Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 596. (Read 2347641 times)

can you send me that too? i was also a donator for the spread algo

Yes please - you can send it to my email.  I will PM it to you.

I don't remember if I donated for that cryptonight miner, I just remember that with 750ti you had to use w7 if you wanted that 8x60 (?) launch.

I like to live in past sometimes so I took tsiv's miner and tried. Still w7x64 and some 970's, doing 490-500 each. p0 and and some heavy (~1500) core oc.

I did put power limit to 120% but it doesn't help, they are using max 80%.

-edit- we didn't know about p0 back then, we were saving power and took that memory slider all the way to the left, no effect on hash -)
-edit2- memory controller load 22%

I was a donator for spreadminer , could you share it to me , also...

can you send me that too? i was also a donator for the spread algo

Spreadminer 9+ compiled for cuda 8.0 has a small boost on the gtx 1070. I can send the exe to you.

Restart your miner.

http://yiimp.ccminer.org not work good today ?

I have few LBRY in balance for my hashrate ( over 1ghs = 4x gtx 1070 ) .

still too low unless a 1070 can do 1kh, also i'm expecting the diff to follow soon, so no reason to even bother with it

@SP - whatever happened to the spreadminer 10?

Someone mentioned 750h/s on RX480

http://coinmarketcap.com/currencies/monero/

Monero up another 80%

1Khash will give you $2.27

I think I need to take another look at the code. My private kernal does around 500h/s on the gtx 970.

No good ccminer for monero ?

I think the gtx 1070 is good for cryptonite .

I agree. There's also seems there's a big difference in efficiency between the 1080's and some high factory OC 1070's.

My 1070's at +50Mhz are runing at ~1949 Mhz and I tried lowering PL and increasing the OC but then my rig crashes eventually and I have yet to find time finding the cause.

Anyway, have a great weekend!

Nice. Thank you for the details there @bathrobehero.

I'm a bit more focused on efficiency than raw hashrate, as my electricity cost is at €0.20 kWh, as well as many times I am mining at loss (or rather, not strictly mining for a BTC output, or mining for low key hoarding).

But the other thing I would highlight, especially with the larger cards (1070/80, 970/80), is that sometimes, and only to moderate extent, you might lower your power usage without any loss of hashrate. This is a function of operating temperature, clock & voltage controls.

The default clocks for my cards (factory OC'ed), achieve roughly ~340 MH/s @ 180W, with the cards running in the range of 70°C.
I can't overclock them much at this range, without hitting temperature and/or TDP throttling. A little increase is possible, but marginal, and with a good deal of power usage increase.

My tweaked settings run at the factory OC, then +140 clock, -1000 memclock, and 85% TDP, which comes out a temperature of 65°C (fan settings unchanged).
The hashrate stands exactly the same as the factory settings, yet my power usage is down by at least 15%.

Obviously results will vary across different cards, vendors, power circuits, ASIC quality, etc. But the gist of it is, that most times, for any given absolute max hashrate at default settings, you can usually get away with better settings to produce the same hashrate at lesser power usage.
I haven't ventured in similar testing for memory intensive algorithms though, I only ever tweaked to any meaningful extent on compute heavy algorithms.

In your case, the 1070 G1 with +50 Mhz OC at 100% Power limit. Would probably be perfectly stable with a tighter power limit, lowered memory clock, and higher core clock - again - producing the same hashrate, but with lesser power usage. It's not something that would interest everyone though... Things will quickly get messy as one changes from algo to algo, with the cards behaving quite differently from one to the other, and sometimes requiring significant reconfiguration.

Enough mining chatter, weekend here I go! Cheers 

That rig has i5 4690k@default.

I am not sure about this yet but it might find blocks more often than it should againts rigs with more hash+weaker cpu.

1070 G1 with only +50 Mhz OC at 100% Power limit: 270Mhs/178W so 1.48 Mh/W
90% PL: 263/162W so 1.62
80% PL: 252Mhs/142W so 1.77
70% PL: 243Mhs/125W so 1.94
60% PL: 226Mhs/107W so 2.11
50% PL: 202Mhs/91W so 2.21

Take the numbers with a grain of salt as I didn't wait much between readings and the power consumption figures are from nvidia-smi.

Anyway, if you do the math higher hashrate trumps efficiency as it almost asways do (if you have average electricity prices). In fact, I've yet to run into a situation where lowering the power limit would have lead to more profit...

With $0.14 kWh LBRY would have to be priced as low as roughly 7k sats for higher efficiency to matter more than hashrate - at which point people would not mine it.

That's a weird one and counterintuitive, but I think I'd call it a feature rather than a bug.
I would expect the extra load on the CPU would cause scheduling delays for other processes like ccminer, although
it wouldn't affect the GPU hashing at all.

The only thing that comes to mind, very speculatively, is thread affinity. With some CPU cores fully loaded CPU mining
the ccminer threads are more likely to always run on the same cores possibly improving cache performance.

I still have that "cpu bug" with lbry, never a found reason for that. Tdp limited 8-pin 970 does 13% more when running cpuminer at the same time with ccminer.

150-153MH/ gpu, cannot test power usage because it takes almost half an hour to get everything synced and running. 160-170 watts maybe?

MY EVGA CLASSIFIED 980TI GETS 220-250MH/s--

I don't have a wattage reading, never bothered to check.  Likely more than your 1080.  My EVGA 750ti cards get 50MH/s each.  No wattage reading again, though.

Right now I am going to re-arrange cards and power supplies. The plan is to populate the 750ti rig with RX 460 cards, and reload Linux from scratch on a 32GB SSD.  The 750 ti cards will move onto a uATX MB and a 128GB SSD in a 4 card rig.  The rig will have Win 7 x64 and dual boot Lubuntu 16.04 with CUDA 8.0.  That way I will be able to auto switch in Win 7, or compile the latest CCminer in Lubuntu.  The RX 460 cards can mine Ethereum (ETH) at 10-12MH/s each, as fast as a GTX 960.  My 960 cards are destined to move to a 4-card rig as soon as it is affordable.  Energy-wise, the RX 460 cards are between a GTX 750ti and a GTX 960.  EVGA Precision-X 16, their overclock utility, works with a maximum of 4 cards.  SO, 4-card GTX rigs; smaller motherboard, smaller power supply (PSU), better stability and control.  A 600 wat PSU is more than enough for 4 750ti cards.

NOTE:  I just received a payment from YIIMP Library Credits.       --scryptr