ethminer-0.9.41-genoil-1.1

[virasog]

The Claymore new eth miner is about 3-5% faster than the ethminer. Is that also faster than 1.0.6?

[Marvell1]

The Claymore new eth miner is about 3-5% faster than the ethminer. Is that also faster than 1.0.6?

I dont know if its verifed that this is for sure 5% faster unless mining pure eth ? also it does not work with stratum and worker name password so keep that in mind

[Genoil]

The Claymore new eth miner is about 3-5% faster than the ethminer. Is that also faster than 1.0.6?

I have only tested 1.0.7 against claymore's on a 7950 in ETH only mode. 1.0.7 has a different kernel but shouldn't affect speed on AMD that much. My miner showed slightly higher hash rates but that could be due to a different way of calculation.

[Casdinyard]

The Claymore new eth miner is about 3-5% faster than the ethminer. Is that also faster than 1.0.6?

I have only tested 1.0.7 against claymore's on a 7950 in ETH only mode. 1.0.7 has a different kernel but shouldn't affect speed on AMD that much. My miner showed slightly higher hash rates but that could be due to a different way of calculation.

Is your 1.0.7 faster than the previous versions of 1.0.6 or 1.0.4? Do they use the same mining kernel?

[Genoil]

The Claymore new eth miner is about 3-5% faster than the ethminer. Is that also faster than 1.0.6?

I have only tested 1.0.7 against claymore's on a 7950 in ETH only mode. 1.0.7 has a different kernel but shouldn't affect speed on AMD that much. My miner showed slightly higher hash rates but that could be due to a different way of calculation.

Is your 1.0.7 faster than the previous versions of 1.0.6 or 1.0.4? Do they use the same mining kernel?

It's a different kernel but not faster on AMD. I had a stab at trying to bring VGRPS usage down to 64 but didn't succeed. As a nice side-effect though, I managed to make it faster on Nvidia GPUs, almost as fast as the CUDA kernel. That can be useful for people who have issue compiling the CUDA kernel (i.e. OSX, Fedora).

[genuin]

cool performance for this,but i cant coding so i'will try to help my friend 

[Ayers]

The Claymore new eth miner is about 3-5% faster than the ethminer. Is that also faster than 1.0.6?

I have only tested 1.0.7 against claymore's on a 7950 in ETH only mode. 1.0.7 has a different kernel but shouldn't affect speed on AMD that much. My miner showed slightly higher hash rates but that could be due to a different way of calculation.

Is your 1.0.7 faster than the previous versions of 1.0.6 or 1.0.4? Do they use the same mining kernel?

It's a different kernel but not faster on AMD. I had a stab at trying to bring VGRPS usage down to 64 but didn't succeed. As a nice side-effect though, I managed to make it faster on Nvidia GPUs, almost as fast as the CUDA kernel. That can be useful for people who have issue compiling the CUDA kernel (i.e. OSX, Fedora).

it's the 1.07 version private or you will make it public? we need to pay for it?

[Marvell1]

The Claymore new eth miner is about 3-5% faster than the ethminer. Is that also faster than 1.0.6?

I have only tested 1.0.7 against claymore's on a 7950 in ETH only mode. 1.0.7 has a different kernel but shouldn't affect speed on AMD that much. My miner showed slightly higher hash rates but that could be due to a different way of calculation.

Is your 1.0.7 faster than the previous versions of 1.0.6 or 1.0.4? Do they use the same mining kernel?

It's a different kernel but not faster on AMD. I had a stab at trying to bring VGRPS usage down to 64 but didn't succeed. As a nice side-effect though, I managed to make it faster on Nvidia GPUs, almost as fast as the CUDA kernel. That can be useful for people who have issue compiling the CUDA kernel (i.e. OSX, Fedora).

so is 1.0.7 faster on gtx 970s ?

[maxvall]

Anyone not busy trying out dual-mining ETH/DCR with Claymore's new toy, please test a new 1.0.7 feature in the "107" branch: stratum failover.

1. Building without ETH_STRATUM fails (because usage of undefined m_fport)
2. IMO you broke farm failover functionality
3. IMO farm/stratum failover mixing makes confusing

Code:

else if ((arg == "-FF" || arg == "-FS" || arg == "--farm-failover" || arg == "--stratum-failover") && i + 1 < argc)
{
	string url = argv[++i];
	if (mode == OperationMode::Stratum) //!!! m_farmFailOverURL is initialised only in stratum mode
	{
		size_t p = url.find_last_of(":");
		if (p > 0)
		{
			m_farmFailOverURL = url.substr(0, p);
			if (p + 1 <= url.length())
				m_fport = url.substr(p + 1); //!!! m_fport is not defined without ETH_STRATUM 
		}
		else
		{
			m_farmFailOverURL = url;
		}
	}
}

P.S. I perform 24h+ testing of 1.0.6 + "added mutex to stratum pending calls" + "disable farm stop on reconnect", ALL OK

[Genoil]

The Claymore new eth miner is about 3-5% faster than the ethminer. Is that also faster than 1.0.6?

