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

[Claymore]

You might find this interesting or funny Claymore?
Looking at a comparison of Intel chipsets (2500K quad processer) and AMD chipsets (Phenom 555 dual core):
The AMD GPUs find more shares for Ethereum on an AMD chipset and with the dual core processor with 1 core deactivated (save some electricity) and overclocked to 3.5GHZ than the Intel chipset and quad core processor at any cores and any clocks.
The Nvidia GPUs find more shares for Ethereum on an Intel chipset and with a quad core processor with 2 deactivated cores (save some electricity) and the two remaining cores set to 3.3Ghz then on the AMD chipset and processor at any cores and clocks.
It looks like: you need the AMD chipset to maximise the AMD GPU's efficiency and you need the Intel chipset to maximise the efficiency of the Nvidia GPUs for crypto mining of Dagger

It is funny. Shares are related to hashrate and luck, that is all. More total shares in test - less luck involved. To exclude luck it's a good idea to test at least 1GH/s for 24 hours.
Anyway, if you change CPU/chipset/anything and miner shows same hashrate, it means that you have same speed, same earnings in long term.

[thevictimofuktyranny]

You might find this interesting or funny Claymore?
Looking at a comparison of Intel chipsets (2500K quad processer) and AMD chipsets (Phenom 555 dual core):
The AMD GPUs find more shares for Ethereum on an AMD chipset and with the dual core processor with 1 core deactivated (save some electricity) and overclocked to 3.5GHZ than the Intel chipset and quad core processor at any cores and any clocks.
The Nvidia GPUs find more shares for Ethereum on an Intel chipset and with a quad core processor with 2 deactivated cores (save some electricity) and the two remaining cores set to 3.3Ghz then on the AMD chipset and processor at any cores and clocks.
It looks like: you need the AMD chipset to maximise the AMD GPU's efficiency and you need the Intel chipset to maximise the efficiency of the Nvidia GPUs for crypto mining of Dagger

It is funny. Shares are related to hashrate and luck, that is all. More total shares in test - less luck involved. To exclude luck it's a good idea to test at least 1GH/s for 24 hours.
Anyway, if you change CPU/chipset/anything and miner shows same hashrate, it means that you have same speed, same earnings in long term.

Yelp, I was thinking it might be that Nvidia do their primary coding work for their GPUs on intel chipset/CPUs and then do the compatibility crossover coding for AMD stuff afterwards.

Whilst, AMD guys and gals do their primary coding work for their GPUs on their own Chipsets/CPUs and then do the compatibility crossover coding for Intel chipsets/cpus.

Obviously, this for Nvidia Compute and AMD openCL calculations.

I'd guess they make the video gaming drivers perfectly optimised for whatever chipset/cpu combination a person is going too use.

[alrami]

Hello, can someone tell me what the power consumption of gtx 970 in dual mining?

it depends from your settengs and OC level

But better will be to find NVIDIA-SMI in nvidia directory

than run in a comand promt mode - "nvidia-smi -q --display=power"

you will see the curent power usage of your nvidia cards, reported by system (win)

[peoples]

is someone running claymore 6.2 with RX4xx card under linux ?

I have realy poor hashrates, also miner uses lot of CPU power.

Code:

ETH: 08/21/16-22:14:34 - New job from eu1.ethpool.org:3333                                                             
ETH - Total Speed: 32.101 Mh/s, Total Shares: 0, Rejected: 0, Time: 00:01                                              
ETH: GPU0 3.602 Mh/s, GPU1 2.451 Mh/s, GPU2 11.019 Mh/s, GPU3 5.707 Mh/s, GPU4 4.465 Mh/s, GPU5 4.857 Mh/s  

i guess its also related to the tons of

Code:

 VM_CONTEXT1_PROTECTION_FAULT_ADDR  
VM_CONTEXT1_PROTECTION_FAULT_STATUS 

errors every second while mining ?
No problems with genoil miner. So not sure if its driver or miner related ?

Running Ubuntu 16.04 with  amdgpu-pro_16.30.3-315407 driver.

[calculatorduck]

I'm getting about 10-15% slower speeds on my GTX 1070 compared to genoil's 1.1.7 compiled with CUDA 8.

This is on Win10 anniversary update with the latest nvidia drivers.

[dephcon]

Claymore,

If you're unable to underclock with a specific clockrate using cclock/mclock is it possible to accomplish it using the dumb percentage based clock adjustment as per wattman?

[mineromineroso]

v6.2:

- released version for Linux with nvidia support.
- fixed Ethereum solo mining mode.
- added "-cvddc" and "-mvddc" options for adjusting voltages for latest AMD 4xx cards.
- now you can turn on/off cards in runtime with "0"..."9" keys.
- bug fixes.

Thank you for the voltage new feature! Its really useful for the memory. But if we use only one voltage value for the gpu, we lose the option to undervolt more if the gpu throttles. For example, if the cpu throttles between 1075mhz and 1145mhz we can go also from 860mV to 900mV. But using only one voltage value, this value needs to be the biggest value for the higher frequency.

