CCminer(SP-MOD) Modded GPU kernels.

[induktor]

With GPU Pascal ( gtx 1070 - 1080 ) ccminer work on windows 10 in future ? .

Already works fine with win 10
I use 80 build (but not cuda 8 - that gave poor performance results) with following results - 32mhs quark and 16mhs lirare2v for 1080 and 24mhs and 14mhs for 1070

So ~23%-26% increase from 970 to 1070 assuming OC. That's pretty poor so far.

We still have a piece missing, power consumption vs the 970  or kh/power ratio.

right now in ethereum my rig has 145KH/W

if hash/power ratio worth it i will consider buy some 1070

[bathrobehero]

With GPU Pascal ( gtx 1070 - 1080 ) ccminer work on windows 10 in future ? .

Already works fine with win 10
I use 80 build (but not cuda 8 - that gave poor performance results) with following results - 32mhs quark and 16mhs lirare2v for 1080 and 24mhs and 14mhs for 1070

So ~23%-26% increase from 970 to 1070 assuming OC. That's pretty poor so far.

We still have a piece missing, power consumption vs the 970  or kh/power ratio.

right now in ethereum my rig has 145KH/W

if hash/power ratio worth it i will consider buy some 1070

Though we're only looking at reference cards at this point and not the different factory OC flavors, it still looks like the increase in efficiency will not warrant replacing 9xx series cards to 1070s with the expected prices anytime soon.
And the reference 970s and 1070s have almost the same power consumption so factory OC cards will likely scale similarily. But there are a lot of other factors so who knows.

[Longsnowsm]

@bathrobehero So far your observations appear exactly what I have been thinking as I have seen some of the numbers showing up for the new cards.  Unless optimizations of the miners can somehow grind out some more hashpower it appears that there isn't much advantage if any so far with the new generation of cards.  Of course time and optimized code will tell the tale.  So far it has been kind of disappointing.

[hmage]

https://gist.github.com/epixoip/6ee29d5d626bd8dfe671a2d8f188b77b

Hashcat benchmark results on 1080.

[vaulter]

With GPU Pascal ( gtx 1070 - 1080 ) ccminer work on windows 10 in future ? .

Already works fine with win 10
I use 80 build (but not cuda 8 - that gave poor performance results) with following results - 32mhs quark and 16mhs lirare2v for 1080 and 24mhs and 14mhs for 1070

So ~23%-26% increase from 970 to 1070 assuming OC. That's pretty poor so far.

We still have a piece missing, power consumption vs the 970  or kh/power ratio.

right now in ethereum my rig has 145KH/W

if hash/power ratio worth it i will consider buy some 1070

I didnt try 1070 on ethereum but as fro the 1080 its 120w at 23mhs on eth
as for 1070 I think it will be the same or less for eth in terms of power and i think it will be around 22mhs for 1070
as for quark its 130w at 24mhs lira was 120w at 14mhs

[vaulter]

Decred sp-mod #9 does 3.3 ghash on the 980ti. 2.7 for the 1080 is low

yes but with what consumption? 1080 can do 3300 with oc too, but consuming only 180w, the 980ti will consume like 250 or more

Sources?

Hypothesizing is fine, but mark it as such. Don't state something as fact unless you have sources for it.

mmmmm 1080 nice


is the above run at stock?

wondering what the power draw is ? if you don't have a watt meter, what is the percentage of TDP that the card is averaging ?

it has double the hashrate of a 970 with the wattage of one for decred, which is really good, but it cost 3 times more, not worth it for the moment, also decred is bad to mine versus eth

Decred is so bad right now that it makes less profit than mining lyra2v2 with a 750Ti or quark with a GTX970, that bad  

Yup, still getting hammered by the Claymore AMD miner... If only Nvidia had something like that.

Real test is going to be what a 1080 gets in Eth with the fucked up memory issues since the 1080 uses a different type of memory and bus.

https://bitcointalk.org/index.php?topic=1464999.msg14885194#msg14885194

this guy is an insider, he tested both 1080 and 1070, they perform better than any gpu right now on any algo minus ethereum

and this without counting optimization and proper drivers

Not sure I believe 'all at 180w' unless they have a watt meter. A lot of people thought 970s wouldn't go over TDP or close to it. Doesn't show clock speeds or TDP either.

Currently I hear the 1080 throttles in games too at stock speeds, doesn't mention whether or not the fan profile was increased or if that happens under prolonged mining.

