CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 647.

joblo

legendary

Activity: 1470

Merit: 1114

Quote from: pallas on June 16, 2016, 05:47:32 AM

lyra2re (v1) is dead).

Not dead yet, have you taken a look at Nicehash lately? I have no clue what coin still uses lyra2RE
but it's been paying well on Nicehash for a few months,

It peaked at 1.59 BTC/GH/Day (0.53 normalized for nvidia) a couple of hours ago and even higher on CPU.

sp_

legendary

Activity: 2954

Merit: 1087

Team Black developer

Quote from: Amph on June 16, 2016, 07:36:35 AM

48MH on lyra for 1080 is equal to around 32MH(30% slower) on a 1070, and a 32MH is almost 50% higher versus 22MH on a 970

The gtx 970 does 25MHASH with my private bins and stable OC.

25->32 is 28% faster. But then again, I guess the 1070 can be overclocked as well

. I just checked the local prices, and I can get two brand new gtx 970's for one 1070.

giagge

legendary

Activity: 1134

Merit: 1001

I do not know whether to take 1x 1070 gtx or 2x amd 480 for mining all algo .

Amph

legendary

Activity: 3248

Merit: 1070

Quote from: pallas on June 16, 2016, 07:00:59 AM

Quote from: Amph on June 16, 2016, 06:56:29 AM

Quote from: pallas on June 16, 2016, 05:47:32 AM

Honestly I don't think there is much to do on pascal as compared to maxwell, from the miner point of view.
I hope I'm proven wrong, but we should be happy with that 50% to 80% increase in hash/power ratio.
Excluding ether, of course (neoscrypt just needs to be tuned and lyra2re (v1) is dead).

80%? i didn't see any 80% in the 970 vs 1070, if it was the case, i would be more than happy, it would be a must buy for me

according to:

http://cryptomining-blog.com/7973-testing-the-nvidia-geforce-gtx-1080-founders-edition-for-crypto-mining/

1080 vs 970 is up to 120% faster (see lyra2v2 or nist5 for example).

1070 should be about 30% slower, so well over 80% faster than 970, in those cases.

Power usage similar, maybe even lower on 1070.

too bad this is not the case the 1070 is getting like 34MH on lyra, which is only 50% faster than a 970, i dunno about nist, can not find anything

...wait i think your math on % is wrong there, an increase of 50% is equal to a decrease of 30%, so being slower 30% is in fact equal to being faster of 50% not 80%

48MH on lyra for 1080 is equal to around 32MH(30% slower) on a 1070, and a 32MH is almost 50% higher versus 22MH on a 970

now based on the bench i see that a 1070 should be around 70% faster than a 970, so there is definetily room for an improvement in the code

pallas

legendary

Activity: 2716

Merit: 1094

Black Belt Developer

Quote from: Amph on June 16, 2016, 06:56:29 AM

Quote from: pallas on June 16, 2016, 05:47:32 AM

Honestly I don't think there is much to do on pascal as compared to maxwell, from the miner point of view.
I hope I'm proven wrong, but we should be happy with that 50% to 80% increase in hash/power ratio.
Excluding ether, of course (neoscrypt just needs to be tuned and lyra2re (v1) is dead).

80%? i didn't see any 80% in the 970 vs 1070, if it was the case, i would be more than happy, it would be a must buy for me

according to:

http://cryptomining-blog.com/7973-testing-the-nvidia-geforce-gtx-1080-founders-edition-for-crypto-mining/

1080 vs 970 is up to 120% faster (see lyra2v2 or nist5 for example).

1070 should be about 30% slower, so well over 80% faster than 970, in those cases.

Power usage similar, maybe even lower on 1070.

Amph

legendary

Activity: 3248

Merit: 1070

Quote from: pallas on June 16, 2016, 05:47:32 AM

Honestly I don't think there is much to do on pascal as compared to maxwell, from the miner point of view.
I hope I'm proven wrong, but we should be happy with that 50% to 80% increase in hash/power ratio.
Excluding ether, of course (neoscrypt just needs to be tuned and lyra2re (v1) is dead).