This doesn't matter if you use a fixed gpu frequency, but when using all the gpu power states may be will be interesting to add a voltage table according to that gpu states.

Anyway, seems that from state 3 to 7 the voltage scalability is lineal, so, maybe its easyer to implement a method to add/subtract voltage to all states.

[TheLAWNoob]

v6.2:

- released version for Linux with nvidia support.
- fixed Ethereum solo mining mode.
- added "-cvddc" and "-mvddc" options for adjusting voltages for latest AMD 4xx cards.
- now you can turn on/off cards in runtime with "0"..."9" keys.
- bug fixes.

Thank you for the voltage new feature! Its really useful for the memory. But if we use only one voltage value for the gpu, we lose the option to undervolt more if the gpu throttles. For example, if the cpu throttles between 1075mhz and 1145mhz we can go also from 860mV to 900mV. But using only one voltage value, this value needs to be the biggest value for the higher frequency.

This doesn't matter if you use a fixed gpu frequency, but when using all the gpu power states may be will be interesting to add a voltage table according to that gpu states.

Anyway, seems that from state 3 to 7 the voltage scalability is lineal, so, maybe its easyer to implement a method to add/subtract voltage to all states.

At this point you might as well edit the BIOS and flash all your GPUs.

[gary_aut]

v6.2:

- released version for Linux with nvidia support.
- fixed Ethereum solo mining mode.
- added "-cvddc" and "-mvddc" options for adjusting voltages for latest AMD 4xx cards.
- now you can turn on/off cards in runtime with "0"..."9" keys.
- bug fixes.

Thank you for the voltage new feature! Its really useful for the memory. But if we use only one voltage value for the gpu, we lose the option to undervolt more if the gpu throttles. For example, if the cpu throttles between 1075mhz and 1145mhz we can go also from 860mV to 900mV. But using only one voltage value, this value needs to be the biggest value for the higher frequency.

This doesn't matter if you use a fixed gpu frequency, but when using all the gpu power states may be will be interesting to add a voltage table according to that gpu states.

Anyway, seems that from state 3 to 7 the voltage scalability is lineal, so, maybe its easyer to implement a method to add/subtract voltage to all states.

And how will you test gpu running stable for hours in all different states ? When Claymore add such Options the Miner will be more complicated to use and bound to current AMD-Drivers (Wattman - 7 States). I use a fixed GPU Frequency with Powerlimit + to disable Throtteling and then I will undervolt to get less Powerdrain. The Powerlimit Option is only a Limiter.

cg

Gary

[bgminersz]

I'm having problems with temperature control under ubuntu 14.04.5 and claymore 6.2 beta 2. The  miner is working but no fan and temp are displayed, nor -tt command is working.
Please let me know what to do in order to fix this problem.

[adamvp]

I suppose a big error... I have noticed if pool goes offline (like nanopool now) miner hangs with my whole rig...

Besides I noticed yesterday that 6.2 version shows only 32celsius on one card, restart helps

[TheLAWNoob]

I suppose a big error... I have noticed if pool goes offline (like nanopool now) miner hangs with my whole rig...

Besides I noticed yesterday that 6.2 version shows only 32celsius on one card, restart helps

I can't go on nanopool's website, but mining is fine...

[adamvp]

There was short break in mining 2hours ago:http://etc.nanopool.org/account/0xbc2e6739781028ebcac5dd81c60f85470d26c621
In sia too.
I have two rigs: linux and CDM 5.3 which resumed after that break
And Windows with CDM 6.2 which hanged with whole rig...

Additionaly : now I switched to eth  only Mode on epool.  Works fine but I have no worker on stats page:
I have such line  in epool. Txt

Code:

       POOL: uk1.epool.io:8008, WALLET:  0xbc2e6739781028ebcac5dd81c60f85470d26c621, PSW: x, WORKER: rusMineruk, ALLPOOLS: 1
POOL: uk2.epool.

:, :  , : x,

[roitpls]

installed the latest driver and latest APP SDK.

And another problem.
The syslog errors are continuously

Code:

[ 7669.222221] amdgpu 0000:06:00.0: GPU fault detected: 147 0x0d608401
[ 7669.222222] amdgpu 0000:06:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x081047AC
[ 7669.222223] amdgpu 0000:06:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x0E084001
[ 7669.222224] VM fault (0x01, vmid 7) at page 135284652, read from 'TC7' (0x54433700) (132)

Linux is a problem, always.
Well, it seems AMD does not support ADL v10 in Linux yet. In Windows it works fine. I will build a system with Ubuntu with latest drivers and RX480 to confirm it.
"VM_CONTEXT1_PROTECTION_FAULT_ADDR" error - not sure that it is a problem of miner, it can be issue in drivers. What Linux version? What drivers version? I tried to reproduce it but it seems I'm doing something wrong.

