Author

Topic: Gateless Gate Sharp 1.3.8: 30Mh/s (Ethash) on RX 480! - page 172. (Read 214431 times)

hero member
Activity: 906
Merit: 507
I just tried v3 and I can only seem to get 220 on my 390, 200 on my 290, and both 280 and 270 where lower but sgminer is only building a bin for my Hawaii I tried choosing just the 270 with -d 0 and it still just build hawaii bin

Are you using Linux? The assembly kernel is not ready for Linux yet.
no win 10 pro
sr. member
Activity: 728
Merit: 304
Miner Developer
I just tried v3 and I can only seem to get 220 on my 390, 200 on my 290, and both 280 and 270 where lower but sgminer is only building a bin for my Hawaii I tried choosing just the 270 with -d 0 and it still just build hawaii bin

Are you using Linux? The assembly kernel is not ready for Linux yet.
sr. member
Activity: 728
Merit: 304
Miner Developer
Is there something specific in the ASM that requires modification for linux?
There is a Linux specific requirement, unfortunately.

It turned out the most recent kernel optimization requires MORE work. *sigh*
So, I will just set it aside for now and prepare the assembly version for Linux now.
hero member
Activity: 906
Merit: 507
I just tried v3 and I can only seem to get 220 on my 390, 200 on my 290, and both 280 and 270 where lower but sgminer is only building a bin for my Hawaii I tried choosing just the 270 with -d 0 and it still just build hawaii bin
legendary
Activity: 2174
Merit: 1401
I just uploaded a new pre-release:

https://github.com/zawawawa/gatelessgate/releases/tag/v0.1.3-pre0

The new assembly version is for GCN1 and Windows only for now.
I will work on the Linux version today.
As always, I appreciate your feedback, donations, and even stars on GitHub. Enjoy!

Is there something specific in the ASM that requires modification for linux? Or do you just mean you haven't compiled that linux version yet...if thats the case I can easily compile it myself.
sr. member
Activity: 450
Merit: 255
With latest GG 0.1.3-pre0 my 280x shows me ~250sol/s (Win8.1x64, 15.12). And what I want to say - this is great result! Needs only 15+% speedup, that hits CM11.1(up to 295 sol/s). Thanks!

Are you running the default bat file cause it only runs with a Intensity of 8 I get best results at intensity 10
Here is my testing Bat for the 7870XT and 270x
I have both a Amd and Nvidia in this system so it's Platform 1 for me but could be 0 for you, I also do all my testing Donating hash to the Dev
Code:
@echo off
@set GPU_FORCE_64BIT_PTR 0
@set GPU_MAX_HEAP_SIZE 100
@set GPU_USE_SYNC_OBJECTS 1
@set GPU_MAX_ALLOC_PERCENT 100
@set GPU_SINGLE_ALLOC_PERCENT 100
gatelessgate.exe -k equihash  --gpu-threads 2 --gpu-platform 1 -I 10 -o stratum+tcp://us1-zcash.flypool.org:3333 -u t1NwUDeSKu4BxkD58mtEYKDjzw5toiLfmCu.Test -p z
pause
member
Activity: 78
Merit: 10
--gpu-platform 1

I got this with --gpu-platform 1. With platform 0 it writes, that can't find GPUs, so platform 0 is CPU.
sr. member
Activity: 652
Merit: 266
With latest GG 0.1.3-pre0 my 280x shows me ~250sol/s (Win8.1x64, 15.12). And what I want to say - this is great result! Needs only 15+% speedup, that hits CM11.1(up to 295 sol/s). Thanks!

And for my nvidia 1070 rig - I get error right after start:
"Error -54: Reading results buffer for "ALGO_EQUIHASH failed. (clEnqueueReadBuffer)"

Btw - nice miner coloring Grin
--gpu-platform 1
member
Activity: 78
Merit: 10
With latest GG 0.1.3-pre0 my 280x shows me ~250sol/s (Win8.1x64, 15.12). And what I want to say - this is great result! Needs only 15+% speedup, that hits CM11.1(up to 295 sol/s). Thanks!

And for my nvidia 1070 rig - I get error right after start:
"Error -54: Reading results buffer for "ALGO_EQUIHASH failed. (clEnqueueReadBuffer)"

Btw - nice miner coloring Grin
hero member
Activity: 672
Merit: 500
Nice. Now we need linux version.
Thank you zawawa for hard work will donate soon.
full member
Activity: 224
Merit: 100
CryptoLearner
No problem. Good news, peeps! I was able to achieve 453 sol/s on 7990 without the GCN assembly.
That should translate into 500 sol/s with the GCN assembly.
The trick was to run the slot loads/radix sort and collision search concurrently. I'm getting really good at this!