80%? i didn't see any 80% in the 970 vs 1070, if it was the case, i would be more than happy, it would be a must buy for me

vaulter

full member

Activity: 174

Merit: 100

Quote from: Amph on June 16, 2016, 05:25:47 AM

Quote from: bensam1231 on June 16, 2016, 05:12:29 AM

Quote from: Amph on June 16, 2016, 05:02:33 AM

Quote from: bensam1231 on June 16, 2016, 03:08:50 AM

AND Neoscrypt (half the performance of 970). Those are basically all the algos that are worth mining right now besides Lyra2. CMB did some testing, couple people posted it here. Considering there is only one algo for 1xxx series to mine, expect that one to start dipping down to the other shit algos like blake.

So you're essentially mining for one coin with the 1xxx series.

Not one to hate, but your math in this case is seriously lacking.

In other news, Nicehash really needs a auto-upload mode for their benchmark mode in their miner. Right now, the only place to go to look at GPUs is like here: http://62.212.74.86/~mining/list/index.php?brand=amd

That's a giant list of shit, mixed clocks, versions of sgminer (for instance), multiple GPUs in some cases(!?!?), and you can barely sort it let alone separate it. A bit of automation in this case would be super helpful. I was trying to find some data on the latest versions of sgminer, but it's a shit show.

you will never know what the best algo will be for the future, all i know is that i will never mine for one algo, EVER, this si like having an asic, which i hate to death

my math is right, and it is based on my consumption, but you are actually losing money by missing the 50% more hash that a 1070 can give to you right now over the 970 or over any other amd

also what you're talking about when saying "half the performance"? currently a 1070 can mine at 1MH on neoiscrypt which is 50% more than a 970, and this without any optimization for pascal

Lyra2 IS one algo. The other two, NeoS and Dagger, that actually earn money the 1XXX series does not perform well in, actually worse then a 970.

I agree, I'll never mine one algo either... And in this case the 1XXX series only can mine one algo at any sort of meaningful rate. As such, anyone buying one will only mine that algo and we will see equalization with the other algos except Dagger (Ethereum).

Where are you getting your results for the 1070 on Neoscrypt? According to cryptominingblog, the 1080 is doing 357khs on Neoscrypt, which makes sense as both 1XXX have a 256bit bus and are crippled on Dagger. Neoscrypt is also very memory intensive, although not as much as dagger.

http://cryptomining-blog.com/

I also agree you have to think long term, but right now you're going to lose money buying a 1xxx over other cards due to their craptastic hashrate in 2 of the 3 big algos, low availability, and crappy super hiked up prices.

https://bitcointalksearch.org/topic/m.15059281

i think they are using a worse version, this guy is getting 900 on neoscrypt, but bears in mind that the pascal architecture need a code rewrite to be able to reach full potential

it's not far stretched to think that there will be a 50% boost in the future, on all algo, we just need the coder...

Quote from: sp_ on June 16, 2016, 05:17:41 AM

Quote from: Amph on June 16, 2016, 05:07:48 AM

you can not compare like this, you optimized like no tomorrow your 980ti, but you didn't the same with the 1080, with optimization a 1080 can probably do 30% more hash than a 980ti, this is basing on the benchmark
not talking about intensity, here talking about rewrite the code for the new architecture that is pascal, which has yet to be unveiled

True. But I will not do the job.

well that's bad, if we collect some btc maybe you will change your mind? otherwise we need to ask to someone else

As for 1080 - it has gddr5x - and thats where problems comes for memory related algos
As for 1070 - in win 10 you have 650khs for neoscrypt
On win 8.1 i think i saw 900khs on neoscrypt - but i need to doublecheck - mind you in every algo it is around or beats 980ti with much less power consumption and seriously beating it in eth
Im on holidays right now and hope to test 480 when i come back and then we will see if how is the winner overall - the question with 480 - it will definitly wont ve able to beat 1070 in any algo - but how close it can get to it - both in perf and power - because we know the price

pallas

legendary

Activity: 2716

Merit: 1094

Black Belt Developer

Honestly I don't think there is much to do on pascal as compared to maxwell, from the miner point of view.
I hope I'm proven wrong, but we should be happy with that 50% to 80% increase in hash/power ratio.
Excluding ether, of course (neoscrypt just needs to be tuned and lyra2re (v1) is dead).