I have only tested 1.0.7 against claymore's on a 7950 in ETH only mode. 1.0.7 has a different kernel but shouldn't affect speed on AMD that much. My miner showed slightly higher hash rates but that could be due to a different way of calculation.

Is your 1.0.7 faster than the previous versions of 1.0.6 or 1.0.4? Do they use the same mining kernel?

It's a different kernel but not faster on AMD. I had a stab at trying to bring VGRPS usage down to 64 but didn't succeed. As a nice side-effect though, I managed to make it faster on Nvidia GPUs, almost as fast as the CUDA kernel. That can be useful for people who have issue compiling the CUDA kernel (i.e. OSX, Fedora).

it's the 1.07 version private or you will make it public? we need to pay for it?

No, I just need a few people to test it and confirm the new stratum failover feature works. Also, somebody taking the chance to use the native stratum miner with a serious production rig for a longer period to confirm stability would be awesome. I just don't have that hardware

I just pushed a win64 binary here: https://github.com/Genoil/cpp-ethereum/blob/107/releases/ethminer-0.9.41-genoil-1.0.7b1.zip?raw=true

so is 1.0.7 faster on gtx 970s ?

No it isn't in CUDA mode (-U), although I relaxed the restrictions on block size. They used to be 32/64/128, but now you get just use any multiple of 8. Even not a multiple of 8 but that won't give you any share. Useful block sizes that you could try are 32,64,128, 224, 288, 448 and 896. In my experience it doesn't do that much, although combined with differetn global work sizes, it may yield a bit better results.

In OpenCL mode however (-G), all NVidia GPUs should expect a big boost. Still not on the level of native CUDA, but close.

-- edit-- oh just missed @maxvalls reply. I guess I'll have to look into these issues! Thanks!
-- edit2-- i think i fixed it. binary also updated


 

[Genoil]

Anyone not busy trying out dual-mining ETH/DCR with Claymore's new toy, please test a new 1.0.7 feature in the "107" branch: stratum failover.

1. Building without ETH_STRATUM fails (because usage of undefined m_fport)
2. IMO you broke farm failover functionality
3. IMO farm/stratum failover mixing makes confusing

Code:

else if ((arg == "-FF" || arg == "-FS" || arg == "--farm-failover" || arg == "--stratum-failover") && i + 1 < argc)
{
	string url = argv[++i];
	if (mode == OperationMode::Stratum) //!!! m_farmFailOverURL is initialised only in stratum mode
	{
		size_t p = url.find_last_of(":");
		if (p > 0)
		{
			m_farmFailOverURL = url.substr(0, p);
			if (p + 1 <= url.length())
				m_fport = url.substr(p + 1); //!!! m_fport is not defined without ETH_STRATUM 
		}
		else
		{
			m_farmFailOverURL = url;
		}
	}
}

P.S. I perform 24h+ testing of 1.0.6 + "added mutex to stratum pending calls" + "disable farm stop on reconnect", ALL OK

Thanks for testing. God it's an awful mess isn't it

[JuanHungLo]

Genoil, I am trying but I can't get it to work.  What would be the config for using stratum if the following?

# Host and port for your workers
HOST = "192.168.1.101"
PORT = 8080

# Coin address where money goes
WALLET = "<YOUR ETH PAYMENT ADDRESS>"

# Main pool
POOL_HOST = "eth-us.dwarfpool.com"
POOL_PORT = 8008

# Failover pool
POOL_FAILOVER_ENABLE = True
POOL_HOST_FAILOVER = "eth-eu.dwarfpool.com"
POOL_PORT_FAILOVER = 8008

Currently my miners uses the following:

@echo off
setx GPU_MAX_ALLOC_PERCENT 95
setx GPU_USE_SYNC_OBJECTS 1
setx GPU_MAX_HEAP_SIZE 100
setx GPU_SINGLE_ALLOC_PERCENT 95
setx GPU_FORCE_64BIT_PTR 0
ethminer.exe --farm-recheck 200 -t 2 -G -F http://192.168.1.101:8080/myMiner --cl-extragpu-mem 0 --cl-local-work 256 --cl-global-work 16384

Thanks, in advance.  <hoping>

[Genoil]

Genoil, I am trying but I can't get it to work.  What would be the config for using stratum if the following?

