The link to polyminer is right there... except there is a 1 behind it... polyminer1... seems to be what you looking for.

I see lots of C/C++ code... some of it is related to randomhash, some of that could be avoided with a delphi to cuda interface library which I have but no others are on the internet as far as I know... so some unneccessary work was done, then again all that c/c++ code will benefit somewhat from c/c++ compiler optimizations perhaps not a lot though. Then there is additional c/c++ source code not really related to random hash only but also stratum and pool mining and such...

Overall it's somewhat interesting... I have GT 520 with only 48 cuda cores... I think this is a fermi based cuda card. I might purchase a GT 720/730 or something like that... which will have 384 cuda cores but still passively cooled and even cooler than GT 520 Not sure what architecture GT 720/730 is or something like that... the asus manufacturer version with blue heatsink. Will look into this later, might also give a nice boost to world of warships Will have to order it online though... cause like PC shops probably don't have it.

I will still try this polyminer/rhminer on this cuda fermi architecture to see if it can work on that as well, if not might require some recompiles here and there and maybe adjustments... older cuda architectures seems supported to... so not sure if rhminer use special cuda 9 features or something... it might though... problem for my machine is visual studio 2017 community edition completely fucked up the installation and deinstallation and probably won't ever work on this system, though this reminds me I do have a windows 10 virtual machine with visual studio 2017 community edition installed, this could probably be used to test/compile this c/c++ source code... that's a little bit interesting... it would be a bit slow to run this vm but doable... could even install additional software... old vmware 8 software though, 32 bit even... but enough to run this... though no cuda support probably... I wonder if newer virtual machine software can "pass through cuda..." like "pass-through graphics drivers" hmmm...

Today there will probably be an AMD presentation about "next horizon"... probably chiplet design of new processor. I also watched some benchmark results. Zen/Threadripper 2920x processor seems like a nice choice if I were to build a new system... this processor can do it all, be productive and still game... the higher models have gaming issues with infinity fabric... the higher models have memory latency issues cause it most be passed from core to core which kinda sucks. I hope their next design will be many cores around a single memory chip to keep memory latencies consistent... that would be awesome.  Then I may wait till 2019 to buy such a system. There is another somewhat problem with this 2920x (promontory chipset) it probably contains hardware backdoors, but nowadays all hardware is pretty vunerable... so perhaps this is a somewhat mute point, but still a concern...

I was wondering what this rhminer thing was during the testing... so now I know... thx jason for informing us about polyminer1's rhminer for cuda and such... not sure if it also runs on opencl (?) anyway was polyminer1 paid to do this work ? I would guess so ?

https://github.com/PascalCoin/PascalCoin/blob/master/PIP/PIP-0020.md

Perhaps not though, this pip does not mention anyway salary/payment figures

This rhminer does have option to either pay or not pay 1% to polyminer I wonder what happens if it is disabled... will it do something weird then ?

@Jason what are your plans for your coinotron mining pool ?!? Will all your miners upgrade to new randomhash mining software ? How is that proceeding ?

Also does anybody have any idea how this will affect nanopool ?!?!?

I am not sure if polyminer1 read this thread my question for him would be:

1. Did you "profile/benchmark" randomhash with delphi/pascal code first to see where major bottlenecks occur in randomhash ? Or was randomhash simply converted to c/c++ without doing any bottleneck/performance analysis in delphi/pascal code ?

Hi,

I see you tasklow on the blockchain lol... with new rhminer... and some other peeps too.

Tried a few settings.

First rhminer complained about error error no credentials given or something but this was probably the windows firewall still blocking, this was a bit difficult to notice since the entire system went slow cause I tried default string which used 4 threads, only dual core system. After shutting rhminer down the firewall question remained and I answered it... to allow it.

Then re-running rhminer finally worked for cpu setup, example below:

C:\test>rhminer.exe -v 2 -r 20 -s http://127.0.0.1:4109 -cpu 0 -cputhreads 1 -e
xtrapayload HelloWorld

  rhminer v0.9 beta for CPU and NVIDIA GPUs by polyminer1 (http://github.com/pol
yminer1)
  NVIDIA CUDA SDK 9.2

  Donations : Pascal account 529692-23
  Donations : Bitcoin address 19GfXGpRJfwcHPx2Nf8wHgMps8Eat1o4Jp