Amph

legendary

Activity: 3248

Merit: 1070

Quote from: bensam1231 on June 16, 2016, 05:12:29 AM

Quote from: Amph on June 16, 2016, 05:02:33 AM

Quote from: bensam1231 on June 16, 2016, 03:08:50 AM

AND Neoscrypt (half the performance of 970). Those are basically all the algos that are worth mining right now besides Lyra2. CMB did some testing, couple people posted it here. Considering there is only one algo for 1xxx series to mine, expect that one to start dipping down to the other shit algos like blake.

So you're essentially mining for one coin with the 1xxx series.

Not one to hate, but your math in this case is seriously lacking.

In other news, Nicehash really needs a auto-upload mode for their benchmark mode in their miner. Right now, the only place to go to look at GPUs is like here: http://62.212.74.86/~mining/list/index.php?brand=amd

That's a giant list of shit, mixed clocks, versions of sgminer (for instance), multiple GPUs in some cases(!?!?), and you can barely sort it let alone separate it. A bit of automation in this case would be super helpful. I was trying to find some data on the latest versions of sgminer, but it's a shit show.

you will never know what the best algo will be for the future, all i know is that i will never mine for one algo, EVER, this si like having an asic, which i hate to death

my math is right, and it is based on my consumption, but you are actually losing money by missing the 50% more hash that a 1070 can give to you right now over the 970 or over any other amd

also what you're talking about when saying "half the performance"? currently a 1070 can mine at 1MH on neoiscrypt which is 50% more than a 970, and this without any optimization for pascal

Lyra2 IS one algo. The other two, NeoS and Dagger, that actually earn money the 1XXX series does not perform well in, actually worse then a 970.

I agree, I'll never mine one algo either... And in this case the 1XXX series only can mine one algo at any sort of meaningful rate. As such, anyone buying one will only mine that algo and we will see equalization with the other algos except Dagger (Ethereum).

Where are you getting your results for the 1070 on Neoscrypt? According to cryptominingblog, the 1080 is doing 357khs on Neoscrypt, which makes sense as both 1XXX have a 256bit bus and are crippled on Dagger. Neoscrypt is also very memory intensive, although not as much as dagger.

http://cryptomining-blog.com/

I also agree you have to think long term, but right now you're going to lose money buying a 1xxx over other cards due to their craptastic hashrate in 2 of the 3 big algos, low availability, and crappy super hiked up prices.

https://bitcointalksearch.org/topic/m.15059281

i think they are using a worse version, this guy is getting 900 on neoscrypt, but bears in mind that the pascal architecture need a code rewrite to be able to reach full potential

it's not far stretched to think that there will be a 50% boost in the future, on all algo, we just need the coder...

Quote from: sp_ on June 16, 2016, 05:17:41 AM

Quote from: Amph on June 16, 2016, 05:07:48 AM

you can not compare like this, you optimized like no tomorrow your 980ti, but you didn't the same with the 1080, with optimization a 1080 can probably do 30% more hash than a 980ti, this is basing on the benchmark
not talking about intensity, here talking about rewrite the code for the new architecture that is pascal, which has yet to be unveiled

True. But I will not do the job.

well that's bad, if we collect some btc maybe you will change your mind? otherwise we need to ask to someone else

sp_

legendary

Activity: 2954

Merit: 1087

Team Black developer

Quote from: Amph on June 16, 2016, 05:07:48 AM

you can not compare like this, you optimized like no tomorrow your 980ti, but you didn't the same with the 1080, with optimization a 1080 can probably do 30% more hash than a 980ti, this is basing on the benchmark
not talking about intensity, here talking about rewrite the code for the new architecture that is pascal, which has yet to be unveiled

True. But I will not do the job.

bensam1231

legendary

Activity: 1764

Merit: 1024

Quote from: Amph on June 16, 2016, 05:02:33 AM

Quote from: bensam1231 on June 16, 2016, 03:08:50 AM

AND Neoscrypt (half the performance of 970). Those are basically all the algos that are worth mining right now besides Lyra2. CMB did some testing, couple people posted it here. Considering there is only one algo for 1xxx series to mine, expect that one to start dipping down to the other shit algos like blake.

