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

[ocminer]

Claymore Port for Sia added:

sia.suprnova.cc:7777

Claymore string for merged ETC and SIA Mining:

ethdcrminer64.exe -epool etc-eu.suprnova.cc:3333 -ewal suprnova.1 -epsw x -esm 3 -dpool sia.suprnova.cc:7777 -dwal suprnova.1 -dpsw x -dcoin sia -allpools 1


Notice: Still 20% BONUS (=MORE) per BLOCK on Sia !

Hashrate on the Dashboard might be a bit flaky, look at https://sia.suprnova.cc/index.php?page=statistics&action=graphs the Graphs, if your Hashrate is stable there - all is good.

seems to be working fine, anyone got issues ?

[MarkAz]

Hey Claymore, really like your miner, been running it on a bunch of machines really without issues.  Very excited to see the -ttli feature, and have been fiddling around with it, with mixed results.  I can tell you are using some simple logic for controlling fan speed and intensity - really too simple for this application.  While on some of my machines it achieves a decent level of stability, on many it does what's called 'flapping', meaning it will oscillate up and down, as it chases the correct setting.  This is actually a really common problem in industrial automation - and specifically in oven control (not cooking ovens, but ovens used with circuit boards).  The way they achieve stability is by using an algorithm called a 'PID' - you can read about it here:

https://en.wikipedia.org/wiki/PID_controller

It's actually very easy to implement - the only challenge is tuning the PID, but I don't imagine that would be that big a deal, even the auto-tuners out there would do a radically better job then what you've currently got implemented.  Then you'll get a rock solid line in terms of hashing performance.

Anyway, hope that helps - as the idea behind it awesome, just looking forward to it being more consistent...

[bctopics]

Getting 2 mh/s with my 980ti. Suggestions?

This is my start file

Code:

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:4444 -ewal 0DC6337fb811eb32473060Ca620F010E996EC71c -epsw x -dpool stratum+tcp://sia-us-east1.nanopool.org:7777 -dwal e8632cecaa2617cb2acb8b3c14ae34fafbca9fc89cc4eaa13d60a27dd18a7a5b7332cf22ab75/7970/bctopics@gmail.com -dcoin sia -dcri 25 -allpools 1

driver version 372.54

[n0x1ous]

v6.1:

- Fixed issue with high CPU load for NVIDIA cards.
- NVIDIA: now two builds are available: for CUDA 7.5 and for CUDA 6.5.
- Fixed issue with resetting WattMan settings.

NOTE: Still no Linux version yet, I need to have stable Windows version firstly.
NOTE2: I don't see good hashrates in cuda 6.5, at least in Win7. May be some specific drivers version must be used?

How do we make it use cuda 6.5 path?

[rednoW]

Getting 2 mh/s with my 980ti. Suggestions?

driver version 372.54

install older driver for Maxwell

[Claymore]

Hey Claymore, really like your miner, been running it on a bunch of machines really without issues.  Very excited to see the -ttli feature, and have been fiddling around with it, with mixed results.  I can tell you are using some simple logic for controlling fan speed and intensity - really too simple for this application.  While on some of my machines it achieves a decent level of stability, on many it does what's called 'flapping', meaning it will oscillate up and down, as it chases the correct setting.  This is actually a really common problem in industrial automation - and specifically in oven control (not cooking ovens, but ovens used with circuit boards).  The way they achieve stability is by using an algorithm called a 'PID' - you can read about it here:
https://en.wikipedia.org/wiki/PID_controller
It's actually very easy to implement - the only challenge is tuning the PID, but I don't imagine that would be that big a deal, even the auto-tuners out there would do a radically better job then what you've currently got implemented.  Then you'll get a rock solid line in terms of hashing performance.
Anyway, hope that helps - as the idea behind it awesome, just looking forward to it being more consistent...

I have tech degree and know about this algo (though I learned it many years ago). And I'm going to implement it for "-tt" option, it's really a good idea, but I cannot find time again and again though it is in my todo list for a long time. For other options like "-ttdcr" or "-ttli" this algorithm may fail because they are related to chip temperature which is adjusted by drivers or by "-tt" option, so two (or even three if you set both "-ttdcr" and "-ttli") alogrithms will try to control temperature on chip, it may cause same waves.
In other words: for example, you use AMD drivers to manage fan ("-tt 1" or "-tt 0") and also set "-ttli" option. So both drivers and miner will try to manage temperature - result depends on target temperatures for both algos and on what algos are used, but waves are possible in many cases.

