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

[mikhan]

Claymore, I'm using latest 6.1 beta:
1. it still shows "6.0" at the beginning screen
2. i got 5x RX480 in the rig and the miner shows only GPU0-GPU3 fans and temps, where is GPU4? All 5 GPUs are mining. There are no errors in logs about temps and fans

1. Ok, fixed.
2. https://bitcointalk.org/index.php?topic=1433925.msg15950514#msg15950514

2. I didn't touch crossfire settings at all. Funny stuff that I managed to see all 5 temps and fans after I put that one card out from the motherboard and put it in using 16x-16x riser. Don't knoww what happened but it works now.

[scryptr]

First of all, thank you for making the software, there is a problem
My GTX1080 is much slower than the friend of GTX1070, especially the ETH count, I am about 1080 in 20MHz, 1070 of friends can reach 30MHz, ask if there is where the setting is not correct?

TOP END GTX CARDS--

The 1080 uses a different memory type, DDR5X, and it does not perform as well on Dagger-Hashimoto.  It is a design difficulty.  My 980ti is only slightly better than a 970, 22MH/s vs. 20MH/s, and again, the memory design is slightly different between the two cards.  Other algorithms see a big difference in performance.

The cards were not designed to hash Dagger-Hashimoto.       --scryptr

[ozoner99]

Hi Claymore,

Feature Suggestion : Make remote manager able to email you a notification when a miner goes offline for X minutes, or drops below Y% of normal efficiency (and you can define X and Y).

Thanks.

[cyr524]

First of all, thank you for making the software, there is a problem
My GTX1080 is much slower than the friend of GTX1070, especially the ETH count, I am about 1080 in 20MHz, 1070 of friends can reach 30MHz, ask if there is where the setting is not correct?

TOP END GTX CARDS--

The 1080 uses a different memory type, DDR5X, and it does not perform as well on Dagger-Hashimoto.  It is a design difficulty.  My 980ti is only slightly better than a 970, 22MH/s vs. 20MH/s, and again, the memory design is slightly different between the two cards.  Other algorithms see a big difference in performance.

The cards were not designed to hash Dagger-Hashimoto.       --scryptr

Thank you for your advice.
What else do I have to say.... Smashed to buy 1070, it seems that only this one way......

[mkimid]

If the miner doesn't work with some pool - let me know its address and port, I'll check it.

Hi,
I have tested with Parity v1.3.0,
with Miner 6.1 version can support eth_getWork() but, I think, you have forgot to update eth_submitWork()

Code:

22:56:51:051	8e0	eth: job is the same
22:56:51:051	8e0	new buf size: 0
22:56:51:082	1398	ETH: put share nonce cac153d80d7fd593
22:56:51:082	1398	ETH round found 1 shares
22:56:51:097	8e0	ETH: 08/17/16-22:56:51 - SHARE FOUND - (GPU 0)
22:56:51:097	8e0	http send request: {"id":4,"method":"eth_submitWork","params":["0xcac153d80d7fd593","0xd53369c5e3fa6087c7513a92e666643c3289f0027ddaf38a07434b610076c121","0x22d5cfb6500ef4654c3347dd60a67bd05783f209bc623dc595f86c7e50490223"]}

22:56:51:097	8e0	http got answer: {"jsonrpc":"2.0","error":{"code":-32600,"message":"Invalid request","data":null},"id":null}
22:56:51:097	8e0	ETH: Submitted share rejected!

22:56:51:269	8e0	parse packet: 253

It is the same error which has been before

[xeridea]

First of all, thank you for making the software, there is a problem
My GTX1080 is much slower than the friend of GTX1070, especially the ETH count, I am about 1080 in 20MHz, 1070 of friends can reach 30MHz, ask if there is where the setting is not correct?

TOP END GTX CARDS--

The 1080 uses a different memory type, DDR5X, and it does not perform as well on Dagger-Hashimoto.  It is a design difficulty.  My 980ti is only slightly better than a 970, 22MH/s vs. 20MH/s, and again, the memory design is slightly different between the two cards.  Other algorithms see a big difference in performance.

The cards were not designed to hash Dagger-Hashimoto.       --scryptr

GDDR5X has higher latency, and this causes it to run slowly for Dagger-Hashimoto.  It would be good for LBRY though, which is not memory bound, and is currently profitable on Nvidia cards (about the same as Eth+Sia for AMD)

[mkimid]

on 6.1.3, after starting my fans resets to autospeed, but if I set the option -tt 0 I can`t keep track of temperatures

Try this:
https://mega.nz/#!HphjhRyI!ynSMiXk4NfNWsLbjmD03noY93Zk5mboCq1UacB-S570

Excellent !!! everything is working

OK, after 15 hours solo I have 9 blocks rejections and no one confirmed that something was wrong with 6.1.4b+parity 1.3

I go back to Genoil ....

even getwork error has fixed in this version, but, still have an error when submit share,

[Marvell1]

Clock control with rx480 is working great in 6.x, but i miss voltage control. Sometimes my miners lose power due to a external problem with some kind of spike of some factory in the neighborhood (when it happens, its always the same hour in the morning). Miners restart and launch claymore miner, but Wattman returns to default settings. This is 21mh/s instead of 27mh/s per card (rx480 4gb). I need to set memory to 2200mhz with 900mV and also underclock gpu in all steps.
Its possible to implement voltage control with claymore miner with the new ADL version?

yeah this would be huge for me as well

[Claymore]

If the miner doesn't work with some pool - let me know its address and port, I'll check it.