970 it's actually under tdp, those that have it above, don't know what power limit is and how to use it on msi afterburner, and or they are buying 8+6 pin gpu with again PL at 100%, which push the wattage to unnecessary level

i'm now at around 125w for 19.5MH with a 970

There isn't a correlation between power connectors and wattage besides on 750tis which generally don't come with a external power connector. I have some 6x2 EVGAs that use less wattage then Asus cards with a 6+2 connector. It's all based on how they're binned and that has nothing to do with power connectors.

You can't trust reported numbers on the PC itself, you have to use a watt meter. They've been proven wrong time and time again.

i have used both, and wattmeter display the same thing, unless now you want to say that even my wattmeter is broken

actually the wattmeter will report only the full consumption, including the whole system which will drain an additional 50-70w, so it's not that good to measure with the wattmeter the correct wattage of your gpu

result are the same 135w at 20.5 mega and 125w at 19.5 mega for each 970, the smi thing got boosted, it's pretty accurate now

What are your exact settings for 970 (including afterburner and cuda intensity for miner)?

[Amph]

Decred sp-mod #9 does 3.3 ghash on the 980ti. 2.7 for the 1080 is low

yes but with what consumption? 1080 can do 3300 with oc too, but consuming only 180w, the 980ti will consume like 250 or more

Sources?

Hypothesizing is fine, but mark it as such. Don't state something as fact unless you have sources for it.

mmmmm 1080 nice


is the above run at stock?

wondering what the power draw is ? if you don't have a watt meter, what is the percentage of TDP that the card is averaging ?

it has double the hashrate of a 970 with the wattage of one for decred, which is really good, but it cost 3 times more, not worth it for the moment, also decred is bad to mine versus eth

Decred is so bad right now that it makes less profit than mining lyra2v2 with a 750Ti or quark with a GTX970, that bad  

Yup, still getting hammered by the Claymore AMD miner... If only Nvidia had something like that.

Real test is going to be what a 1080 gets in Eth with the fucked up memory issues since the 1080 uses a different type of memory and bus.

https://bitcointalk.org/index.php?topic=1464999.msg14885194#msg14885194

this guy is an insider, he tested both 1080 and 1070, they perform better than any gpu right now on any algo minus ethereum

and this without counting optimization and proper drivers

Not sure I believe 'all at 180w' unless they have a watt meter. A lot of people thought 970s wouldn't go over TDP or close to it. Doesn't show clock speeds or TDP either.

Currently I hear the 1080 throttles in games too at stock speeds, doesn't mention whether or not the fan profile was increased or if that happens under prolonged mining.

970 it's actually under tdp, those that have it above, don't know what power limit is and how to use it on msi afterburner, and or they are buying 8+6 pin gpu with again PL at 100%, which push the wattage to unnecessary level

i'm now at around 125w for 19.5MH with a 970

There isn't a correlation between power connectors and wattage besides on 750tis which generally don't come with a external power connector. I have some 6x2 EVGAs that use less wattage then Asus cards with a 6+2 connector. It's all based on how they're binned and that has nothing to do with power connectors.

You can't trust reported numbers on the PC itself, you have to use a watt meter. They've been proven wrong time and time again.

i have used both, and wattmeter display the same thing, unless now you want to say that even my wattmeter is broken

actually the wattmeter will report only the full consumption, including the whole system which will drain an additional 50-70w, so it's not that good to measure with the wattmeter the correct wattage of your gpu

result are the same 135w at 20.5 mega and 125w at 19.5 mega for each 970, the smi thing got boosted, it's pretty accurate now

If you use the same system, system load doesn't matter between different cards you test.

We're talking about card TDP, it's not always accurate. More then one person has confirmed that in this thread, myself included.

There isn't a correlation between power connectors and wattage besides on 750tis which generally don't come with a external power connector. I have some 6x2 EVGAs that use less wattage then Asus cards with a 6+2 connector. It's all based on how they're binned and that has nothing to do with power connectors.

You can't trust reported numbers on the PC itself, you have to use a watt meter. They've been proven wrong time and time again.

i have used both, and wattmeter display the same thing, unless now you want to say that even my wattmeter is broken

actually the wattmeter will report only the full consumption, including the whole system which will drain an additional 50-70w, so it's not that good to measure with the wattmeter the correct wattage of your gpu

result are the same 135w at 20.5 mega and 125w at 19.5 mega for each 970, the smi thing got boosted, it's pretty accurate now

Hi
I agree with bensam1231 on this one, you can only trust an AC wattmeter, sometimes GPU power reading is close, but a LOT of times is WAY OFF reality

an AC power meter measures the real thing, that is, what you pay every month!, and that's what matters the most. how much $$$ costs you to have your farm in operation.

the very first thing i do when i assemble a miner, i put everything except the GPUs, and power the rig, wait until the os loads, and with the system running and fully OS loaded, I measure how much does it use, in my case is 16W right now. (that includes the LCD  and power bar), then I put all cards in.