Log   06:16:16   Selecting CPU (GPU1) AMD Athlon(tm) 64 X2 Dual Core Processor 3
800+ to mine on 2 cores with 1 threads
Net   06:16:16   Solomining on deamon 127.0.0.1:4109
Log   06:16:16   MiniWeb: Webserver started on port 7111
Net   06:16:16   Received new Work #1. Work target 0x000136df (diff 0.00001257)
Miner 06:16:26   Shares: Accepted 0  Rejected 0  Failed 0 Up for 00:00:10
Miner 06:16:26   Total: CPU XXXXXXXXX H/S. AVG XXXXXXX.00 (X = censored LOL)

The webserver at 7111 is a bit concerning though, but ok

Then I also tried gpu... actually tried this first I think, but unfortunately eventually it gave this error:

C:\test>rhminer.exe -v 2 -r 20 -s http://127.0.0.1:4109 -gpu 0 -gputhreads 32 -
extrapayload HelloWorld

  rhminer v0.9 beta for CPU and NVIDIA GPUs by polyminer1 (http://github.com/pol
yminer1)
  NVIDIA CUDA SDK 9.2

  Donations : Pascal account 529692-23
  Donations : Bitcoin address 19GfXGpRJfwcHPx2Nf8wHgMps8Eat1o4Jp

Log   06:11:54   Selecting GPU0 GeForce GT 520 to mine with 32 threads
Net   06:11:54   Solomining on deamon 127.0.0.1:4109
Log   06:11:54   MiniWeb: Webserver started on port 7111
Net   06:11:54   Received new Work #1. Work target 0x000165e2 (diff 0.00001091)
GPU0  06:11:54   CUDA call error in cudaSetDevice(m_deviceID) : CUDA driver vers
ion is insufficient for CUDA runtime version.

A quick google learns this is because cuda 9 does not support fermi anymore, last version was cuda 8.

The nice thing is rhminer is a bit more efficient then the pascalcoinminer... something like two or three or four times faster on my system... so that's a nice software upgrade to start with

I tried the kepler version which is lowest/closest to fermi as can be seen here:

https://en.wikipedia.org/wiki/CUDA

Perhaps it's possible to re-compile rhminer with cuda 8 so it can still run on fermi... not sure if rhminer uses cuda 9 features or something... hmmm...

Bye,
  Skybuck.

I would say try more threads.  I am stress testing the TESTNET with The Beast.

15 threads at once.  Will it crash or not in 7 hours?

Take a guess...  I called it The Beast for a reason, you know.   

Update:

6 hours on the TESTNET with CPU maxed out.  Temps did not go above 60 degrees.  Thanks for guessing @Skybuckflying.

BTW.  @Polyminer did an amazing job.  He spent a lot of his own time on this.  I am not removing the dev fee.  That message "Thanks..." is hardcoded.
@Skybuckflying.  He was nice enough to provide the source code and tell you how to remove the dev fee.  Read the GitHub page.

Can you extract some block with this hashrate?

@Polyminer

I'm not sure if you check here.

I ran a test on TESTNET.  A combined GPU+CPU mining for several hours.

15 CPU threads + 400 on each 1080ti (I have two in my Linux system).

The average HR was higher than CPU only and GPU miner did find blocks.


Power usage fluctuates, but GPUs run very cool.


I'm not sure how you managed 200 H/S out of a 1060 6GB.  Maybe 1080tis are not suitable.  I have not guessed the correct amount of threads.

The miner did not crash.  OS:  Kali Linux.

Cheers.

~Thomas

This problem is almost everyone who used Microsoft products mate. So much garbage is stuffed from the first days that it is difficult to work with time. In this case, the Apple is all the better. But this is a privilege of a closed operating system rather.

Hi @tlaskows, Indeed CPU+GPU mining is a little bit slower on the cpu side because of the CUDA driver and some internal stuff that occurs in the miner (mainly gpu mining handling !)

As it was stated many times, GPU mining of RandomHash is slower by design. But it has some room for improvement that will come in the later weeks (if all goes well).

If I read correct, you have 400 GPU threads per 1080. I'm, sure you can raise that substantially. Look at the "Ideal CUDA thread count" on the rhminer main page : https://github.com/polyminer1/rhminer#ideal-cuda-threads-count

Still ~2000 H/s for a system like that is awesome in comparisons to typical systems (i5, i3...)

Cheers.
Polyminer1

Hi,

I posted the official announcement of the rhminer here : https://bt.irlbtc.com/view/5065304.msg47684510#msg47684510

Feel free to askquestions there.

Or join the official discord channel at https://discord.gg/x4Yaq4

cheers.
Polyminer1

Is there a setup guide for PascalCoinMiner_4.0 anywhere? I don't see anything on Github or the Pascal site.