So you're essentially mining for one coin with the 1xxx series.

Not one to hate, but your math in this case is seriously lacking.

In other news, Nicehash really needs a auto-upload mode for their benchmark mode in their miner. Right now, the only place to go to look at GPUs is like here: http://62.212.74.86/~mining/list/index.php?brand=amd

That's a giant list of shit, mixed clocks, versions of sgminer (for instance), multiple GPUs in some cases(!?!?), and you can barely sort it let alone separate it. A bit of automation in this case would be super helpful. I was trying to find some data on the latest versions of sgminer, but it's a shit show.

you will never know what the best algo will be for the future, all i know is that i will never mine for one algo, EVER, this si like having an asic, which i hate to death

my math is right, and it is based on my consumption, but you are actually losing money by missing the 50% more hash that a 1070 can give to you right now over the 970 or over any other amd

also what you're talking about when saying "half the performance"? currently a 1070 can mine at 1MH on neoiscrypt which is 50% more than a 970, and this without any optimization for pascal

Lyra2 IS one algo. The other two, NeoS and Dagger, that actually earn money the 1XXX series does not perform well in, actually worse then a 970.

I agree, I'll never mine one algo either... And in this case the 1XXX series only can mine one algo at any sort of meaningful rate. As such, anyone buying one will only mine that algo and we will see equalization with the other algos except Dagger (Ethereum).

Where are you getting your results for the 1070 on Neoscrypt? According to cryptominingblog, the 1080 is doing 357khs on Neoscrypt, which makes sense as both 1XXX have a 256bit bus and are crippled on Dagger. Neoscrypt is also very memory intensive, although not as much as dagger.

http://cryptomining-blog.com/

I also agree you have to think long term, but right now you're going to lose money buying a 1xxx over other cards due to their craptastic hashrate in 2 of the 3 big algos, low availability, and crappy super hiked up prices.

Amph

legendary

Activity: 3248

Merit: 1070

Quote from: sp_ on June 16, 2016, 05:05:47 AM

Quote from: Amph on June 16, 2016, 05:02:33 AM

Quote from: sp_ on June 16, 2016, 01:55:34 AM

Quote from: Amph on June 16, 2016, 12:44:55 AM

assuming you increased some algo to 10-20%, the same will happen for the 1080, so add 20% more on top of that hashrate...

This is not the case. I have a few that has tested the "public" private bins, and they are sometimes slower than the opensource.

My buyable private Spreadcoin #9 is up 0.6MHASH on the 1070 and 1MHASH on the 1080 compared to the opensource.

1070: 6000 -> 6600
1080: 7500 -> 8500

you still need to tweak it for the pascal, you can squeeze some more hash from it

I think playing with the -x parameter can give a boost.

i mean from a kernel point of view, pascal it's a new architecture, there should be a considerable boost, with the correct code

did you look at it already?

Quote from: sp_ on June 16, 2016, 04:17:38 AM

Quote from: giagge on June 16, 2016, 03:35:29 AM

THIS IS NOT MY CARD MSI gtx 1080 Gaming X on lyra2v2 NiceHashMiner.v1.5.1.0

43.6MHASH is good, but I can do 41 MHASH stable on my 980ti.

So the 1080 is only 6.34% faster than a pimped 980ti. (private kernal (sp-mod #7) with overclocking)
You can pick up a used 980ti for half the price of the 1080 on ebay...

The 1080 use around 180Watt and the 980ti use 250Watt.

you can not compare like this, you optimized like no tomorrow your 980ti, but you didn't the same with the 1080, with optimization a 1080 can probably do 30% more hash than a 980ti, this is basing on the benchmark

not talking about intensity, here talking about rewrite the code for the new architecture that is pascal, which has yet to be unveiled

sp_

legendary

Activity: 2954

Merit: 1087

Team Black developer

Quote from: Amph on June 16, 2016, 05:02:33 AM

Quote from: sp_ on June 16, 2016, 01:55:34 AM

Quote from: Amph on June 16, 2016, 12:44:55 AM

assuming you increased some algo to 10-20%, the same will happen for the 1080, so add 20% more on top of that hashrate...