[Claymore]

v6.1:

- Fixed issue with high CPU load for NVIDIA cards.
- NVIDIA: now two builds are available: for CUDA 7.5 and for CUDA 6.5.
- Fixed issue with resetting WattMan settings.

NOTE: Still no Linux version yet, I need to have stable Windows version firstly.
NOTE2: I don't see good hashrates in cuda 6.5, at least in Win7. May be some specific drivers version must be used?

How do we make it use cuda 6.5 path?

You don't have to use cuda path. To use cuda6.5 you should copy files from "cuda6.5" folder (download miner v6.1 and find this folder) to main miner folder.

[GH]

Hey Claymore, really like your miner, been running it on a bunch of machines really without issues.  Very excited to see the -ttli feature, and have been fiddling around with it, with mixed results.  I can tell you are using some simple logic for controlling fan speed and intensity - really too simple for this application.  While on some of my machines it achieves a decent level of stability, on many it does what's called 'flapping', meaning it will oscillate up and down, as it chases the correct setting.  This is actually a really common problem in industrial automation - and specifically in oven control (not cooking ovens, but ovens used with circuit boards).  The way they achieve stability is by using an algorithm called a 'PID' - you can read about it here:
https://en.wikipedia.org/wiki/PID_controller
It's actually very easy to implement - the only challenge is tuning the PID, but I don't imagine that would be that big a deal, even the auto-tuners out there would do a radically better job then what you've currently got implemented.  Then you'll get a rock solid line in terms of hashing performance.
Anyway, hope that helps - as the idea behind it awesome, just looking forward to it being more consistent...

I have tech degree and know about this algo. And I'm going to implement it for "-tt" option, it's really a good idea, but I cannot find time again and again though it is in my todo list for a long time. For other options like "-ttdcr" or "-ttli" this algorithm may fail because they are related to chip temperature which is adjusted by drivers or by "-tt" option, so two (or even three if you set both "-ttdcr" and "-ttli") alogrithms will try to control temperature on chip, it may cause same waves.
In other words: for example, you use AMD drivers to manage fan ("-tt 1" or "-tt 0") and also set "-ttli" option. So both drivers and miner will try to manage temperature - result depends on target temperatures for both algos and on what algos are used, but waves are possible in many cases.

In fact I was quite impressed how good that "simple logic" works. It is the best way we found to keep the cards at the temperature we want - at least if they have sufficient headroom to remove the heat (=fan speed below 100%).

[qqqq]

It stuck and do not anything with 6.5 Cuda or Higher (Win7x64 368.39 Gtx 970) At 6.0 ver of miner i got 51 MH\s with 3 cards but 100 % load of CPU.

Code:

DUAL MINING MODE ENABLED: ETHEREUM+SIACOIN
ETH: HTTP SOLO mode
 SC - connecting to siamining.com:9980/miner/header?address=e325c219e6acb9e57b8f
5adf5c60def3028ecb12f9b25a0c78cbf59ba97e4104c2ad1c5183fb
Ethereum HTTP requests time (-etht) is set to 200 ms
Watchdog enabled
Remote management is enabled on port 3333

 SC: Connected (siamining.com:9980/miner/header?address=e325c219e6acb9e57b8f5adf
5c60def3028ecb12f9b25a0c78cbf59ba97e4104c2ad1c5183fb)

In log seems infinity things

Code:

21:57:03:935 c68 http got answer: 000000007fffffffffffffffffffffffffffffffffffffffffffffffffffffff000000000000008850f1c65f09c211543bf9cdabadbc12e93ed0a9311db853b500000000000000003562b3570000000091c8c36f7ac9beef090b9ed0716168f612650606d89601496c419f6371c71cb2
21:57:03:940 c68 SIA job updated
21:57:04:010 d84 parse packet: 242
21:57:04:010 d84 new buf size: 0
21:57:04:225 d84 parse packet: 242
21:57:04:225 d84 new buf size: 0
21:57:04:435 d84 parse packet: 242
21:57:04:435 d84 new buf size: 0
21:57:04:655 d84 parse packet: 242
21:57:04:655 d84 new buf size: 0
21:57:04:870 d84 parse packet: 242
21:57:04:870 d84 new buf size: 0
21:57:05:085 d84 parse packet: 242
21:57:05:085 d84 new buf size: 0
21:57:05:300 d84 parse packet: 242
21:57:05:300 d84 new buf size: 0
21:57:05:515 d84 parse packet: 242
21:57:05:515 d84 new buf size: 0
21:57:05:730 d84 parse packet: 242
21:57:05:730 d84 new buf size: 0
21:57:05:935 d84 parse packet: 242
21:57:05:935 d84 new buf size: 0
21:57:06:156 d84 parse packet: 242
21:57:06:156 d84 new buf size: 0
21:57:06:376 d84 parse packet: 242
21:57:06:376 d84 new buf size: 0
21:57:06:591 d84 parse packet: 242
21:57:06:591 d84 new buf size: 0
21:57:06:796 d84 parse packet: 242
21:57:06:796 d84 new buf size: 0
21:57:07:011 d84 parse packet: 242
21:57:07:011 d84 new buf size: 0
21:57:07:216 d84 parse packet: 242
21:57:07:216 d84 new buf size: 0
21:57:07:416 d84 parse packet: 242
21:57:07:416 d84 new buf size: 0
21:57:07:621 d84 parse packet: 242
21:57:07:621 d84 new buf size: 0
21:57:07:821 d84 parse packet: 242
21:57:07:821 d84 new buf size: 0
21:57:07:841 6e0 recv: 51
21:57:07:841 6e0 srv pck: 50
21:57:07:846 6e0 srv bs: 0
21:57:07:846 6e0 sent: 140
21:57:08:026 d84 parse packet: 242
21:57:08:026 d84 new buf size: 0
21:57:08:231 d84 parse packet: 242
21:57:08:231 d84 new buf size: 0
21:57:08:431 d84 parse packet: 242
21:57:08:431 d84 new buf size: 0
21:57:08:636 d84 parse packet: 242
21:57:08:641 d84 new buf size: 0
21:57:08:841 d84 parse packet: 242
21:57:08:841 d84 new buf size: 0
21:57:08:946 c68 http send request:
21:57:09:041 d84 parse packet: 242
21:57:09:041 d84 new buf size: 0
21:57:09:081 c68 http got answer: 000000007fffffffffffffffffffffffffffffffffffffffffffffffffffffff000000000000008850f1c65f09c211543bf9cdabadbc12e93ed0a9311db853b500000000000000003562b357000000000b43f5cb248868ce0c6b729b77dfc55f8643a340bf368057a7592d42115f0114
21:57:09:086 c68 SIA job updated
21:57:09:251 d84 parse packet: 242
21:57:09:251 d84 new buf size: 0
21:57:09:471 d84 parse packet: 242
21:57:09:471 d84 new buf size: 0
21:57:09:691 d84 parse packet: 242
21:57:09:691 d84 new buf size: 0
21:57:09:901 d84 parse packet: 242
21:57:09:901 d84 new buf size: 0
21:57:10:106 d84 parse packet: 242
21:57:10:106 d84 new buf size: 0

Looks like it can't make a DAG.

[n0x1ous]

v6.1:

- Fixed issue with high CPU load for NVIDIA cards.
- NVIDIA: now two builds are available: for CUDA 7.5 and for CUDA 6.5.
- Fixed issue with resetting WattMan settings.

NOTE: Still no Linux version yet, I need to have stable Windows version firstly.
NOTE2: I don't see good hashrates in cuda 6.5, at least in Win7. May be some specific drivers version must be used?

How do we make it use cuda 6.5 path?

You don't have to use cuda path. To use cuda6.5 you should copy files from "cuda6.5" folder (download miner v6.1 and find this folder) to main miner folder.

Thanks....no difference on performance with 980ti/Win10/372.54.  Getting 4 MH/s per card

[goofy73]

Claymore, is it possible to make updating miner via ethman?!

[hatshepsut93]

Is GTX 1060 supported in version 6.1? I have very low hashrates (3.83 for ETH and 38.393 for DCR), while my system is freezing. I have Windows 10 1607 and 372.54 drivers.

