Gateless Gate Sharp 1.3.8: 30Mh/s (Ethash) on RX 480!

[QuintLeo]

Since I am away from home until Wednesday and do not have access to dedicated graphics cards, I just decided to try potential replacements for gatelessgate.py as I feel more comfortable with C++ than Python and I think the Python component of SA v5 is rather lacking as far as functionality is concerned. I am planning to evaluate sgminer-gm and nheqminer. I hope they should bring GG on par with Claymore's in terms of usability.

 IMO go with sgminer by a wide margin over the Nicehash one, it has a lot better pool support built in already (the Nicehash one has to be modded to work with anything BUT Nicehash) and I personally prefer the interface quite a bit.

[1713585376]

1713585376

[1713585376]

1713585376

[1713585376]

1713585376

[1713585376]

1713585376

[1713585376]

1713585376

[zawawa]

I was able to build sgminer-gm with VC2015 Community Edition while I was away from home. I will try to port GG's kernel to sgminer-gm. Ideally, I would like to swap gatelessgate.py out, replace it with sgminer-gm, and make sgminer-gm compatible with NVIDIA video cards.

[reb0rn21]

You might take a look at https://github.com/nicehash/nheqminer
at least for NV its nicehash source of their solver

also Claymore 11 started using assembler and he managed a lot to boost a speed for older AMD HW as tahiti, hawai (280x, 380, 390, fury, nano) etc, no RX optimization

[toptek]

nice once you do build it into sgminer-gm which is actually so much better then CM ,i will be dropping my use of CM the only thing CM has over sgminer-gm is speed it can't touch sgminer-gm other wise the Auto fan , API  a lone can out do CM, only reason i don't use sgminer-gm now for Zec is speed, i use it for ETH and XMR and it works just as good as any of CM stuff does for XMR And ETH if not better after you fine tune it .I guess the other nice thing is it is one miner does it all .  with lots of nice options that work with any algorithm that it has support for .  if you don't like the old school BFG/CGMIner look just use the flag

 

Code:

"text-only": true

 it makes it look like CM without the colors kind of or  any of the newer miners out but i like that look my self .

[zawawa]

You might take a look at https://github.com/nicehash/nheqminer
at least for NV its nicehash source of their solver

also Claymore 11 started using assembler and he managed a lot to boost a speed for older AMD HW as tahiti, hawai (280x, 380, 390, fury, nano) etc, no RX optimization

I am planning to port my stuff to nheqminer as well.
 
It is not all that surprising that his miner got a huge boost as I already disclosed my plan for optimizations with a GCN assembler. It is rather amusing to think about how far I can push this "competition" and whether open-source miners can surpass closed-source counterparts in performance. I will work on the GCN assembly version of my kernel when I get back home tonight.

[m1n1ngP4d4w4n]

Nice, keep up the good work  

Will you add Nvidia Cuda then if you branch-out to nheqminer, or just take inspirations/code from the solvers to make it even faster for AMD ?

cheers

[jstefanop]

You might take a look at https://github.com/nicehash/nheqminer
at least for NV its nicehash source of their solver

also Claymore 11 started using assembler and he managed a lot to boost a speed for older AMD HW as tahiti, hawai (280x, 380, 390, fury, nano) etc, no RX optimization

I am planning to port my stuff to nheqminer as well.
 
It is not all that surprising that his miner got a huge boost as I already disclosed my plan for optimizations with a GCN assembler. It is rather amusing to think about how far I can push this "competition" and whether open-source miners can surpass closed-source counterparts in performance. I will work on the GCN assembly version of my kernel when I get back home tonight.

It can still be "open sourced" and not disclose your plans ahead of time Of course that defeats the purpose of open-source collaboration, which is the current sad state of crypto-currencies.

Ironic isent it? A project that was founded upon open source for the betterment of man kind, which at its founding nature was supposed to be the opposite of greed and corruption...has turned into just that...

[QuintLeo]

You might take a look at https://github.com/nicehash/nheqminer
at least for NV its nicehash source of their solver

also Claymore 11 started using assembler and he managed a lot to boost a speed for older AMD HW as tahiti, hawai (280x, 380, 390, fury, nano) etc, no RX optimization

I am planning to port my stuff to nheqminer as well.
 

 That's "in addition to sgminer" not "instead of sgminer" I hope?

 The only thing nheqminer has going for it is if you WANT to work with NiceHash for auto-switch algorythm mining - overall the sgminer interface and controls are a TON better.

[m1n1ngP4d4w4n]

You might take a look at https://github.com/nicehash/nheqminer
at least for NV its nicehash source of their solver

also Claymore 11 started using assembler and he managed a lot to boost a speed for older AMD HW as tahiti, hawai (280x, 380, 390, fury, nano) etc, no RX optimization

I am planning to port my stuff to nheqminer as well.
 

 That's "in addition to sgminer" not "instead of sgminer" I hope?

 The only thing nheqminer has going for it is if you WANT to work with NiceHash for auto-switch algorythm mining - overall the sgminer interface and controls are a TON better.

Agreed, i would wish to have this interface for Nvidia mining (sg-miner), i love it.

