Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 486. (Read 2347659 times)

   
The price has now dropped to 0.0084btc /khash

0.0084 *0.06 * 740 = $0.37296

Yup I'm getting 65 s/s per card on my 750ti's SP.  When I get a chance ..I'll hook my watt meter on my 6 card rig and see what I get.

I was looking at AMD but I think I'll continue with Nvidia, the efficiency of Nvidia I like, now thinking of buying another 1070 and keep the 750ti until the end of the year.

Wrong. The 750ti's are still doing $0.56 per card @ 60 sol/s with the nicehash miner(z-cash). 44 watt in the wall.

0.0128 *0.06 * 740 = $0.56

try to use all identical cards.
i'll go for a 1070, if you sell your 3 750Ti, that should pay for about 50% of what a 1070 costs.
that's what i did, i sell all my 750ti's (i had like 30  ) and buy 1070's my power bill decreased like 30% 

750ti where fantastic cards, the best ones i ever used, but their time for mining is over IMHO.

Thanks for your answers guys, all your advice is welcome.

Another option is not to do anything until there is improvement in mining profit.

amd only cost less but profit is the same on etheruem between a 480 and a 1070, actually a 1070 is a little better, but it cost double

and since zcash is no longer so profitable, you can't count on that

Thanks!

This is a good point, AMD at the moment seems a more logical option maybe I should reconsider and look for some amd although I plan to give Nvidia a chance since the other GPU`s are all Nvidia.

Best bet is to go AMD right now. Nvidia really doesn't have any upsides until Equihash gets figured out and we start hitting a wall on hash speed, then efficiency starts to matter more.

That being said, with a rig that small, I'd go for 1060s though, they'll probably be more in your price range and density isn't a issue.

Thanks for your reply, the price of 1 GTX 1060 6GB and 2 GTX 1050ti 4GB are almost the same, around 300 dollars, GTX 1070 around $ 400

Go for 1060 or 1070, depending on price and your setup

Off topic

Hi, I currently have 3 GTX 1070 and 3 GTX 750ti, I have decided to change two GTX750 ti`s, my doubt is for which GPU to change
2 GTX 1050ti 4GB
1 GTX 1060 6G
or make the effort and buy a GTX1070, I hope someone can guide me

Thanks!

*sorry for my english

I merged the changes from the other forks and got a small speedup. Not worth a release

Equihash is different from the other hashing algos. The limit depends on the algorithm. It's a bit like sorting numbers. You have bubble sort, mergesort,quick sort, radix sort etc etc. Each algorithm has it's own limitations. The maximum speed is unknown. There is no known limit.

BTW all the bullshit about 'memory limitations' people were tooting their horns about as far as Equihash goes is just that, bullshit. Someone came up with a random ass extrapolated number and then the tried to justify it using logic. This is also called a confirmation bias.

Claymore is currently pushing past 300sol on 390 and working on another 'good speedup'.

As I've mentioned before, both here and on the Equihash forums. MCU is a pretty good indicator of how 'memory strapped' a algo is. Even at 100%, you can still optimize code so it's more efficient at memory usage. People are fucking dumb and take one thing way too far and believe it's a 'brick wall' without ever spending any time programming or working on a miner themselves, then spread that shit everywhere to get people to believe them.

1070 for instance on Nicehash is doing 65% MCU while operating at 180sol. TDP for cards is also closer to 60% as well (meaning there is another 40% of card thermal headroom mem/gpu). Theoretical increase is another 33%, without optimizing for memory transfers or efficiency, since both are available.

Meaning 240sol is the bare minimum for a 1070, not the maximum and that's just based off of what we can read off the cards while they're operating. And doesn't include power efficiency or memory optimizations.

That isn't to say Equihash doesn't depend on memory, it's just not in the same boat as Ethereum. Stupid eth-babies.

Sp you think this week you have a time to released new zcash_spmod ? .

Hmmm they don't sell Palit's in the US so I haven't used them. Out of all the cards I have, MSI reds seem to have the best memory OCs. I can get +700 stable while dual mining Eth (which is maximum load situation), +900 for ZEC.

That said I don't go out of my way to min/max all my cards. I usually find a stable clock that works across the board on all the cards and blanket OC them (like 1975 core clock and 600 memory). I don't have enough time to troubleshoot constant reboots on quite a few rigs just for a couple extra cents per day.


Speaking of which. It used to be when a OC failed, say with Maxwell, the cards would fall over into a lower power state and there was no way to actually fix it without a reboot. So when you start a miner back up, the card would sit at 400mhz while all the other cards would be operating normally. Pascal seems to have 'fixed' that, when a OC fails, the driver resets and the card seems like it's working perfectly fine... even though it's unstable. This makes it incredibly hard to find the card which a OC isn't working on without trial and error by declocking a random card you think isn't working properly or blanket declock all the cards on a rig.

Is there anyway to get the old behavior of having a faulty card stay in a low power state after it fails like with Maxwell or figure out which card is failing? I haven't found a meaningful way of doing this yet.

same, with samsung and new bios (micron fix) I even get 2500 (stable on same but some rejects on other cards) on ETH but zcash is write heavy and it will cresh so i keep them 2375 too, need to test 24h stability for more

It can do 2400 mem with ETH but only 2375 ZEC. I need do pump memory voltage in palit thundermaster for this memory clocks