[Dev] NeoScrypt GPU Miner - Public Beta Test
-
cgminer was not stable with wolfs kernel, also old nvidia went slower.(280x was fast)
im trying now 280x/sgminer over night.
-
cgminer was not stable with wolfs kernel, also old nvidia went slower.(280x was fast)
im trying now 280x/sgminer over night.
try lowering core clock to 1100 and run w 48 i 15 g2 on 280x/7970 getting between 130 and 150 per card
-
try lowering core clock to 1100 and run w 48 i 15 g2 on 280x/7970 getting between 130 and 150 per card
He’s right - it runs the cards a bit harder; you’ll probably need to drop your clocks a bit.
-
Okay, done. I’m pretty sure it works, but haven’t tested on a Windows installation. This is a zip of my kernel, slighly modified for SGMiner, as well as all the other kernels included on the github’s develop branch, and a Win64 binary. Static compile, no DLLs, just like my standard SGMiner builds on Litecointalk. Also GPG signed, like my standard builds. Someone please test for me and ensure it works.
https://ottrbutt.com/sgminer/neoscrypt/sgminer5-neoscrypt-11-02-2014.zip
And of course, GPG sigs for those that check them (you should be): https://ottrbutt.com/sgminer/neoscrypt/sgminer5-neoscrypt-11-02-2014.zip.sig
Thank you very much wolf. Once again you have come to our rescue and given us our Win64 sgminer builds :) You sir are a legend. Much gratitude and love
-
Thank you very much wolf. Once again you have come to our rescue and given us our Win64 sgminer builds :) You sir are a legend. Much gratitude and love
lol, thanks.
-
lol, thanks.
You are welcome. Please post your FTC address so we can tip you for your hard work.
BTW, your sgminer5-neoscrypt is working **almost **flawlessly, but for some reason sgminer5 doesn’t generate bins properly when using different worksizes. For example, my 7950 uses w128, and my 7870 uses w32, but this sgminer5 only generates bins with w128 for both cards. I get higher hashes on it with my 290x vs cgminer-neoscrypt. Not anything crazy, but 2-3 Kh/s faster. It’s the same speed as cgminer-neoscrypt, but still having a unified sgminer is much better than using crap, old, cgminer.
-
Did I hear correctly… We have a unified miner now?
edit
I’ve know about it for a bit, but yeah, this is awesome. I’m testing out that 64bit compile now.
-
Did I hear correctly… We have a unified miner now?
edit
I’ve know about it for a bit, but yeah, this is awesome. I’m testing out that 64bit compile now.
Just be aware of the worksize issue I reported above. If you use multiple cards with different worksizes it may not work for you. I am still testing sgminer5 to see if I can get different worksizes for different cards to work properly, but no luck yet.
-
I’m just testing it out on a GT 630M.
-
sgminer --algorithm neoscrypt --no-extranonce -I 8 -g 1 -w 256 -o stratum+tcp://ftceu.nut2pools.com:5567
-
You are welcome. Please post your FTC address so we can tip you for your hard work.
BTW, your sgminer5-neoscrypt is working **almost **flawlessly, but for some reason sgminer5 doesn’t generate bins properly when using different worksizes. For example, my 7950 uses w128, and my 7870 uses w32, but this sgminer5 only generates bins with w128 for both cards. I get higher hashes on it with my 290x vs cgminer-neoscrypt. Not anything crazy, but 2-3 Kh/s faster. It’s the same speed as cgminer-neoscrypt, but still having a unified sgminer is much better than using crap, old, cgminer.
First, just so you know - the port was done by an official SGMiner dev who I happen to know - I just annoyed him until he did it and then compiled it, along with adding my kernel and changing it just a tiny bit for SGMiner.
Anyway, my FTC address: 6gD49TDWtSseGno9Hi17XahzejF1BKAtSb
-
Alright, so after testing sgminer5 here are my observations:
It’s the same speed as cgminer-neoscrypt 3.7.8 with wolf0’s kernel.
It works great if you are going to be switching from mining FTC to other coins with different algorithms (sgminer5 is intended to be a unified miner).
It works great if you have one graphics card in your computer.