[NiceHashSupport]

You might take a look at https://github.com/nicehash/nheqminer
at least for NV its nicehash source of their solver

also Claymore 11 started using assembler and he managed a lot to boost a speed for older AMD HW as tahiti, hawai (280x, 380, 390, fury, nano) etc, no RX optimization

I am planning to port my stuff to nheqminer as well.
 

 That's "in addition to sgminer" not "instead of sgminer" I hope?

 The only thing nheqminer has going for it is if you WANT to work with NiceHash for auto-switch algorythm mining - overall the sgminer interface and controls are a TON better.

Agreed, i would wish to have this interface for Nvidia mining (sg-miner), i love it.

You cannot set clocks and fan speeds with sgminer for Nvidia, because different API is needed. I think most of things could be doable with latest NVSMI lib, but someone has to code all that.

[ioglnx]

Okay time for me to get back to coding ..I will look into the NVML stuff.

[laik2]

You might take a look at https://github.com/nicehash/nheqminer
at least for NV its nicehash source of their solver

also Claymore 11 started using assembler and he managed a lot to boost a speed for older AMD HW as tahiti, hawai (280x, 380, 390, fury, nano) etc, no RX optimization

I am planning to port my stuff to nheqminer as well.
 

 That's "in addition to sgminer" not "instead of sgminer" I hope?

 The only thing nheqminer has going for it is if you WANT to work with NiceHash for auto-switch algorythm mining - overall the sgminer interface and controls are a TON better.

Agreed, i would wish to have this interface for Nvidia mining (sg-miner), i love it.

You cannot set clocks and fan speeds with sgminer for Nvidia, because different API is needed. I think most of things could be doable with latest NVSMI lib, but someone has to code all that.

There are already 2 projects oriented for both NV and AMD.
ccminer is more nvidia oriented
sgminer is more AMD(supports sysfs and ADL currently - sgminer-gm)
It will be easy to port nvidia kernel to ccminer and AMD to sgminer-gm.

[zawawa]

sgminer-gm seems to be working fine with an old SA kernel both with RX 480 and with GTX 1060.
(Obviously, it's really slow and hardware monitoring is not available for the latter.)
Porting GG's kernel to sgminer-gm should not be that difficult if SA's kernel is already working.
Let's see if I can quickly hack up a new version...

[kenshirothefist]

sgminer-gm seems to be working fine with an old SA kernel both with RX 480 and with GTX 1060.
(Obviously, it's really slow and hardware monitoring is not available for the latter.)
Porting GG's kernel to sgminer-gm should not be that difficult if SA's kernel is already working.
Let's see if I can quickly hack up a new version...

Please, make yourself a favour and don't dig into sgminer. Even though it is a great multi-algorithm miner and also we at NiceHash support it for long time, it has too much legacy and is too bloated. It goes back to original minerd, through cgminer, into sgminer, with lots "ifs" inside. Modern algorithms and modern gpus need lightweight miners, and nhqminer is one of them, you're really welcome to try working on nheqminer (it can be easily extended with support for other algorithms, besides equihash).

[zawawa]

sgminer-gm seems to be working fine with an old SA kernel both with RX 480 and with GTX 1060.
(Obviously, it's really slow and hardware monitoring is not available for the latter.)
Porting GG's kernel to sgminer-gm should not be that difficult if SA's kernel is already working.
Let's see if I can quickly hack up a new version...

Please, make yourself a favour and don't dig into sgminer. Even though it is a great multi-algorithm miner and also we at NiceHash support it for long time, it has too much legacy and is too bloated. It goes back to original minerd, through cgminer, into sgminer, with lots "ifs" inside. Modern algorithms and modern gpus need lightweight miners, and nhqminer is one of them, you're really welcome to try working on nheqminer (it can be easily extended with support for other algorithms, besides equihash).

Thank you for your concerns. I can definitely see the "bloatness" you are talking about after playing with the code for a while. It is just that I am familiar with sgminer as a user and that it is easier to experiment with the GCN assembly with it. I have a pretty good sense of how the code works at this point, so I should be able to finish the porting sooner than later and move on.

[zawawa]

In fact, instead of porting GG's kernel as it is now, I am thinking about redesigning GG's kernel so that it can take full advantage of global syncs and GDS. I am not going to disclose too much this time around until I have a working code, but I will keep you guys posted.

[hopfi2k]

In fact, instead of porting GG's kernel as it is now, I am thinking about redesigning GG's kernel so that it can take full advantage of global syncs and GDS. I am not going to disclose too much this time around until I have a working code, but I will keep you guys posted.

highly appreciated!

[zawawa]

I just ported GG's old kernel to sgminer-gm and I am already getting 198 sol/s on RX 480 LOL
I knew Python was no good for the miner...

[zawawa]

I just uploaded a new version of the miner:

https://github.com/zawawawa/gatelessgate

With the user-interface and back-end of sgminer, Gateless Gate is much more feature-rich than before.
Also, it's already faster on RX 480 without the GCN assembly. Yay!

Update: There was a build error on Linux, but it has already been fixed.

[toptek]

DO i need to compile it for windows if i click on the windows version

I get this


https://github.com/zawawawa/gatelessgate/releases/download/v0.1.0/gatelessgate-0.1.0-win64.zip

dead link..