Hi,
I have tested with Parity v1.3.0,
with Miner 6.1 version can support eth_getWork() but, I think, you have forgot to update eth_submitWork()

Code:

22:56:51:051	8e0	eth: job is the same
22:56:51:051	8e0	new buf size: 0
22:56:51:082	1398	ETH: put share nonce cac153d80d7fd593
22:56:51:082	1398	ETH round found 1 shares
22:56:51:097	8e0	ETH: 08/17/16-22:56:51 - SHARE FOUND - (GPU 0)
22:56:51:097	8e0	http send request: {"id":4,"method":"eth_submitWork","params":["0xcac153d80d7fd593","0xd53369c5e3fa6087c7513a92e666643c3289f0027ddaf38a07434b610076c121","0x22d5cfb6500ef4654c3347dd60a67bd05783f209bc623dc595f86c7e50490223"]}

22:56:51:097	8e0	http got answer: {"jsonrpc":"2.0","error":{"code":-32600,"message":"Invalid request","data":null},"id":null}
22:56:51:097	8e0	ETH: Submitted share rejected!

22:56:51:269	8e0	parse packet: 253

It is the same error which has been before

Fixed, please confirm:
https://mega.nz/#!WoYEDLzb!2wBQPWVdvqOfkkRMoR8LJ_TgO76cTvtVXL5dpRI1h24

[maxmad_x]

Hi Claymore,
first of all thank you for creating such a great software.
 
I have been using dual miner 6.1 on my GTX 1060s. Win 10 anniversary edition. WDDM 2.1. getting 21Mh/s
Memory controller load is 100%
Memory used is 1936MB out of 6GB available.
GPU load is 100%

I have OCed  0/+700
Am I getting bottlenecked at the memory controller? Can you do something to allocate more memory and decrease load on the controller?  
Thank you

[nightclaw]

Hi Claymore,
first of all thank you for creating such a great software.
 
I have been using dual miner 6.1 on my GTX 1060s. Win 10 anniversary edition. WDDM 2.1. getting 21Mh/s
Memory controller load is 100%
Memory used is 1936MB out of 6GB available.
GPU load is 100%

I have OCed  0/+700
Am I getting bottlenecked at the memory controller? Can you do something to allocate more memory and decrease load on the controller?  
Thank you

Please tell me a version driver u use ..  cause i try 1060 x2  use all  claymore 6.0, 6.1  still 4 mh/s

[Gpumanijak]

Guys i have problem with my hashrate, i have 390x sappphire nitro and 390 gigabyte g1, but my sapphire card goes downclock and 390 g1 stable. I am using claymore dual miner, my other components are ssd 250gb samsung, 16gb ram, z68s-g43 (g3), i5 3570, LEPA MaxBron B1000...ati drivers 15.12... How to solve this i tried power limit 23+ on sapphire and then is stable ,is it safe to do this and is there another way, bcs  when i raise powerl limit vrm goes 89-92c??? Gpu 0 is 390x and gpu1 is 390 https://s3.postimg.org/evxh82zrn/gpu.png

[alrami]

My 980ti is only slightly better than a 970, 22MH/s vs. 20MH/s, and again, the memory design is slightly different between the two cards.  Other algorithms see a big difference in performance.

The cards were not designed to hash Dagger-Hashimoto.       --scryptr

Sorry, what is your 980ti manufacturer?
it must be near a 32Mh...

[maxmad_x]

Hi Claymore,
first of all thank you for creating such a great software.
 
I have been using dual miner 6.1 on my GTX 1060s. Win 10 anniversary edition. WDDM 2.1. getting 21Mh/s
Memory controller load is 100%
Memory used is 1936MB out of 6GB available.
GPU load is 100%

I have OCed  0/+700
Am I getting bottlenecked at the memory controller? Can you do something to allocate more memory and decrease load on the controller?  
Thank you

Please tell me a version driver u use ..  cause i try 1060 x2  use all  claymore 6.0, 6.1  still 4 mh/s

Hello, I am using 369 (latest version) Make sure you go to nvidia website and install CUDA toolkit 7.5 or 8.0 whichever you like.
Perform a clean install of drivers when drivers are getting installed (you will have to select "custom installation" option) and then you will get a clean installation prompt. Have you installed anniversary windows 10 update? You will HAVE TO install that before you install new drivers.
so 1) Uninstall current nvidia drivers
2) install windows 10 anniversary edition
3) install latest nvidia drivers
4) install nvida CUDA toolkit
and you should be good to go.
Non OC version gets me 17 Mh/s while OC (just memory OC is necessary) will give you about 20-22 Mh/s variable

Good luck

[scryptr]

My 980ti is only slightly better than a 970, 22MH/s vs. 20MH/s, and again, the memory design is slightly different between the two cards.  Other algorithms see a big difference in performance.