# Host and port for your workers
HOST = "192.168.1.101"
PORT = 8080

# Coin address where money goes
WALLET = "<YOUR ETH PAYMENT ADDRESS>"

# Main pool
POOL_HOST = "eth-us.dwarfpool.com"
POOL_PORT = 8008

# Failover pool
POOL_FAILOVER_ENABLE = True
POOL_HOST_FAILOVER = "eth-eu.dwarfpool.com"
POOL_PORT_FAILOVER = 8008

Currently my miners uses the following:

@echo off
setx GPU_MAX_ALLOC_PERCENT 95
setx GPU_USE_SYNC_OBJECTS 1
setx GPU_MAX_HEAP_SIZE 100
setx GPU_SINGLE_ALLOC_PERCENT 95
setx GPU_FORCE_64BIT_PTR 0
ethminer.exe --farm-recheck 200 -t 2 -G -F http://192.168.1.101:8080/myMiner --cl-extragpu-mem 0 --cl-local-work 256 --cl-global-work 16384

Thanks, in advance.  <hoping>

My miner does't support dwarfpool with native stratum. I'm not going to implement a proprietary protocol that is confusing the hell out of devs because it's called stratum, while in fact it isn't following the protocol spec. Unless I get paid to do it of course
 

[maxvall]

I did performance comparison 1.0.6 with Claymore's miner in single ETH mode:
1.0.6: 74,5 MHs (about 71MHs according to pool metrics)
Claymore: 66,7 MHs
It's about -10% of performance
Rig: R7 370 x5, Catalyst 15.12, Win7 64bit, RAM 4GB + PAGEFILE 8GB

[adaseb]

Here are my speeds with Windows 7 and various AMD GPUs.

Stock Ethminer: 73MHS
Version 1.0.4: 73MHS
Version 1.0.6: 67MHS
Version 1.0.7: 73MHS

I used

--cl-local-work 64 --cl-global-work 12288 -t 5


Maybe someone knows how to tweak the latest version for faster speeds

[Genoil]

I did performance comparison 1.0.6 with Claymore's miner in single ETH mode:
1.0.6: 74,5 MHs (about 71MHs according to pool metrics)
Claymore: 66,7 MHs
It's about -10% of performance
Rig: R7 370 x5, Catalyst 15.12, Win7 64bit, RAM 4GB + PAGEFILE 8GB

Pitcairns are known to perform worse with Claymore' miner. He's working on it. My 7950 hashes equally on the two miners.

Here are my speeds with Windows 7 and various AMD GPUs.

Stock Ethminer: 73MHS
Version 1.0.4: 73MHS
Version 1.0.6: 67MHS
Version 1.0.7: 73MHS

I used

--cl-local-work 64 --cl-global-work 12288 -t 5


Maybe someone knows how to tweak the latest version for faster speeds

Ah good to be back at the original speed. Wasn't even aware 1.0.6 being slower

[Vaccinus]

1.0.6 is also bugged i think because i'm still getting that strange unknown error after the dag reset, i'm reverting to the last 4b3, it was working good

[PcChip]

Hi Genoil,

Thanks for making a great miner!

I have your latest version (that you pushed a few hours ago?) running great on Win8.1 , 980Ti

I'm trying to get it to run on Ubuntu Xenial (GTX 680) , however I get the following when running your cmake command:

CMake Error at cmake/FindPackageHandleStandardArgs.cmake:138 (message):
  Could NOT find CryptoPP (missing: CRYPTOPP_INCLUDE_DIR CRYPTOPP_LIBRARY)
  (Required is at least version "5.6.2")


I tried searching for CryptoPP to install it, however google searches only return people having trouble building this

Thanks for any suggestions you may have!

(edit: wow, just realized how old my sig is... I've been out of the crypto game since like 2011 when I sold tons of bitcoins for $7 each...)

[calculatorduck]

So I tried a version compiled with CUDA 6.5 on Windows 10 with the nvidia 347.52 driver and I'm only getting about 3mh/s on an GTX 760. This same build gets me about 13-14mh/s on Win7.

Is Win10 just totally screwed up?

[maxvall]

Here are my speeds with Windows 7 and various AMD GPUs.

Stock Ethminer: 73MHS
Version 1.0.4: 73MHS
Version 1.0.6: 67MHS
Version 1.0.7: 73MHS

I used

--cl-local-work 64 --cl-global-work 12288 -t 5


Maybe someone knows how to tweak the latest version for faster speeds

Ah good to be back at the original speed. Wasn't even aware 1.0.6 being slower

It strange, I get same hashrate on all versions 1.0.3-1.0.7
Win7, Catalyst 15.12, --cl-local-work 128 --cl-global-work 16384