This is not the case. I have a few that has tested the "public" private bins, and they are sometimes slower than the opensource.

My buyable private Spreadcoin #9 is up 0.6MHASH on the 1070 and 1MHASH on the 1080 compared to the opensource.

1070: 6000 -> 6600
1080: 7500 -> 8500

you still need to tweak it for the pascal, you can squeeze some more hash from it

I think playing with the -x parameter can give a boost.

Amph

legendary

Activity: 3248

Merit: 1070

Quote from: sp_ on June 16, 2016, 01:55:34 AM

Quote from: Amph on June 16, 2016, 12:44:55 AM

assuming you increased some algo to 10-20%, the same will happen for the 1080, so add 20% more on top of that hashrate...

This is not the case. I have a few that has tested the "public" private bins, and they are sometimes slower than the opensource.

My buyable private Spreadcoin #9 is up 0.6MHASH on the 1070 and 1MHASH on the 1080 compared to the opensource.

1070: 6000 -> 6600
1080: 7500 -> 8500

you still need to tweak it for the pascal, you can squeeze some more hash from it

Quote from: bensam1231 on June 16, 2016, 03:08:50 AM

AND Neoscrypt (half the performance of 970). Those are basically all the algos that are worth mining right now besides Lyra2. CMB did some testing, couple people posted it here. Considering there is only one algo for 1xxx series to mine, expect that one to start dipping down to the other shit algos like blake.

So you're essentially mining for one coin with the 1xxx series.

Not one to hate, but your math in this case is seriously lacking.

In other news, Nicehash really needs a auto-upload mode for their benchmark mode in their miner. Right now, the only place to go to look at GPUs is like here: http://62.212.74.86/~mining/list/index.php?brand=amd

That's a giant list of shit, mixed clocks, versions of sgminer (for instance), multiple GPUs in some cases(!?!?), and you can barely sort it let alone separate it. A bit of automation in this case would be super helpful. I was trying to find some data on the latest versions of sgminer, but it's a shit show.

you will never know what the best algo will be for the future, all i know is that i will never mine for one algo, EVER, this si like having an asic, which i hate to death

my math is right, and it is based on my consumption, but you are actually losing money by missing the 50% more hash that a 1070 can give to you right now over the 970 or over any other amd

also what you're talking about when saying "half the performance"? currently a 1070 can mine at 1MH on neoscrypt which is 50% more than a 970, and this without any optimization for pascal

sp_

legendary

Activity: 2954

Merit: 1087

Team Black developer

Quote from: pallas on June 16, 2016, 04:35:02 AM

the 980ti uses 250 if not overclocked, it will use 300 or more to reach 41 MH/s.
don't compare apples and oranges.

I need to double check, but I think the lyra2v2 is using less power than the other algos.
The 1080 is using 176 Watt in lyra2v2 vs 180 Watt in decred so you're probobly right. The 980ti is closer to 300 watt@41mhash

pallas

legendary

Activity: 2716

Merit: 1094

Black Belt Developer

Quote from: sp_ on June 16, 2016, 04:17:38 AM

Quote from: giagge on June 16, 2016, 03:35:29 AM

THIS IS NOT MY CARD MSI gtx 1080 Gaming X on lyra2v2 NiceHashMiner.v1.5.1.0

43.6MHASH is good, but I can do 41 MHASH stable on my 980ti.

So the 1080 is only 6.34% faster than a pimped 980ti. (private kernal (sp-mod #7) with overclocking)
You can pick up a used 980ti for half the price of the 1080 on ebay...

The 1080 use around 180Watt and the 980ti use 250Watt.

the 980ti uses 250 if not overclocked, it will use 300 or more to reach 41 MH/s.
don't compare apples and oranges.

sp_

legendary

Activity: 2954

Merit: 1087

Team Black developer

Quote from: giagge on June 16, 2016, 03:35:29 AM

THIS IS NOT MY CARD MSI gtx 1080 Gaming X on lyra2v2 NiceHashMiner.v1.5.1.0

43.6MHASH is good, but I can do 41 MHASH stable on my 980ti.

