more interesting feature than getting temps for Linux for 4xx cards.

No Problem, thats why i try it more often
I agree that this seems to be strange for windows users, but in linux this is quite normal.
The /sysfs/ folder mentioned above is a virtual file system from the Kernel (and there drivers) itself.
The above paths mentioned are not files, they are more like APIs where you can control the driver.
For example you can also control the FAN Speed.

Code:

echo 255 > /sys/class/drm/card1/device/hwmon/hwmon0/pwm1

will set the fanspeed of card1 to 100%

This realy is an important feature of your miner, and im pretty sure that bigger mining farms are not running windows

[Claymore]

Third try.
@Claymore
Please update us on this one.

I noticed that I miss some messages sometimes, sorry.
I'm not linux dev and reading/parsing some files to get temps looks strange for me and will take some time to implement and test it properly. Therefore I will not do it right now, though I will do it in future anyway because it seems there is no "normal" way now, again thanks to AMD devs, they don't care about such things as backward compatibility. Also, this way allows monitoring only, fans management won't work.
Right now I add new coin and I think it's more interesting feature than getting temps for Linux for 4xx cards.

No Problem, thats why i try it more often
I agree that this seems to be strange for windows users, but in linux this is quite normal.
The /sysfs/ folder mentioned above is a virtual file system from the Kernel (and there drivers) itself.
The above paths mentioned are not files, they are more like APIs where you can control the driver.
For example you can also control the FAN Speed.

Code:

echo 255 > /sys/class/drm/card1/device/hwmon/hwmon0/pwm1

will set the fanspeed of card1 to 100%
This realy is an important feature of your miner, and im pretty sure that bigger mining farms are not running windows

Ok, I will check it soon.

[z0n0]

Third try.
@Claymore
Please update us on this one.

I noticed that I miss some messages sometimes, sorry.
I'm not linux dev and reading/parsing some files to get temps looks strange for me and will take some time to implement and test it properly. Therefore I will not do it right now, though I will do it in future anyway because it seems there is no "normal" way now, again thanks to AMD devs, they don't care about such things as backward compatibility. Also, this way allows monitoring only, fans management won't work.
Right now I add new coin and I think it's more interesting feature than getting temps for Linux for 4xx cards.

No Problem, thats why i try it more often
I agree that this seems to be strange for windows users, but in linux this is quite normal.
The /sysfs/ folder mentioned above is a virtual file system from the Kernel (and there drivers) itself.
The above paths mentioned are not files, they are more like APIs where you can control the driver.
For example you can also control the FAN Speed.

Code:

echo 255 > /sys/class/drm/card1/device/hwmon/hwmon0/pwm1

will set the fanspeed of card1 to 100%

This realy is an important feature of your miner, and im pretty sure that bigger mining farms are not running windows

I agree. I also think fan control is a must.

[Naimon]

OK this is driving me nuts.  Claymore v 6.2 or 6.4.  I can't hash more than 3.5 MH Eth on any of my new NVIDIA 1070 ( 2 MSI and 2 ASUS).  Just reinstalled Windows 10 64bit and using 372.54 Anniversary update driver

ASUS Card memory - GPUZ shows is Micron.  Let me power off and check the MSI

I might just be reading it wrong, but I just want to confirm... You updated the nvidia drivers and updated Windows to the anniversary update?

Rookie mistake.  I only used the Windows 10 anniversary update driver.  I thought that was a Nvidia  thing , didn't realize there is actually a Windows 10 anniversary  updates as I normally used to service packs.  This fix the issue.  What is the best OC  level?? I am doing +575 for memory and +150 for clock to get 29.5MH eth n 300 sia.  I see others reporting that the 1070 can get 29 Eth and up to 900 on Sia.

P.S Thanks so much for the help

What's your -dcri at?  I run -dcri 60 on my 1070 rig, which gets me 29.5MH Eth and 550 Sia with similar clocks to yours, power set to 65%.