It works great if you have multiple graphics cards, BUT only if those cards all use the same worksize.
If you try and specify custom worksizes for multiple cards either through passing -w x,x through command line, or by specifying “worksize” : “x,x”, in config what happens is ONLY the first worksize is applied to ALL GPUs, so the resulting bin files will not be optimized for one or more cards.
If you use a multi-GPU setup, and the cards have different worksizes I recommend sticking with cgminer 3.7.8, as the custom worksizes will be applied properly, and the bin files will be optimized for each card.
Note: Copying bin files from cgminer 3.7.8 over to sgminer5 and renaming them to work in sgminer does not work for instances where you are trying to get multiple cards with different worksizes to work.
sgminer5-neoscrypt-11-02-2014 results:
Sapphire AMD 290x Vapor-X 167.5 KH/s
Sapphire AMD HD 7950 145 KH/s
Visiontek AMD HD 7850 63 KH/s
cgminer 3.7.8 with Wolf0’s kernel results:
AMD HD 7870 71 KH/s
-
First, just so you know - the port was done by an official SGMiner dev who I happen to know - I just annoyed him until he did it and then compiled it, along with adding my kernel and changing it just a tiny bit for SGMiner.
Anyway, my FTC address: 6gD49TDWtSseGno9Hi17XahzejF1BKAtSb
I already knew that wolf. Just wanted to report my findings on using the miner for FTC. If I have a chance I will report the issue on the sgminer-dev github.
-
Alright, so after testing sgminer5 here are my observations:
It’s the same speed as cgminer-neoscrypt 3.7.8 with wolf0’s kernel.
It works great if you are going to be switching from mining FTC to other coins with different algorithms (sgminer5 is intended to be a unified miner).
It works great if you have one graphics card in your computer.
It works great if you have multiple graphics cards, BUT only if those cards all use the same worksize.
If you try and specify custom worksizes for multiple cards either through passing -w x,x through command line, or by specifying “worksize” : “x,x”, in config what happens is ONLY the first worksize is applied to ALL GPUs, so the resulting bin files will not be optimized for one or more cards.
If you use a multi-GPU setup, and the cards have different worksizes I recommend sticking with cgminer 3.7.8, as the custom worksizes will be applied properly, and the bin files will be optimized for each card.
Note: Copying bin files from cgminer 3.7.8 over to sgminer5 and renaming them to work in sgminer does not work for instances where you are trying to get multiple cards with different worksizes to work.
I have a multi-GPU setup - I use diff worksizes, and it’s fine.
-
I have a multi-GPU setup - I use diff worksizes, and it’s fine.
You are on linux though right?
-
You are on linux though right?
Yes, but I also use a config file, which could be it, too.
-
Yes, but I also use a config file, which could be it, too.
I am using config file, plus I have also tried passing -w 48,32 and --worksize 48,32 to command line and it still only generates bins with w48. Take a look at this screenshot:
The 7870 Pitcairn is what is taking a hit with sgminer5. It’s not a huge hit though, only about 5 Kh/s with w48 instead of w32, but using same config on cgminer 3.7.8 Neoscrypt it generates the bins with w48 and w32 respectively.
-
I am using config file, plus I have also tried passing -w 48,32 and --worksize 48,32 to command line and it still only generates bins with w48. Take a look at this screenshot:
The 7870 Pitcairn is what is taking a hit with sgminer5. It’s not a huge hit though, only about 5 Kh/s with w48 instead of w32, but using same config on cgminer 3.7.8 Neoscrypt it generates the bins with w48 and w32 respectively.
Odd - works for me, so maybe it is a Windows issue.
-
Odd - works for me, so maybe it is a Windows issue.
Yeah. Maybe I should just bite the bullet and install a Linux distro on my miner.
-
Ugh, I hate the AMD OpenCL compiler - not only does it sometimes mangle code, but it is STUPID. This thing should not have graduated from compiler kidnergarten. It’s also full of bugs - I just figured out it’s flat-out ignoring some of my loop unrolling instructions. I wrote a far better chacha round function and my hash on 280X dropped from 270kh/s to under 200kh/s. Now I need to make it into a macro so I can unroll it manually.