So the 1080 is only 6.34% faster than a pimped 980ti. (private kernal (sp-mod #7) with overclocking)
You can pick up a used 980ti for half the price of the 1080 on ebay...

The 1080 use around 180Watt and the 980ti use 250Watt.

sp_

legendary

Activity: 2954

Merit: 1087

Team Black developer

The point is that Etherum is worth $ 1,590,652,810, and million of dollars are given out the miners every day. (estimated $500 000 000 worth of GPU's mining)
Monacoin (lyra2v2 ) is worth $ 2,122,648
Vertcoin (lyra2v2) is worth $ 873,020

and

LyraBar
Verge
Unitus
MobileCash
Zoom

are worth pennies. So even if the 1080 is the fastest card to mine lyra2v2( shitcoins) , it is slow in etherum (This is where the money is right now).

giagge

legendary

Activity: 1134

Merit: 1001

Quote from: sp_ on June 09, 2016, 07:49:02 AM

Spmod #7 lyra2v2 results

standard clocks

980ti (1280 core clock) 35600 (g1 gaming)
970 (1366 core clock) 22600 (g1 gaming)
960 (1350 core clock) 14500 (msi gaming)
750ti (1275 core clock) 6700 (asus strix)

so 970+960 would do around 37100 on 1350mhz core

Both tsiv and djm34 had kernals with the shuffle instruction a few days after the launch. This nicehash kernal has a simple crypt pass to avoid kernal extraction and reverse engineering. So my bet is that it is a private kernal has been used for a long time in binary form before it was bought by nicehash and spread.

THIS IS NOT MY CARD MSI gtx 1080 Gaming X on lyra2v2 NiceHashMiner.v1.5.1.0

Anyone Know the result for AMD 390X ?

Thanks .

bensam1231

legendary

Activity: 1764

Merit: 1024

Quote from: Amph on June 15, 2016, 01:45:31 PM

Quote from: joblo on June 15, 2016, 11:14:25 AM

Quote from: Amph on June 15, 2016, 10:33:33 AM

Quote from: sp_ on June 15, 2016, 06:49:56 AM

I'm not sure I will do the pascal sp-mod. I think I will move to AMD and 512bit bus. 1070/1080 is not attractive when it comes to price/vs hashrate.

unless you are talking about new amd, current amd line up can not compete on any algo minus etheruem with 1070, not to mention that the 1070 can be improved further and reach a nice 2x hashrate on any algo comapred to 970, eth excluded,

but who care eth is not the only profitable coin...

Current Pascal price/performance seems to be inline with Maxwell for most algos so there's not much incentive
to upgrade yet.

actually no, a 1070 consume only 100w or less and have 50% more hash in any algo(ok no etheruem but i don't care i don't mine for one coin) versus a 970, and this without any tweak i'm sure it can be pushed to 100%

a easy math is telling me that mining now with a 970 instead of a 1070, will only make you lose 40-50 euro a month, for each gpu, this counting the 50% more performance and less consumption, even with cheap energy still around 30 euro are lost each month

so buying a 1070 now for 500 euro, or buying it in 4 months for 380 euro is the same thing....

AND Neoscrypt (half the performance of 970). Those are basically all the algos that are worth mining right now besides Lyra2. CMB did some testing, couple people posted it here. Considering there is only one algo for 1xxx series to mine, expect that one to start dipping down to the other shit algos like blake.

So you're essentially mining for one coin with the 1xxx series.

Not one to hate, but your math in this case is seriously lacking.

In other news, Nicehash really needs a auto-upload mode for their benchmark mode in their miner. Right now, the only place to go to look at GPUs is like here: http://62.212.74.86/~mining/list/index.php?brand=amd

That's a giant list of shit, mixed clocks, versions of sgminer (for instance), multiple GPUs in some cases(!?!?), and you can barely sort it let alone separate it. A bit of automation in this case would be super helpful. I was trying to find some data on the latest versions of sgminer, but it's a shit show.

Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 647. (Read 2347641 times)