Nice work man, you rock  Wink
sr. member
Activity: 728
Merit: 304
Miner Developer
No problem. Good news, peeps! I was able to achieve 453 sol/s on 7990 without the GCN assembly.
That should translate into 500 sol/s with the GCN assembly.
The trick was to run the slot loads/radix sort and collision search concurrently. I'm getting really good at this!
member
Activity: 129
Merit: 10
Scratch that, I managed to get it running with new kernels... but only 180sols on modded rx470s

[2017-01-30 00:14:16] GPU0                | (5s):178.4 (avg):179.1h/s | A:10 R:0 HW:0 WU:6.308/m                   
[2017-01-30 00:14:16] GPU1                | (5s):182.1 (avg):178.9h/s | A:6 R:0 HW:0 WU:3.785/m                   
[2017-01-30 00:14:16] GPU2                | (5s):173.4 (avg):176.6h/s | A:6 R:0 HW:0 WU:3.785/m                   
[2017-01-30 00:14:16] GPU3                | (5s):179.3 (avg):179.8h/s | A:13 R:0 HW:0 WU:8.201/m                   
[2017-01-30 00:14:16] GPU4                | (5s):179.9 (avg):179.4h/s | A:8 R:0 HW:0 WU:5.047/m                   
[2017-01-30 00:14:16] GPU5                | (5s):179.9 (avg):178.1h/s | A:10 R:0 HW:0 WU:6.308/m   

Is the windows version using the equihash-gcn1-ocl12-32bit.asm kernel?



Yes. Be patient, guys. I'm working on something here.

Sorry, I just read backwards and see "The new assembly version is for GCN1 and Windows only for now."
sr. member
Activity: 728
Merit: 304
Miner Developer
Scratch that, I managed to get it running with new kernels... but only 180sols on modded rx470s

[2017-01-30 00:14:16] GPU0                | (5s):178.4 (avg):179.1h/s | A:10 R:0 HW:0 WU:6.308/m                   
[2017-01-30 00:14:16] GPU1                | (5s):182.1 (avg):178.9h/s | A:6 R:0 HW:0 WU:3.785/m                   
[2017-01-30 00:14:16] GPU2                | (5s):173.4 (avg):176.6h/s | A:6 R:0 HW:0 WU:3.785/m                   
[2017-01-30 00:14:16] GPU3                | (5s):179.3 (avg):179.8h/s | A:13 R:0 HW:0 WU:8.201/m                   
[2017-01-30 00:14:16] GPU4                | (5s):179.9 (avg):179.4h/s | A:8 R:0 HW:0 WU:5.047/m                   
[2017-01-30 00:14:16] GPU5                | (5s):179.9 (avg):178.1h/s | A:10 R:0 HW:0 WU:6.308/m   

Is the windows version using the equihash-gcn1-ocl12-32bit.asm kernel?



Yes. Be patient, guys. I'm working on something here.
member
Activity: 129
Merit: 10
Scratch that, I managed to get it running with new kernels... but only 180sols on modded rx470s

[2017-01-30 00:14:16] GPU0                | (5s):178.4 (avg):179.1h/s | A:10 R:0 HW:0 WU:6.308/m                   
[2017-01-30 00:14:16] GPU1                | (5s):182.1 (avg):178.9h/s | A:6 R:0 HW:0 WU:3.785/m                   
[2017-01-30 00:14:16] GPU2                | (5s):173.4 (avg):176.6h/s | A:6 R:0 HW:0 WU:3.785/m                   
[2017-01-30 00:14:16] GPU3                | (5s):179.3 (avg):179.8h/s | A:13 R:0 HW:0 WU:8.201/m                   
[2017-01-30 00:14:16] GPU4                | (5s):179.9 (avg):179.4h/s | A:8 R:0 HW:0 WU:5.047/m                   
[2017-01-30 00:14:16] GPU5                | (5s):179.9 (avg):178.1h/s | A:10 R:0 HW:0 WU:6.308/m   

Is the windows version using the equihash-gcn1-ocl12-32bit.asm kernel?

member
Activity: 129
Merit: 10
So I guess I cant just take the 0.1.3 kernel/equihash* files and recompile sgminer-gm 5.5.5 on linux.   I re-defined a bunch of variables that no longer exist in equihash-param.h that ocl.c wanted to be defined, and I got it to compile, but it segfaults during runtime.   So there must be more to it Smiley
sr. member
Activity: 450
Merit: 255
OK Heavily Modded RX 470 Comparison
GG -I 10 = 198-215h/s
GG -I 11 = 202-218h/s

CM = 219-230h/s - Fee = 214.5-225.5h/s

270x matches the 7870XT in comparison
sr. member
Activity: 728
Merit: 304
Miner Developer
Easy, easy. One thing at a time.

I think I finally figured out the last secret to Claymore's performance.
Since the maximum memory throughput is dictated by the hardware limitations, it is imperative to maintain a steady flow of memory transfer without interruptions. It's as simple as that.

Oops, my wife is calling. Gotta go!
hero member
Activity: 610
Merit: 500
zawawa Hi what about PASCminer
sr. member
Activity: 728
Merit: 304
Miner Developer
im still confused on how to build it on ubuntu 16.04 Smiley))

i got the source and compiled it ... running it i get sgminer 5.5.4 and same hashrate im getting from genesis sgminer 5.5.5 ... R9-390x: 285 h/s and RX470: 188 h/s?!


The speeds should be pretty much the same because sgminer-gm has my kernel and I haven't updated yet for Linux.
The new kernel should be ready tonight, hopefully...
Jump to: