CCminer(SP-MOD) Modded GPU kernels.

[bathrobehero]

I have no idea about the obstacles of developing a GPU miner for Steem but I suspect it would be super profitable.

[1713856817]

1713856817

[bensam1231]

Memory runs cool and doesn't need to be cooled.

Generally true, but for Ethereum I measured the memory being as high as 98°C on a WF3OC 970 (and 104°C on the mini 1 fan models).
Meanwhile the GPU core was limited to 70 or 75°C and not even reaching that.

Reducing what is basically the intensity (–cuda-grid-size) helps but at the loss of hashrate.


Cryptomining blog did an article on it: http://cryptomining-blog.com/7501-nvidia-geforce-gtx-970-thermal-images-for-ethererum-and-decred-mining/




Regarding SIA, KlausT's SIA CUDA miner now supports poolmining but it's kind of glitchy and reports slower speeds even locally. And even lower speeds poolside.

Now with Claymore's dual Eth/Sia miner it will probably go way down in profitability.

There is almost always a correlation between memory temps and GPU temps (usually because the GPU heats up the entire card). The Crypto article shows exactly that. Decred mines at higher GPU temps then Ethereum because it stresses the GPU harder.

Curious how you measured your memory temps. Do you have a infrared temperature gun? I've never heard of memory temps being that high.


Siacoin is at breakeven with other coins now. It doesn't make anymore then say mining Lyra.

[antantti]

The 390 g1 gaming comes with a 6 pin and a 8 pin connector.

6pin= 75Watt
8pin= 150Watt
Motherboard 66 watt

MAX power is 291 watt, but the wall readings is abit higher. (Powered usb risers)

there is nothing stopping the card to draw more than those "max" values.
only thing that matters is power meter at the wall plug.

yes in fact in that cse some of those thing, will simply melt away

and, again, pure Mh/s don't matter: it's efficiency that matters (Mh/s)/W, unless you run on free power.

yeah also 300w for 40MH is not that impressive actually, because 6 x 970 can do the same, 120MH at 900w(6 x 140w + system) and without any tweaking just PL limited to 60%, like i've done in my case

if we were to ignore the consumption we would still use monsters like 7990 or 290x2(this can do 51MH)

I don't know about you guys but for me it is profit that matters.

Most of the time some powersaving brings you more profits but not always. Best hash/ watt and I am loosing profits everyday with every card I have.

When reading these watt/hash threads this starts to look like some loudest whisper contest. Like the_ultimate_mining_GPU would consume zero watts and hash 2Mh/s eth. Or like winner of GP race was that guy who had most fuel left after race.

Really, do your math. Many of you are loosing profits.

I am still happily running some 295x2. Like every hawaii they do at least 34MH/ core without mods. If I ask they deliver, currently I am not asking so much.

[pallas]

The 390 g1 gaming comes with a 6 pin and a 8 pin connector.

6pin= 75Watt
8pin= 150Watt
Motherboard 66 watt

MAX power is 291 watt, but the wall readings is abit higher. (Powered usb risers)

there is nothing stopping the card to draw more than those "max" values.
only thing that matters is power meter at the wall plug.

yes in fact in that cse some of those thing, will simply melt away

and, again, pure Mh/s don't matter: it's efficiency that matters (Mh/s)/W, unless you run on free power.

yeah also 300w for 40MH is not that impressive actually, because 6 x 970 can do the same, 120MH at 900w(6 x 140w + system) and without any tweaking just PL limited to 60%, like i've done in my case

if we were to ignore the consumption we would still use monsters like 7990 or 290x2(this can do 51MH)

I don't know about you guys but for me it is profit that matters.

Most of the time some powersaving brings you more profits but not always. Best hash/ watt and I am loosing profits everyday with every card I have.

When reading these watt/hash threads this starts to look like some loudest whisper contest. Like the_ultimate_mining_GPU would consume zero watts and hash 2Mh/s eth. Or like winner of GP race was that guy who had most fuel left after race.

Really, do your math. Many of you are loosing profits.

I am still happily running some 295x2. Like every hawaii they do at least 34MH/ core without mods. If I ask they deliver, currently I am not asking so much.

Of course it's profit that matters!
I was just saying that Mh/s/W is more precise than just Mh/s.
(Mh/s/W - cost) is obviously even better.
I tend to reiterate the efficiency trallalla' because many miners are overclocking like hell and don't even use a watt meter, thinking they are gaining more :-D

[bathrobehero]

Memory runs cool and doesn't need to be cooled.

Generally true, but for Ethereum I measured the memory being as high as 98°C on a WF3OC 970 (and 104°C on the mini 1 fan models).
Meanwhile the GPU core was limited to 70 or 75°C and not even reaching that.

Reducing what is basically the intensity (–cuda-grid-size) helps but at the loss of hashrate.


Cryptomining blog did an article on it: http://cryptomining-blog.com/7501-nvidia-geforce-gtx-970-thermal-images-for-ethererum-and-decred-mining/




Regarding SIA, KlausT's SIA CUDA miner now supports poolmining but it's kind of glitchy and reports slower speeds even locally. And even lower speeds poolside.

Now with Claymore's dual Eth/Sia miner it will probably go way down in profitability.

There is almost always a correlation between memory temps and GPU temps (usually because the GPU heats up the entire card). The Crypto article shows exactly that. Decred mines at higher GPU temps then Ethereum because it stresses the GPU harder.

Curious how you measured your memory temps. Do you have a infrared temperature gun? I've never heard of memory temps being that high.


Siacoin is at breakeven with other coins now. It doesn't make anymore then say mining Lyra.

Yes, I measured with an IR gun.

More or less what's being used/stressed more is what heats up more but we haven't had such a memory heavy algo as dagger. Note, just because some algos use a lot of memory doesn't mean it's using it as actively.

[Amph]

a 970 cost less than 200 euro now, the dual mining is possible for nvidia too, just no one wants to do it, i remember you were working on one....

where? link please,....

it's private if it ever exist...

[pallas]

a 970 cost less than 200 euro now, the dual mining is possible for nvidia too, just no one wants to do it, i remember you were working on one....

where? link please,....

it's private if it ever exist...

maybe he was referring to the 970 at less than 200

[Amph]

The 390 g1 gaming comes with a 6 pin and a 8 pin connector.

6pin= 75Watt
8pin= 150Watt
Motherboard 66 watt

MAX power is 291 watt, but the wall readings is abit higher. (Powered usb risers)

there is nothing stopping the card to draw more than those "max" values.
only thing that matters is power meter at the wall plug.

yes in fact in that cse some of those thing, will simply melt away

and, again, pure Mh/s don't matter: it's efficiency that matters (Mh/s)/W, unless you run on free power.

yeah also 300w for 40MH is not that impressive actually, because 6 x 970 can do the same, 120MH at 900w(6 x 140w + system) and without any tweaking just PL limited to 60%, like i've done in my case

if we were to ignore the consumption we would still use monsters like 7990 or 290x2(this can do 51MH)

I don't know about you guys but for me it is profit that matters.

Most of the time some powersaving brings you more profits but not always. Best hash/ watt and I am loosing profits everyday with every card I have.

When reading these watt/hash threads this starts to look like some loudest whisper contest. Like the_ultimate_mining_GPU would consume zero watts and hash 2Mh/s eth. Or like winner of GP race was that guy who had most fuel left after race.

Really, do your math. Many of you are loosing profits.

I am still happily running some 295x2. Like every hawaii they do at least 34MH/ core without mods. If I ask they deliver, currently I am not asking so much.

yeah but final profit is minus cost of electricity, you must factor this, and there is no 0 electricity

it's all about if you need more efficency, for me it's important, and i understand that for other is not that important, because they have 5 cent electricity or lower(still important for them if they have 100 gpu or more)

the same for initial investment, it's not that important you can sell later your card, your card will not be worth zero in few months if ever, especially if they are new generation gpu

a 970 cost less than 200 euro now, the dual mining is possible for nvidia too, just no one wants to do it, i remember you were working on one....

where? link please,....

it's private if it ever exist...

maybe he was referring to the 970 at less than 200

ah i sold mine for 220 few days ago on ebay(g1 gaming), below 200 i see them on different private website here in euro

[bensam1231]

The 390 g1 gaming comes with a 6 pin and a 8 pin connector.

6pin= 75Watt
8pin= 150Watt
Motherboard 66 watt

MAX power is 291 watt, but the wall readings is abit higher. (Powered usb risers)

there is nothing stopping the card to draw more than those "max" values.
only thing that matters is power meter at the wall plug.

yes in fact in that cse some of those thing, will simply melt away

and, again, pure Mh/s don't matter: it's efficiency that matters (Mh/s)/W, unless you run on free power.

yeah also 300w for 40MH is not that impressive actually, because 6 x 970 can do the same, 120MH at 900w(6 x 140w + system) and without any tweaking just PL limited to 60%, like i've done in my case

if we were to ignore the consumption we would still use monsters like 7990 or 290x2(this can do 51MH)

I don't know about you guys but for me it is profit that matters.

Most of the time some powersaving brings you more profits but not always. Best hash/ watt and I am loosing profits everyday with every card I have.

When reading these watt/hash threads this starts to look like some loudest whisper contest. Like the_ultimate_mining_GPU would consume zero watts and hash 2Mh/s eth. Or like winner of GP race was that guy who had most fuel left after race.

Really, do your math. Many of you are loosing profits.

I am still happily running some 295x2. Like every hawaii they do at least 34MH/ core without mods. If I ask they deliver, currently I am not asking so much.

Of course it's profit that matters!
I was just saying that Mh/s/W is more precise than just Mh/s.
(Mh/s/W - cost) is obviously even better.
I tend to reiterate the efficiency trallalla' because many miners are overclocking like hell and don't even use a watt meter, thinking they are gaining more :-D

Also have seen this. There is a point where efficiency doesn't matter. You're paying for a high end GPU that can spit out mad amounts of hashrate. There is a point at which saving on power is better then more hashrate (depending on your power cost), but power cost has to start eating close to 50% of your revenue.

If you overvolt (and overclock), that's where power efficiency starts going wildly out of wack. Overclocking doesn't use that much more wattage compared to the hashrate you get out of things. Overvolting is where you start using a lot more. The relationship between clock speed and voltage is logarithmic, meaning you have to put a lot more in and get a lot less out at higher clock speeds.

If you can get higher clocks at the same voltage, it's practically free performance.

[pallas]

If you can get higher clocks at the same voltage, it's practically free performance.

No: even at the same voltage, higher clock speed will use more power:

"
There are several factors contributing to the CPU power consumption; they include dynamic power consumption, short-circuit power consumption, and power loss due to transistor leakage currents.

The dynamic power consumption originates from the activity of logic gates inside a CPU. When the logic gates toggle, energy is flowing as the capacitors inside them are charged and discharged. The dynamic power consumed by a CPU is approximately proportional to the CPU frequency, and to the square of the CPU voltage.
"

so it's linear but in practice it's more than that because of leakage currents and increased fan speed.

[giagge]

Exist one program to powerful SIACOIN on Nvidia maxwell ?

Now mining at 1255 mh/s on windows 10 with gtx 980 1460 boost gpu .

Siacoin Go Pool miner hashrate:
– GTX 1080 – 1945 MHS
– GTX 1070 – 1466 MHS
– GTX 980 Ti – 1220 MHS
– GTX 970 – 803 MHS
– GTX 950 – 385 MHS
– GTX 750 Ti – 301 MHS
– RX 480 – 872 MHS
– R9 280X – 849 MHS
– R9 290x – 1116 MHS

http://cryptomining-blog.com/8080-siamining-another-open-public-siacoin-sc-gpu-mining-pool/

.

i'm doing 905 with my 970, up to 1G with oc(i'm limiting it to 60% tdp) so a 1070 can do almost 1800 or 1g with oc

SIACOIN IS ANOTHER VALUE ADDED COIN--

Like ETH, it is more than another clone coin.  SiaCoin (SC) stores data in the blockchain, you can both mine SC and offer drive space for data storage in return for SC.  My 980ti mined at 1400MH/s when running undisturbed.

I was hoping that one of the coding talents would add it to CCminer.  SC uses the Blake2b algo.       --scryptr

Thanks for info .

[myagui]

If you can get higher clocks at the same voltage, it's practically free performance.

No: even at the same voltage, higher clock speed will use more power:

"
There are several factors contributing to the CPU power consumption; they include dynamic power consumption, short-circuit power consumption, and power loss due to transistor leakage currents.

The dynamic power consumption originates from the activity of logic gates inside a CPU. When the logic gates toggle, energy is flowing as the capacitors inside them are charged and discharged. The dynamic power consumed by a CPU is approximately proportional to the CPU frequency, and to the square of the CPU voltage.
"

so it's linear but in practice it's more than that because of leakage currents and increased fan speed.

My best results are almost always a combination of overclocking + undervolting. It's just a bit of a pain to find what are the fastest stable settings for each algorithm (and even between releases, as the performance/stress for a given algorithm changes).   

[giagge]

Not miner Nvidia for this coin

https://bitcointalk.org/index.php?topic=1541268.0

http://cryptomining-blog.com/8075-cpu-and-amd-gpu-mining-for-lbry-credits-lbc-now-available/

.

[bensam1231]

If you can get higher clocks at the same voltage, it's practically free performance.

No: even at the same voltage, higher clock speed will use more power:

"
There are several factors contributing to the CPU power consumption; they include dynamic power consumption, short-circuit power consumption, and power loss due to transistor leakage currents.

The dynamic power consumption originates from the activity of logic gates inside a CPU. When the logic gates toggle, energy is flowing as the capacitors inside them are charged and discharged. The dynamic power consumed by a CPU is approximately proportional to the CPU frequency, and to the square of the CPU voltage.
"

so it's linear but in practice it's more than that because of leakage currents and increased fan speed.

And I already mentioned that, but you snipped it out. The amount of hashrate you get out is much higher proportionally to the amount of power use. It's almost negligible, baring overvolting.