[clipto]

After trying a lot of different driver versions, we have it running with CUDA 6.5 and Miner 6.1
Geforce Driver: 350.12

ETH - Total Speed: 21.970 Mh/s, Total Shares: 26, Rejected: 0, Time: 00:08
ETH: GPU0 12.659 Mh/s, GPU1 9.311 Mh/s
SC - Total Speed: 329.548 Mh/s, Total Shares: 36, Rejected: 0
SC: GPU0 189.889 Mh/s, GPU1 139.659 Mh/s

GPU0=GTXTitanX
GPU1=GTX970

----------

At our AMD setup we have this in the BAT file:
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

Is any of that needed for Nvidia too?
Current batfile is this:

ethdcrminer64.exe -epool etc-eu.suprnova.cc:3333 -ewal user.worker -epsw x -esm 3 -dpool sia.suprnova.cc:7777 -dwal worker.user -dpsw x -dcoin sia -allpools 1

[Claymore]

v6.1:

- Fixed issue with high CPU load for NVIDIA cards.
- NVIDIA: now two builds are available: for CUDA 7.5 and for CUDA 6.5.
- Fixed issue with resetting WattMan settings.

NOTE: Still no Linux version yet, I need to have stable Windows version firstly.
NOTE2: I don't see good hashrates in cuda 6.5, at least in Win7. May be some specific drivers version must be used?

How do we make it use cuda 6.5 path?

You don't have to use cuda path. To use cuda6.5 you should copy files from "cuda6.5" folder (download miner v6.1 and find this folder) to main miner folder.

Thanks....no difference on performance with 980ti/Win10/372.54.  Getting 4 MH/s per card

On Win10 you have to use old drivers like 35x.xx. They don't support cuda 7.5, that's why I compiled version for cuda 6.5

Claymore, is it possible to make updating miner via ethman?!

Not yet.

Is GTX 1060 supported in version 6.1? I have very low hashrates (3.83 for ETH and 38.393 for DCR), while my system is freezing. I have Windows 10 1607 and 372.54 drivers.

It is driver issue, miner supports all recent nvidia.

[Claymore]

After trying a lot of different driver versions, we have it running with CUDA 6.5 and Miner 6.1
Geforce Driver: 350.12

ETH - Total Speed: 21.970 Mh/s, Total Shares: 26, Rejected: 0, Time: 00:08
ETH: GPU0 12.659 Mh/s, GPU1 9.311 Mh/s
SC - Total Speed: 329.548 Mh/s, Total Shares: 36, Rejected: 0
SC: GPU0 189.889 Mh/s, GPU1 139.659 Mh/s

GPU0=GTXTitanX
GPU1=GTX970

You don't need all these setx GPU_FORCE_64BIT_PTR etc for nvidia.
Low hashrate, I see >13MH on single 970 on Win7 cuda 6.5 and 359.06 drivers. Though it is bad hashrate anyway, it shows >18MH on Win7 with cuda 7.5 and 368.81 drivers.

[qqqq]

Back to genoil, cause 6.1 not work for me and 6.0 is waste my electricity.

[Flenger]

Driver 372.54 Win7/64 GTX1060 18.7MHs ETC + 250MHs SC dcri 40 without overclocking

Any chance you know your power usage at the wall?

Alas, I do not know

[Claymore]

Back to genoil, cause 6.1 not work for me and 6.0 is waste my electricity.

You mine ETH solo? Not with Parity? Because I got reports about problems with Parity, I changed requests a bit in v6.1, it could cause problems with older version. Changed it to original version, please try this:
https://mega.nz/#!r8IWGIZS!zHiYJ0N2Ywes6ReWoe-VWjMmDCiIfmUGrDInB2nDtJE
Linked with cuda 7.5, just copy .exe it to main miner folder.

[MarkAz]

I have tech degree and know about this algo (though I learned it many years ago). And I'm going to implement it for "-tt" option, it's really a good idea, but I cannot find time again and again though it is in my todo list for a long time. For other options like "-ttdcr" or "-ttli" this algorithm may fail because they are related to chip temperature which is adjusted by drivers or by "-tt" option, so two (or even three if you set both "-ttdcr" and "-ttli") alogrithms will try to control temperature on chip, it may cause same waves.
In other words: for example, you use AMD drivers to manage fan ("-tt 1" or "-tt 0") and also set "-ttli" option. So both drivers and miner will try to manage temperature - result depends on target temperatures for both algos and on what algos are used, but waves are possible in many cases.

