Topic: hsrminer - Nvidia mining software for various algos by palgin&alexkap - page 51. (Read 30791 times)

sure...

OK, I'll check. Won't you mind if I ask you in PM to test nightly build on 1080Ti when fix will be ready?
Unfortunately, I don't have any Ti's, as I know it has non-typical Memory amount and also strange memory interface width, so core tweaking would definitely be needed.

P.S.: situation with Ti's may be caused by this "duplicate avalanche" bug too, but need to check.

Also check pls why one 1080ti works fine and 3x with -d 0,1,2 hash same speed as one card on pool
hash reported by miner is fine

Hi, no, currently no website planned, software is in early beta stage and is not popular at all, need to polish it first, add more algos, features and so on.

Thank you for participation, you need to configure desired pool, username and pass first using command-line options and run via batfile, check hsrminer -h for list of avaliable commands. Ti's are supported.

Thank you for your feedback, will try to move more work to GPU itself, this should lower CPU usage significanlty, but this is how my kernel operates and I need to find workaround for this without speed drops.

However, your 1080 speed is way too low, think it's the "duplicate share" situaton, in that case GPU does "fake" work which isn't counted during hashrate calculation.

P.S: most of my farms are 5xGTX 1060 3G, CPUs on all are AMD FX-4100, current version loads 87% CPU, no crashes overnight.

Hi, have you plan to have a website for this software/project?

1st it is important to make option to solve CPU hog, 99% mining rigs are 4-6 cards in dual core CPU, even i5 can run more then 3 GPU atm

Windows usually flags miner programs as viruses, so nothing unusual there.

_______________________________________

This is what I get when I try to run the miner on my personal computer with 2 x GTX 1080 Ti in it. I tried it with -d 0 and the results are the same. Is it just not initializing the GPUs since they are Ti models? I will try to test it on some 1060s when I get a chance.

Yeah but both 0 and 1 card is names as first card

1070 1264kh, ccminer is doing 1223kh  ---- 5% faster

1080 940, ccminer is doing ~1050kh   ---- 10% slower

yeah I used 2 programs to run each

but CPU usage is high too, duno if you can run more cards on celeron
-------------------------------------------------------------------------------------
4x GTX 1080ti, i5 stuck at 100%, miner start and work 30sec and then exit

Cuda error in func 'neoscrypt_cpu_hash_1080' at line 1430 : the launch timed out and was terminated.

3x GTX 1080ti works per i5 CPU.... there is some speedup of almost 10% (WRONG pool side is no good!)

1x GTX 1080Ti looks to works fine pool side!
-----------------------------------------------------------------------------------------------------------
5x GTX 1070 on celeron -> crash

alexkap didn't reply me, so I prepared Neoscrypt release myself.

LINK: hsrminer_neoscrypt

Tested on 1060 3G, 1070, 1080 (not Ti), due to kernel specific it may not launch on 1060 6G, 1080Ti and Titan series, so I need feedback of that models owners if any want to participate.

***BUG FIXED***

Tested on 1070/1080 mixed rig and on 1060 3G only rig.

Will be waiting for your feedback!

I'd be happy to do it, but I have no source for this now, I've lost it with my HDD at the mid-september.

Ah cool, thanks for the clarification. Best of luck with hsrminer!

Will you be open sourcing your last changes as of August 9th for palginmod? Removing the binaries doesn't replace the need for distributing the source for something you've already distributed.

Thank you, Lyra2v2 and Nist5 will definitely be added, maybe even this year, I was working this way before switching to Neoscrypt.
Equihash is another "hardcore" for me, like Neoscrypt was, so I'll need to rest a bit before another fight with algo 

Used your skunkhash mod before, it was great!!  Look forward to seeing this software develop, will be very happy to send devfee your way once the most profitable algos appear.. Lyra2Rev2, Equihash and Nist5 for me currently.  Wish you good luck and respect your ambition to build a custom miner!!

1300 at stock settings is really darned good for a GTX 1080 on Neoscrypt. I think you will get some traffic once the miner is released since there's quite a few highly profitable Neoscrypt coins out there right now. Good work man!

OK, don't think I can go further with current generation kernel, and I'm a bit tired of Neoscrypt to be fair.

Here's 1080 (not Ti, KFA2) at 100% PL without OC



I'm passing this to alexkap, he will create work distribution code for my kernel, then I'll compile and publish hsrminer-Neoscrypt. How long would it take him - I don't know, but I'll try to push him to finish it faster.

Hi, thanks for your interest, but in fact you're not right.

We've started this project specially to get rid of GPL licensed code and possible legal problems, all kernels in hsrminer are my intellectual property, which I have full right not to publish in source code format. You can say style smells like ccminer, but style similarity is not GPL violation, it's the same as, for example, saying that "Daimler AG" is violating "WAG Group" licenses because their cars also have steering wheels and doors. And program clearly says it was "inspired" by ccminer and EWBF's equihash miner.

This project is not related to any kind of "optimisation" as ccminer-palginmod was. If it was so, why we're losing devfee and not releasing new algos for weeks? ccminer-palginmod had many algos "optimised", why not publish them? Hsrminer kernels are homebrew, and it takes me tons of time (really, I get up at 7AM and go to bed at midnight, spending hours in front of screen) to create something faster than existing software offers. Also, palginmod had no movement since 9th of August this year if you look at github.

It's your right to use this software or not, just to close any speculation around GPL and my violation of it, I'm removing all palginmod binaries from Github, just right time to do it.

Thank you again for raising this question and have a nice day!

Cool work - however please consider this a formal request for the source - is seems the kernels might be related to another project of yours, https://github.com/palginpav/ccminer - which I'd also love the source for. Since the ccminer fork is entirely off a project released under GPL, all the code you've written is also considered GPL.

I'm all for your devfee and am not looking to remove it, however I tend not to run anything I cannot compile myself and I tend to try and do my own optimizations for my linux rigs.

Thanks - looking forward to the releases.

UPDATE: found a way to speedup (10% and more) memory dependant Neoscrypt, but this needs completely other type of kernel, which I'm still learning to write, so hsrminer-Neoscrypt will deifintely be delayed. 

Will try to keep you updated on progerss.

Sorry for the delay, I wish you all pleasant holidays, Merry Christmas and Happy New Year!