It's definitely not a linear relationship. In overclocking higher clock speeds will definitely need a lot more power then lower clock speeds. If it was linear, efficiency would be identical across all clocks as efficiency is a multiplicative relationship.

Fan power usage is next to nothing and temps are more so directly related to operating voltage then clock speeds. At a certain threashold +/- 5% that can be made up simply by chip binning (getting a more efficient chip).

My best results are almost always a combination of overclocking + undervolting. It's just a bit of a pain to find what are the fastest stable settings for each algorithm (and even between releases, as the performance/stress for a given algorithm changes).   

This is really no different then overclocking at stock voltage. You're trying to raise the frequency to the highest possible with the current operating voltage. Undervolting lowers that. Overclocking is definitely quite efficient at getting every last ounce of performance out at a certain power envelope and uses very little power comparatively to going up and down the voltage ladder.

[grrrgrrr]

From my experience 1.1x core/memory clock usually results in slightly less than 1.1x rev. and 1.2x power. The optimal operating point is therefore when 20% of power costs less than 10% of rev. If everything's ideal it's not until power cost reaches 0.5x rev. when one might consider throttling.

I recently heard that 1.1x core/memory clock would result in 1.3x power but I couldn't find the source any more. In that case throttle when power cost = 0.3x rev.


However power efficiency is quite something for us who live in apartments. The rig cannot draw more than x kwh from the wall and we often have more than enough BTCBTCBTC to invest. Building rigs that would get the most profit given x kwh requires quite a bit of down clocking.

[Epsylon3]

sorry to post that here, but lbry cuda miner preview (binaries with sources) is on sale for 0.1 BTC... :p

PM me

will maybe get a tip from sp this time ? :p who know ?

tested working without issues on all cuda arch (3.0 to 6.1) on windows and linux

Code:

GPU #0: Gigabyte GTX 1080, 270.14 MH/s
GPU #1: GeForce GTX 1070, 192.61 MH/s
accepted: 339/339 (diff 0.115), 462.73 MH/s yes!
Stratum difficulty set to 16 (0.06250)
lbry block 11086, diff 9152.967
GPU #0: Gigabyte GTX 1080, 270.92 MH/s
accepted: 340/340 (diff 1.919), 462.74 MH/s yes!

EDIT (14 July) : dont send me money anymore please !

[ICOcountdown.com]

sorry to post that here, but lbry cuda miner preview (binaries with sources) is on sale for 0.1 BTC... :p

PM me

will maybe get a tip from sp this time ? :p who know ?

tested working without issues on all cuda arch (3.0 to 6.1) on windows and linux

Code:

GPU #0: Gigabyte GTX 1080, 270.14 MH/s
GPU #1: GeForce GTX 1070, 192.61 MH/s
accepted: 339/339 (diff 0.115), 462.73 MH/s yes!
Stratum difficulty set to 16 (0.06250)
lbry block 11086, diff 9152.967
GPU #0: Gigabyte GTX 1080, 270.92 MH/s
accepted: 340/340 (diff 1.919), 462.74 MH/s yes!

Contribute to this developer, as he has worked very hard on this miner for lbry, contact me or him https://bitcointalk.org/index.php?topic=1541268.0

[antonio8]

sorry to post that here, but lbry cuda miner preview (binaries with sources) is on sale for 0.1 BTC... :p

PM me

will maybe get a tip from sp this time ? :p who know ?

tested working without issues on all cuda arch (3.0 to 6.1) on windows and linux

Code:

GPU #0: Gigabyte GTX 1080, 270.14 MH/s
GPU #1: GeForce GTX 1070, 192.61 MH/s
accepted: 339/339 (diff 0.115), 462.73 MH/s yes!
Stratum difficulty set to 16 (0.06250)
lbry block 11086, diff 9152.967
GPU #0: Gigabyte GTX 1080, 270.92 MH/s
accepted: 340/340 (diff 1.919), 462.74 MH/s yes!

BTC to address in your Sig?

EDIT: Read to fast. I just noticed the pm you.  

[Epsylon3]

- deleted - dont send me money anymore please

[Amph]

sorry to post that here, but lbry cuda miner preview (binaries with sources) is on sale for 0.1 BTC... :p

PM me

will maybe get a tip from sp this time ? :p who know ?

tested working without issues on all cuda arch (3.0 to 6.1) on windows and linux

Code:

GPU #0: Gigabyte GTX 1080, 270.14 MH/s
GPU #1: GeForce GTX 1070, 192.61 MH/s
accepted: 339/339 (diff 0.115), 462.73 MH/s yes!
Stratum difficulty set to 16 (0.06250)
lbry block 11086, diff 9152.967
GPU #0: Gigabyte GTX 1080, 270.92 MH/s
accepted: 340/340 (diff 1.919), 462.74 MH/s yes!

that is the hashrate?

how much libry with one 1070? it look very profitable