Glad to hear you're familiar with it - I learned it in school as well, but never had an opportunity to use it in real life until a few years ago, and was really impressed by the simplicity and performance of it.  You can actually have multiple PID's controlling different elements that affect eachother - in my oven example, they actually manage the temp of the heating element, and the speed of the exhaust fan each with separate PID's.  Another good example are 3 axis gimbals everyone using with quadcopters or Gopro's, which typically use a separate PID for each axis of the gimbal, so while they all are obviously effecting eachother the system is working in unison.  Where these tend to fail is when they go far out of range - so as long as you put some limits in place, it's typically not an issue.

Right now I use your fan management, which seems to work fine and I like the ability to easily set the target temp (ie; "-tt 70"), plus the limit (ie; "-tstop -85")...  But I also think this is a bit backward, as what it's doing then is kind of maintaining a certain level of performance on the chip, but varying fan speed to maintain it.  I wonder if it would make more sense to keep a constant fan speed, and vary the chip speed to maintain temp - assuming there isn't some performance hit making the change, then this would yield the maximum performance for the target temp.  I imagine you could probably do this now using the fan control with a fixed speed and ttli to vary things, right?

[Marvell1]

New 372.54 Nvidia drivers released today.  No fix for maxwell win 10 poor eth performance unfortunately......

Tested here too, same hashrate still.
Claymore, can you please work this out?
Would be nice if besides our AMD rig, we could use your miner on our Nvidia one too.

The problem is not in the miner. The problem is in nvidia drivers. You have to use old drivers on Win10, new drivers don't work as they must do. But old drivers don't support cuda 7.5 which is used by the miner. So I have to release at least two versions of miner - for cuda 7.5 and for cuda 6.5. It's annoying, especially because I see less hashrate on cuda 6.5.

@claymore.  Still locks up with rx470. This round made it 10hr 42 min.  Always locks up on the Dev fee switch at random times of course. Opencl hang.  Have to shut computer down with power switch anytime this miner hangs which I feel is odd. The graphics card I am using for monitor isn't the rx470.  Not sure what can be done but looks like I'll go back to genoil. That miner doesn't have any loss in hash rate over 24hr and it just works.  Claymore can you do something like genoil where you can throw the opencl.dll into same directory as miner and it use that instead.  Seems to do the trick.

Best Regards
d57heinz

If you think that the problem is in drivers you can place driver files to the folder with miner. But it seems you mine not ETH and not EXP, so miner has to rebuild DAG every time and I don't see any ways to fix it if it hangs on your system. Perhaps you really need different miner for your coin.

I have mutiple versions of 470 cards , MSI, XFX, and a few Saphire OC cards all 4GB and they all work flawlessley with Calymore not the latest version the last one.

version of miner or display driver.. IM on 16.8.2 hotfix and v6cdm.. Couldnt get last version to work either.  Could you give some more info on your command line setup.. Are you serializing? or using a different intensity by chance.  TIA

Could be that its due to being a mix in my system of old and new.. I got a new msi mobo today with 7 pcie so i will start that build off with the rx470 by itself to eliminate any issue my home pc may have  Thanks for letting me know that at least someone is having a good exp with the miner and the new amd cards. gives me hope. Now i just have to figure out why genoil works and this one dont.  I let genoil run for 4 days and it never glitched.  hmmmm

Best Regards
d57heinz

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-us-west1.nanopool.org:9999 -ewal xxxxxxxxxxxxxxxxxxxxxxxxxxxx/Marvell9.57/ola_lawal@yahoo.com -allpools 1 -allcoins 1 -dpool stratum+tcp://sia-us-west1.nanopool.org:7777 -dwal xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx/Marvell9.57/ola_lawal@yahoo.com -dcoin sia -dcri 40 -tt 75

that's what I'm using for my 470s

the 8GB card arrived today and it gets 23mhs stock vs the 20 mhs stock on the 4 gig cards

I also installed the hotfix 16.8.2 over 7.3 and it fixed the issue with upping mem freq over 1700 , all 470s can get to 2050 mem now

Right now I upped the mem to 1750 for all cards and the 8 gig card gets to 24 mhs eth and 467 sia , all the 4 GB cards arfe 21.2 and 427 sia.

I think I could get these cards way higer with a custom bios though all in all they are decent cards with no 480s in stock