CCminer(SP-MOD) Modded GPU kernels.

[sp_]

what real speed on 1060 is monero?

Klaust's opensource cuda 8.0 build is doing 420h/s on stock clock

[1715463178]

1715463178

[1715463178]

1715463178

[1715463178]

1715463178

[1715463178]

1715463178

[Amph]

you can do 8MH with 70% tdp or above 9MH with 100%

You can reduce the power by using the -x parameter. Lower x value, less power.

i get also less hashrate, i made many tests, the best result is 70% tdp(i'm trying 72% now to see if the peak is not above 1000watt, with spreadmod 9 was not possible) with 30 intensity, i get 8MB at around 100 watt



peak 173 watt

[sp_]

Zpool always takes 20-30% of your earnings, so you have to reduce that.

24 hours estimates differences between zpool and yiimp.
20-30% lower estimates. Isn't zpool using the same software as yiimp? Why are the estimates so different..?

[jjjordan]

The old maxwell kernels are no longer available.

sp_ will these miners work on my 970?
Should I use the address in your signature for donation?

[bensam1231]

If you guys didn't get the memo Claymore has a new dual miner for eth+pasc, which is why pasc has been hit so hard. Unless someone improves the Nvidia miner, it's pretty much dead.

480s are earning just as much dual mining as 1070s are just mining normally. -_-

Dual mining eats a lot of energy for RX 480, no ? Not everyone can afford to do that.

At $.1kwh you definitely can. That's average power cost in the US. People are trying to save $.05 on power for $.4 worth of revenue. The relationship between mining income and efficiency isn't linear. Meaning if you use 50 less watts, 100 vs 150watts, it doesn't mean you're 33% more efficient as the cost of power isn't the same as the amount of hash you're trading off.

It's weird, everyone seems to be about efficiency, but no one looks at the actual revenue. Equihash is weird because it runs well even at low clock speeds (and TDP), most algos aren't that way. Unless you have really expensive power, you're going to earn less, that's all on a case by case basis and something you need to calculate yourself though.

Equihash is the ONLY '100watt' 1070 algo, otherwise a 480 uses about the same amount of electricity. It's also has quite a bit less computational power, which is where a 1070 should be acceling at... but it's not right now, either because kernels aren't developed or what it's good at there are too many of us (Nvidia miners) to support.

Guessing too many people picked fights in AMD threads about how efficient Nvidia is, convinced too many others to change sides, and then sunk the ship from the extra weight.

I would definitely trade a $400 video card for a $200 one, and make the same profit. I haven't bought a new 1070 in a few months... I'm saving money, waiting for vega, and seeing what I should do.

1070s and 1060s currently aren't worth it. A 1060 is the easy out option, it's still inferior to buying 480s. ASICs are also a option which are much more stable and ROI is about the same now.

Zpool always takes 20-30% of your earnings, so you have to reduce that.

24 hours estimates differences between zpool and yiimp.
20-30% lower estimates. Isn't zpool using the same software as yiimp? Why are the estimates so different..?

Hashrate and pool luck is what is causing that. The '20-30%' zpool takes wont show up on the pool, you'll simply get paid less. So take the 'actual' amount the pool paid in the last 24hr and then compare it to how much you got paid, that's where you see the 20-30% difference. Taking your pool reported hashrate and multiplying it by the pool 'actual' payrate should give you your actual revenue... and it doesn't with zpool.

[bensam1231]

SP if you can improve SIB that'd be a good idea. Not sure if you've taken a look at it yet. BBR as well. BBR is a really big market.

RIE is another potential market that is CPU only. If someone can look into making a GPU miner for it, that'd also net a decent amount of money.

[sp_]

SP if you can improve SIB that'd be a good idea. Not sure if you've taken a look at it yet. BBR as well. BBR is a really big market.
RIE is another potential market that is CPU only. If someone can look into making a GPU miner for it, that'd also net a decent amount of money.

The opensource SIB kernel is already optimized, difficult to get more than a few percent. Will look into BBR. Seems profitable..

[bensam1231]

SP if you can improve SIB that'd be a good idea. Not sure if you've taken a look at it yet. BBR as well. BBR is a really big market.
RIE is another potential market that is CPU only. If someone can look into making a GPU miner for it, that'd also net a decent amount of money.

The opensource SIB kernel is already optimized, difficult to get more than a few percent. Will look into BBR. Seems profitable..

I'm pretty sure someone already has a GPU miner for RIE as well. There is no way the coin would be as unprofitable as it is right now without one or botnets eating it.

[giagge]

Use hardware wattage for estimate power comsumption , is most precise .

[Nikolaj]

@SP

With my old LBC wallet address, with your LBC SPMOD5, my testing rig doesen't report any hashrate.

