@str4d
I integrated tromp’s optimizations into standalone miner.
Pull request: Integrating tromp solvers into stratum pool miner. by sarath-hotspot · Pull Request #11 · str4d/zcash · GitHub
One issue: While working on standalone miner, after mining some blocks, standalone miner is stopped.
Process is not exited but stopped working.
Raised an issue @github. https://github.com/str4d/zcash/issues/12
ok thanks 
ill get on that in a bit later this afternoon, so by tonight or worst case tomorrow I should have it set up with a 980ti
Hello! Well, yes 5 hours ago I had dreams Yes, you obviously can mine GPU and CPU at the same time. There is no best CPU until there is no the best software implementation. Currently any multi core CPU, I suppose 4 cores is enough and is a good solution. To mine effectively you should run one instance per core. Tromp’s solver is the best open-source for now, but I really hope nobody (assembly wizard 
) will provide his solution and we will get 28 sols/sec for i7 CPU what actually will beat all GPU implementations from power+usd / performance point of view. So, if you are going to buy 980 Ti, better do not until you will see exact CUDA implementation. Here time is critical if we talk bout profitable mining. RAM seems not critical until you want to run a full node. Mixing, which actually provide privacy for Zcash is OFF by default, turning it ON requires 8GB of RAM for node.
I think I need to create a good looking Debian image for that purpose with deployment instructions :).
hmm. link or guide to build?
cool thank you so much for that CPU/GPU answer 
I already have 980ti’s so its not a problem for me to test it out
what I am considering buying after implimentation is 1080s 
Im order a CPU, Ram, MB, SSD and stuff now for an extra rig
before I get those, probably tonight, I will take apart an ETH mining rig (remove the risers lol) and install ubuntu/mint to test out hte 980ti. By this time tomorrow you should have numbers for 980ti.
also my cousin works for Nvidia, though dont know if she can help
Thanks again for the help,
I’ll be back 
as counsel said,
“You should clone (download) https://github.com/tromp/equihash7
and (then extract and then) compile (open terminal from the folder and type in “makeall”). Then run $time ./eqcuda it should show the amount of
solutions and time spent for them. Like in examples above. Tromp’s GTX
provides around 17 sol/s according to his data.”
If You have connections in Nvidia I think it would be very correct and on time to ask for help in solver implementation because it will potentially cause the massive buy of their cards. If people from NVidia will donate to Mr. Tromp or provide simply a grant it will be even more a generous and wise step. Sure, push your cousin.
thanks for that but im looking for the integratred. a little excited to try it
shes just an engineer who works on cards, she does not know crypto well, but sure I can ask, though i don’t know if she can help, or persuade them to give tromp help.
Well, NVidia is a pretty flexible company. She just need to provide her management with a link to this forum and github, better all resource altogether. It won’t take much time for them to consider it, its a golden opportunity for NVidia.
I thought nobody said he was using assembly for AMD processors, not Intel. AMD instruction set not identical to Intel, sooooo. Only nobody knows for sure : )
Well Nobody , according to his statement knows this and his implementation is only going to be made for AMD, for Intel it was already done.
I guess I missed the Intel part. So Nobody has ASM for both Intel and AMD?
Yes, seems so. We need to wait, its possible that he is targeting to CPU contest, I’m almost sure, if so, he will be probably the winner, at least for CPU.
ok i asked, but really not sure how much help she can get/give if any
So nothing for AMD out yet ?
NV didn’t bother to make a fix for Maxwell-based cards when after driver 350.x ethereum mining was ****ed for Windows
Upto 349.x it was working
They just confirmed that its a driver issue under Windows…
Latest commit replaces C bitfields by manual bitfields.
Now at 4 Sol/s on 4GHz i7 single threaded.
Many more micro-optimizations to go…
Btw, xenoncat’s solver is still over 60% faster!
It truly is 