Please keep supporting Linux version  , the driver is still in beta so I'll patiently wait
There is no libatiadlxx.so when using amdgpu pro driver, so I guess they release ADLv10 but not merge it with the driver yet.

I myself use Ubuntu 16.04.1, driver version amdgpu-pro_16.30.3-315407.tar.xz ( this is last month version, I think before ADLv10 release ), try on both stock kernel 4.4 and 4.7 ( not the latest 4.7 kernel ), maybe 4.8 will be better

Then check on /var/log/syslog or journalctl when running claymore an you'll see above error.
The mining works ok anyway, but the syslog is bad "too many write" We can disable syslog to protect hdd.

It looks like a driver problem, but perhaps you can find more

P.S It's easier to find whether your GPU is faulty or not on Linux then Windows ( Windows just directly go to BSOD  or recovery with no log ), I personally think Windows is overkill for mining

I have this exact same issue, causes CPU to be tagged to 100% from all the error logging.

[peoples]

same Problem here.
Also tried  Kernel 4.8 RC2 , same issue.

Do you guys mine at normal speed when seeing this issue ?
With Claymore's miner have realy bad mining results (arround 5 MH/s)

[x201k120]

Thanks for the Linux support!

I'm on Ubuntu 16.04 and amdgpu-pro_16.30.3-315407 with a RX 480 ref
No ADL installed - so its fine that I get an error for that when I start the miner.

AMD Cards available: 1
GPU #0: Ellesmere, 4068 MB available, 14 compute units
POOL/SOLO version
GPU #0: set -etha as 0 (ETH algo for fast cards)
No NVIDIA CUDA GPUs detected.
Total cards: 1
AMD ADL library not found, temperature management for AMD GPUs disabled.
ETH: Stratum - connecting to 'eu1.ethermine.org' <108.61.165.134> port 4444
ETH: Stratum - Connected (eu1.ethermine.org:4444)
ETH: Authorized
ETHEREUM-ONLY MINING MODE ENABLED (-mode 1)
ETH: qtminer stratum mode
Watchdog enabled
Remote management is enabled on port 3333

Setting DAG epoch #70...
Setting DAG epoch #70 for GPU #0
Create GPU buffer for GPU #0
ETH: 08/22/16-11:11:16 - New job from eu1.ethermine.org:4444
ETH - Total Speed: 0.000 Mh/s, Total Shares: 0, Rejected: 0, Time: 00:00
ETH: GPU0 0.000 Mh/s
GPU 0 DAG creation time 3072 ms
Setting DAG epoch #70 for GPU #0 done
ETH: 08/22/16-11:11:34 - New job from eu1.ethermine.org:4444
ETH - Total Speed: 6.100 Mh/s, Total Shares: 0, Rejected: 0, Time: 00:00
ETH: GPU0 6.100 Mh/s
ETH: 08/22/16-11:11:55 - New job from eu1.ethermine.org:4444
ETH - Total Speed: 5.820 Mh/s, Total Shares: 0, Rejected: 0, Time: 00:00
ETH: GPU0 5.820 Mh/s
ETH: 08/22/16-11:11:57 - New job from eu1.ethermine.org:4444
ETH - Total Speed: 6.092 Mh/s, Total Shares: 0, Rejected: 0, Time: 00:00
ETH: GPU0 6.092 Mh/s

... and continues with a hashrate like this

tail /var/log/syslog

Aug 22 11:12:07 ethereumrx480 kernel: [325501.707532] amdgpu 0000:02:00.0: GPU fault detected: 147 0x0d508401
Aug 22 11:12:07 ethereumrx480 kernel: [325501.739385] amdgpu 0000:02:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x081047A6
Aug 22 11:12:07 ethereumrx480 kernel: [325501.801310] amdgpu 0000:02:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x08044001
Aug 22 11:12:07 ethereumrx480 kernel: [325501.863327] VM fault (0x01, vmid 4) at page 135284646, read from 'TC5' (0x54433500) (68)

so it seems the same like for the other linux folk over here with RX* cards. Genoil Ethminer is running fine though.

[Claymore]

I'm checking this issue in Linux, so I will have some news in a few hours.

[jkminkov]

I have the impression that claymore handles unstable pool somewhat conservative comparing to genoil+mining proxy(dwarfpool version), the issue is with going back to original pool when connection is recovered, is there any way to change setting so it goes back to primary pool as soon as possible.

[adamvp]

I have the impression that claymore handles unstable pool somewhat conservative comparing to genoil+mining proxy(dwarfpool version), the issue is with going back to original pool when connection is recovered, is there any way to change setting so it goes back to primary pool as soon as possible.

Yes,  here you have:

Code:

-ftime   failover main pool switch time, in minutes, see "Failover" section below. Default value is 30 minutes, set zero if there is no main pool.

[crackfoo]

I can get BTC directly from mining DCR at zpool, ....?

Yes you can. just mine using your BTC address instead of DCR.