4h wasted on a unknown pool (Maybe your, hardlocked, Zpool BTC address? )

Could anyone report if it's working correctly? In the past I had other problems with your donation systems, also because I paid for a non-fee private version, and now we've got a public one with a fee.

You should release a version without the fee for the old clients (or if you wish due tax purposes: "donators")

Have a look at this issue because for me your kernel it's no-use (and frankly speaking, fishy)

I've sent you an email days ago with questions, please have a look too.

[Velgelm]

i have tested lbc spmod#5 VS alexis version
sp:



Alexis:



For my sp ver non profitable than alexis

[sp_]

4h wasted on a unknown pool (Maybe your, hardlocked, Zpool BTC address? )

Wich pool did you mine on?

[sp_]

i have tested lbc spmod#5 VS alexis version

65% tdp?

What about 80% tdp or 100%?

And what is the intensity setting used. Wich cards do you use in the rig? And run without the -q option..

[sp_]

You should release a version without the fee for the old clients (or if you wish due tax purposes: "donators")

I sendt you LBRY sp-mod #2 03-sept-2016 wich was at the time around 10% faster or more than the opensource

The opensource was improved by alexei/tvpruvot on the 22 Sep 2016 to match the speed of #2.

So you had the faster kernel for 19 days.

You can't expect me to release a faster kernel for free every time the competition improving their work..

LBRY #5 is around 5-7% faster than 1.8.4. This is my tests, on the gtx 970, gtx 10603gb, and gtx 1070.

The numbers depends on tdp, heat, clock settings etc. The pascal cards trottle.
With my latest kernel, I am running at 2075mhz and 1.8.4 is running at 2100mhz. but my kernel is still 7.5% faster than 1.8.4 on a lower clock. (lower power too)

From yiimp's statistics, on the gtx 1060 3gb , you get the best hash/watt with

80watt Power  (70% plimit)
1850mhz coreclock

[Nikolaj]

4h wasted on a unknown pool (Maybe your, hardlocked, Zpool BTC address? )

Wich pool did you mine on?

ccminer -a lbry -o stratum+tcp://yiimp.ccminer.org:3334 -u myoldLBCaddress -p d=128 --show-diff --cpu-priority 4
pause

Working flawlessly with 1.8.4 (I can see the speed and all the other stuff)
With yours (N'5) nothing at all, it seems working correctly on clientside, but on server side (Yiimp) there's absolutely nothing.


Well, the issue is in your statements and selling conditions, the customer's utilization it's totally irrelevant.

You've sold a software that should have been updated with that logic. I can understand that you want to change rules later on, but I've just pointed out another issue with your fee system, hardlocked inyour closed source kernel 

[sp_]

ccminer -a lbry -o stratum+tcp://yiimp.ccminer.org:3334 -u myoldLBCaddress -p d=128 --show-diff --cpu-priority 4
Working flawlessly with 1.8.4 (I can see the speed and all the other stuff)
With yours (N'5) nothing at all, it seems working correctly on clientside, but on server side (Yiimp) there's absolutely nothing.

You need to wait 2 minutes, then you get to mine 120minutes on your pool.

[Nikolaj]

ccminer -a lbry -o stratum+tcp://yiimp.ccminer.org:3334 -u myoldLBCaddress -p d=128 --show-diff --cpu-priority 4
Working flawlessly with 1.8.4 (I can see the speed and all the other stuff)
With yours (N'5) nothing at all, it seems working correctly on clientside, but on server side (Yiimp) there's absolutely nothing.

You need to wait 2 minutes, then you get to mine 120minutes on your pool.

I've left it 4 hours, nothing at all.

I haven't tested on zpool, but on yiimp there's an issue with the version that I've downloaded from your link

sp: in 19 days the btc amount spent on a 10% optimization it's totally useless, as you know. Btw it's not an issue, I'm just telling you to have a look at your code, preventing potential fees and addresses errors. Talking about VGA thermal throttling issue it's wasting time, we've got a very big experience in matter.

[sp_]

sp: in 19 days the btc amount spent on a 10% optimization it's totally useless, as you know.

Then you should direct your hate to the people who opensourced the optimalizations. Not me.
I release faster kernels for people who donate..

[Nikolaj]

I'm not hating anyone dude, I've just pointed out that the release that you've sent me is broken at the moment, on yiimp

As simple as that.

Whoever is selling the intellectual property of others, drawing profit, should be the subject of disapproval, but not for this reason, if you permit, your end users should be affected. We've talked about this in the past and it's pretty clear that unfortunately, talking about software development and releases, these are concrete issues.

Thanks, I'll try if in the next days if I'll have time to retest.

[sp_]

I've just pointed out that the release that you've sent me is broken at the moment, on yiimp

It's working here. try to run without

--show-diff --cpu-priority 4.

And did you download the latest version from the link? (number 4)

I made 3 public betas before the final version. The url of the archieve is the same.