[Naimon]

That is exactly what I want. To enable only 1 nvidia card. It worked with 6.3 and older versions but doesn't work with 6.4

Just checked, you are correct, it's a bug with indexing. Intermediate build with the fix (cuda :
https://mega.nz/#!vsxgXJ6J!leiBtmET6npRXNICCW9LWhaIYe1H5t__9HRt1zLnAy8

Thanks! It works now. Cuda 7,5 in 6.3 is still faster then cuda8 in 6.4. 32.7 vs 32.4 with gtx 1070@100watt

It depends on Windows and drivers version, in Windows 10 cuda8 is faster for 10xx.

I've got Windows 10 and a 1070, and for me cuda 7.5 is faster than cuda 8.

[calculatorduck]

That is exactly what I want. To enable only 1 nvidia card. It worked with 6.3 and older versions but doesn't work with 6.4

Just checked, you are correct, it's a bug with indexing. Intermediate build with the fix (cuda :
https://mega.nz/#!vsxgXJ6J!leiBtmET6npRXNICCW9LWhaIYe1H5t__9HRt1zLnAy8

Thanks! It works now. Cuda 7,5 in 6.3 is still faster then cuda8 in 6.4. 32.7 vs 32.4 with gtx 1070@100watt

It depends on Windows and drivers version, in Windows 10 cuda8 is faster for 10xx.

I've got Windows 10 and a 1070, and for me cuda 7.5 is faster than cuda 8.

Same here, 7.5 is faster on my 1070. On Genoil's miner 8 is much faster. Not sure what the difference is.

[yun9999]

OK this is driving me nuts.  Claymore v 6.2 or 6.4.  I can't hash more than 3.5 MH Eth on any of my new NVIDIA 1070 ( 2 MSI and 2 ASUS).  Just reinstalled Windows 10 64bit and using 372.54 Anniversary update driver

ASUS Card memory - GPUZ shows is Micron.  Let me power off and check the MSI

I might just be reading it wrong, but I just want to confirm... You updated the nvidia drivers and updated Windows to the anniversary update?

Rookie mistake.  I only used the Windows 10 anniversary update driver.  I thought that was a Nvidia  thing , didn't realize there is actually a Windows 10 anniversary  updates as I normally used to service packs.  This fix the issue.  What is the best OC  level?? I am doing +575 for memory and +150 for clock to get 29.5MH eth n 300 sia.  I see others reporting that the 1070 can get 29 Eth and up to 900 on Sia.

P.S Thanks so much for the help

What's your -dcri at?  I run -dcri 60 on my 1070 rig, which gets me 29.5MH Eth and 550 Sia with similar clocks to yours, power set to 65%.

Hello Naimon,

My issue is that if I run 1 or 2 GPU, there's no issues but anything more than that would crash Claymore.  I've tried different power risers as well as motherboard, same issues.  For single card, it works great and I can hash very well, multi GPU, it will crash even when doing stock setting and Eth only.

[gribgo]

There were some comments about running dual power supplies on 1 rig. I didn't see a proper explanation and someone said they thought it was the problem with their rig so let me explain.

You can run the second PSU a couple of ways.

Option 1:
Use a dual PSU cable that connects both PSUs to the motherboard. It is designed for dual PSU configurations. The cable runs connects the second PSU to the same ground so if you shutdown the power on the rig it properly shuts down the power on the second PSU.

Option 2:
Ground short plug
The plug goes into the slot where you'd normally connect your PSU to the motherboard. It completes the ground circuit so the PSU will turn on.  However it stays on and never turns off unless you use the switch on the back of the PSU or you shut down the power via surge protector or whatever. My EVGA G2 Gold came with the plug. I didn't have to buy one and I didn't have to try and make my own.


The problem comes in when people don't connect the second PSU properly. They don't have the right cable or the plug. They try and cross the wires to the MOBO or do other weird things to get the second PSU to work. As long as you don't do that you should be fine.

I could find photos if people need a better explanation.

Both PSUs must be the same power ..like 2X850?