The cards were not designed to hash Dagger-Hashimoto.       --scryptr

Sorry, what is your 980ti manufacturer?
it must be near a 32Mh...

EVGA CLASSIFIED 980ti--

And I've never seen a report of any 980ti getting 32MH/s.  a GTX 1070 might attain 32+MH/s, but 980ti cards will not.  The top-end GTX cards do not "scale" with the number of CUDA units when hashing Dagger-Hashimoto, the way they would with a compute-intensive algorithm.  The only card bigger and better than a 980ti is a Titan, and the Titan does not do well with Dagger-Hashimoto, either.       --scryptr

[qqqq]

on 6.1.3, after starting my fans resets to autospeed, but if I set the option -tt 0 I can`t keep track of temperatures

Try this:
https://mega.nz/#!HphjhRyI!ynSMiXk4NfNWsLbjmD03noY93Zk5mboCq1UacB-S570

Excellent !!! everything is working

OK, after 15 hours solo I have 9 blocks rejections and no one confirmed that something was wrong with 6.1.4b+parity 1.3

I go back to Genoil ....

even getwork error has fixed in this version, but, still have an error when submit share,

I want confirm that something happened with solo ETH (geth 1.4.10) on last revisions of miner. 10 days and no blocks with 150-300 mhs. Claymore could you confirm that ETH solo is ok ? ETC seems fine with accepted blocks (Of course i know about different diffs but anyway).

[alrami]

EVGA CLASSIFIED 980ti--

And I've never seen a report of any 980ti getting 32MH/s.  a GTX 1070 might attain 32+MH/s, but 980ti cards will not.  The top-end GTX cards do not "scale" with the number of CUDA units when hashing Dagger-Hashimoto, the way they would with a compute-intensive algorithm.  The only card bigger and better than a 980ti is a Titan, and the Titan does not do well with Dagger-Hashimoto, either.       --scryptr

i had 06G-P4-4996 (EVGA GeForce GTX 980 Ti FTW GAMING ACX 2.0+)
i was able to reach it at with a bios mod (DTP, voltage, mem and gpu clock up)
but it was 85C hot, so i decide to use 2pcs MSI 970 which are 22Mh at 73C (power was for free. mean just utilize give-or-pay contract)

P.S. i'm not sure that GTX 1070 can meet 32+Mh due to thin memory bandwith. think it will be just little highter than 970

[Claymore]

on 6.1.3, after starting my fans resets to autospeed, but if I set the option -tt 0 I can`t keep track of temperatures

Try this:
https://mega.nz/#!HphjhRyI!ynSMiXk4NfNWsLbjmD03noY93Zk5mboCq1UacB-S570

Excellent !!! everything is working

OK, after 15 hours solo I have 9 blocks rejections and no one confirmed that something was wrong with 6.1.4b+parity 1.3

I go back to Genoil ....

even getwork error has fixed in this version, but, still have an error when submit share,

I want confirm that something happened with solo ETH (geth 1.4.10) on last revisions of miner. 10 days and no blocks with 150-300 mhs. Claymore could you confirm that ETH solo is ok ? ETC seems fine with accepted blocks (Of course i know about different diffs but anyway).

If 5.3 worked for you, 6.1.4b will work too. However, all versions won't work with parity, I fixed it in 6.1.5b only, refer to this msg: https://bitcointalk.org/index.php?topic=1433925.msg15963217#msg15963217

[scryptr]

EVGA CLASSIFIED 980ti--

And I've never seen a report of any 980ti getting 32MH/s.  a GTX 1070 might attain 32+MH/s, but 980ti cards will not.  The top-end GTX cards do not "scale" with the number of CUDA units when hashing Dagger-Hashimoto, the way they would with a compute-intensive algorithm.  The only card bigger and better than a 980ti is a Titan, and the Titan does not do well with Dagger-Hashimoto, either.       --scryptr

i had 06G-P4-4996 (EVGA GeForce GTX 980 Ti FTW GAMING ACX 2.0+)
i was able to reach it at with a bios mod (DTP, voltage, mem and gpu clock up)
but it was 85C hot, so i decide to use 2pcs MSI 970 which are 22Mh at 73C (power was for free. mean just utilize give-or-pay contract)

P.S. i'm not sure that GTX 1070 can meet 32+Mh due to thin memory bandwith. think it will be just little highter than 970

ON THE PREVIOUS PAGE, A POSTER CLAIMED 36MH/s WITH HIS 1070--

So I am doubtful of some claims.  980ti cards don't do well on Dagger-Hashimoto.       --scryptr

[espressodelisi]

Still no linux version of 6.1?
Happy running 5.3 with AMD GPUs but I'd like to try it with NVIDIA GPUs too.