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

[zawawa]

Would there other kernels still work if I wanted to switch to eth or monero

I haven't tested them yet, but there are no reasons they shouldn't work.
I am planning to work on Monero and Ethereum once I am done with Zcash.

[zawawa]

I am very happy to hear that the new version is working for those with AMD cards.
As for NVIDIA cards, I will stick with OpenCL for now as I can do pretty much whatever I want with inline assembly.
I don't see any reasons why I cannot make my miner as fast as other miners on NVIDIA cards, if not faster, once I get really serious about it.
I just want to catch up with Claymore's first. One thing at a time, you know.

[m1n1ngP4d4w4n]

I am very happy to hear that the new version is working for those with AMD cards.
As for NVIDIA cards, I will stick with OpenCL for now as I can do pretty much whatever I want with inline assembly.
I don't see any reasons why I cannot make my miner as fast as other miners on NVIDIA cards, if not faster, once I get really serious about it.
I just want to catch up with Claymore's first. One thing at a time, you know.

Im eager to see a fast opencl miner for nvidia, see if cuda can be beaten, not gonna be an easy task Good luck and keep up the good work man, you rock  

[antantti]

Quick tests with 3x390 stock:

GG 860 sol/s
CM 960 sol/s
Optiminer 1050 sol/s

[UberMakedonische]

Just a quick THANKS!!!
Keep up with the awesome work...
As my addition here is my youtube video about this miner...
https://www.youtube.com/watch?v=K3CT5MIhn4s
Thanks Again

[Za1n]

Just tested this out on a fresh out of the box (non-optimized) RX470 4GB and I am getting 285 sols without changing anything. I also ran the same test with Claymore's V11 for comparison purposes and there I can get ~218 sols, so you are really closing in on the gap, good job!

Considering that the Claymore miner has a fee and your does not, the performance gap is even smaller than it first appears. I also like that you decided to utilize the sgminer interface on this latest release.

[toptek]

got my 480 doing 215 a tad over clocked default bios,  cm11 does 250  at stock setting not over clock fee on. fee off 236

so it's close..

[zawawa]

I was able to fix Monero. Still trying to get ETH working...

[zawawa]

Ethereum mining for AMD cards is working now, too.
The fixes will be included in the next version.

[joaocha]

Ethereum mining for AMD cards is working now, too.
The fixes will be included in the next version.

Please add some kind of monitor support s2

[zawawa]

Ethereum mining for AMD cards is working now, too.
The fixes will be included in the next version.

Please add some kind of monitor support s2

Awesome Miner should work with Gateless Gate. If not, please let me know.

[zawawa]

So Sir. Claymore seems to have given up at 260 sol/s on RX 480.
My experiments are going well, so this target is totally doable, methinks.

[laik2]

You should stick with CUDA for NVIDIA cards, this is the proper way to go.

[FFI2013]

I was able to fix Monero. Still trying to get ETH working...

awesome I like having the ability to switch coins without having to have 5 different miners to switch also a little while back there was multiminer which was a multi os app with the ability to monitor multi devices and coins with coinwars and whattomine api built in the dev stopped updating it but it works perfect with sgminer and can switch coins on the fly if you have time check it out multiminerapp.com and https://github.com/nwoolls/MultiMiner/releases

[zawawa]

You should stick with CUDA for NVIDIA cards, this is the proper way to go.

I was to switch to CUDA, but it turned out that OpenCL was not a limiting factor as far as performance is concerned, which is all I care about.

[zawawa]

I was able to fix Monero. Still trying to get ETH working...

awesome I like having the ability to switch coins without having to have 5 different miners to switch also a little while back there was multiminer which was a multi os app with the ability to monitor multi devices and coins with coinwars and whattomine api built in the dev stopped updating it but it works perfect with sgminer and can switch coins on the fly if you have time check it out multiminerapp.com and https://github.com/nwoolls/MultiMiner/releases

Looks very sleek. Thanks for the info.

[nerdralph]

So Sir. Claymore seems to have given up at 260 sol/s on RX 480.
My experiments are going well, so this target is totally doable, methinks.

9 rounds takes ~1.2GB of external memory bandwidth when you use GDS for the row counters.  The Rx 480 has 256GB/s of theoretical bandwidth, but even with copied straps, you won't get much more than 200GB/s.  200/1.2 = 166.7 itterations per section, * 1.87 sols/i =~ 312 sols/s.  You'll probably have to experiment with SLC and/or GLC memory IO in order to get over 275 sols.

p.s. Claymore and Optiminer are doing it the messy way with dozens of different pre-built kernels.  One kernel per ISA is the most you should need.  With ROCm/OpenCL it looks like the same conventions are used for argument passing for GCN1.1 and 1.2, so a single combined kernel for both may be possible.

[toptek]

Ethereum mining for AMD cards is working now, too.
The fixes will be included in the next version.

Please add some kind of monitor support s2

Awesome Miner should work with Gateless Gate. If not, please let me know.

AWESOME Miner does i got it working .

[jstefanop]

So Sir. Claymore seems to have given up at 260 sol/s on RX 480.
My experiments are going well, so this target is totally doable, methinks.

9 rounds takes ~1.2GB of external memory bandwidth when you use GDS for the row counters.  The Rx 480 has 256GB/s of theoretical bandwidth, but even with copied straps, you won't get much more than 200GB/s.  200/1.2 = 166.7 itterations per section, * 1.87 sols/i =~ 312 sols/s.  You'll probably have to experiment with SLC and/or GLC memory IO in order to get over 275 sols.

Keep up the good work guys...seems like Claymore has once again screwed over linux users, so I will keep donating here and hopefully we can all switch over to this open source miner soon.

[nerdralph]

So Sir. Claymore seems to have given up at 260 sol/s on RX 480.
My experiments are going well, so this target is totally doable, methinks.

9 rounds takes ~1.2GB of external memory bandwidth when you use GDS for the row counters.  The Rx 480 has 256GB/s of theoretical bandwidth, but even with copied straps, you won't get much more than 200GB/s.  200/1.2 = 166.7 itterations per section, * 1.87 sols/i =~ 312 sols/s.  You'll probably have to experiment with SLC and/or GLC memory IO in order to get over 275 sols.

Keep up the good work guys...seems like Claymore has once again screwed over linux users, so I will keep donating here and hopefully we can all switch over to this open source miner soon.

Yeah, he got his panties all in a bunch after I posted how easy it was to dump his kernel code, and Wolf explained how he extracted the miner ELF binaries.  The silly thing is I only did it for fun as I already knew v10 was too slow to be using any asm tricks, so there was no big performance secrets to learn.