With this readings I know how much every card uses  (total power - 18W (mobo), result /3  and i know exactly how much each card is using)

this is not accurate either, because when you run six gpu on a motherboard it consume more in comparison with a empty motherboard, so from 18w like in your case, it may increase to 30-50

the same for cpu, anyway like i said, i did measure it with the wattmeter, and it report the same plus those 50-70w for other component, i'm running additional fan on my system, so it's there, smi is telling the true

That doesn't make sense. Only thing that would cause more load besides the CPU running at full frequency is the cards. You could turn off power saving features like cool'n'quiet or speedstep if you want to be really anal. Adding more cards shouldn't increase system load outside of the cards themselves. Something needs to draw the power.

actually it make sense, when you use more gpu, the motherboard need additional current, so there is a bit of more consumption for that, and the cpu will work a bit harder, we are not talking about crazy vlaue, but those additional 10-20w are there

there was even an article in cryptoblog saying that

for smi, you need to check the last version, it's pretty accurate, you can do it your test, even another user was surprised when he tried again, recently, and anyway like i've said i've tested with the wattmeter to be sure, and they are around the same

also gpu efficiency matter here, i have the 1500i, which has one of the best efficiency out there, so overall the consumption is lower by few watt because of this

there are many variable that simply can not be ignored, it's not so straightforward like many think

What are your exact settings for 970 (including afterburner and cuda intensity for miner)?

55% tdp, but the first gpu crash it seems, it crashed today, intensity is at default, everyhting at default, minuc oc at +150/+150 on afterburner

[tbearhere]

I ordered some of these devices :

http://fr.aliexpress.com/item/AC-80-260V-LCD-Digital-100A-Volt-Watt-Power-Meter-Ammeter-Voltmeter-CT-TTL-free-shipping/32620971490.html

to compare real power while mining... i know there is +10% for the PSU but it could be nice to have a tool to monitor the real price too :p

Epsylon3  What exactly does it do?  Thx

[tbearhere]

With GPU Pascal ( gtx 1070 - 1080 ) ccminer work on windows 10 in future ? .

Already works fine with win 10
I use 80 build (but not cuda 8 - that gave poor performance results) with following results - 32mhs quark and 16mhs lirare2v for 1080 and 24mhs and 14mhs for 1070

How does it do on neoscrypt ?   Thx

[Epsylon3]

I ordered some of these devices :

http://fr.aliexpress.com/item/AC-80-260V-LCD-Digital-100A-Volt-Watt-Power-Meter-Ammeter-Voltmeter-CT-TTL-free-shipping/32620971490.html

to compare real power while mining... i know there is +10% for the PSU but it could be nice to have a tool to monitor the real price too :p

NIIIICEEE!!  I want one!, will order one, any idea if it works under linux?, i guess is just a FTL USB to 232 chip right?
would be a nice addition to every miner, i can even stream the data to the miner´s lcd display!

I was thinking about design something like that, but for that price, i´ll just buy it!, the current transformer alone here costs 15 dollars hehe

there are different models, bulks one, some cheaper without cable and some higher with a case... its self powered else but i will put that in a custom case like an UPS with several outputs, will try to link that to a wemos module (3€) for wireless monitoring ^^

[Liquid71]

Hi tbearhere
question
why do you keep the 750ti there, is insignificant compared to the other ones, and it uses a different compute architecture.
wouldn't be better get rid of it, at least until you solve the issue with the second GTX970 ? which is three times more powerful?

just a thought, 80% of my cards are 750Ti and GTX970, I never ever mixed them, i already have way too many problems with risers, lack of lanes, power limitations, to have to deal with different compute architectures in the same machine hehe.
(EDIT) I just saw the image you posted again, you could disable onboard VGA too, and use the vga on the 16x slot as primary adapter,less driver issues as well.

I'll have to try that...thx  I use 1x pcie power risers.
What do you mean by primary adapter with power risers?
And how do I get to the bios through team-viewer?  Thx

Sorry, the PSU i mention in my previous post was SENTEY SDP850SS (dunno why endup written sentry hehe, automatic corrector maybe)
Primary adapter I mean if you connect your monitor to the rig, the card connected to the first 16x connector, should be the primary adapter (ie connected to the monitor) if you dont conect a monitor (BETTER!) and access remotely, it doesn't matter  

You don't get to the bios from team-viewer, that can only be done in servers, with iLO2 (HP) or DRAC (Dell) (fantastic feature btw!, but only reserved to the best of the best   )

650W, depends on the brand, could be a little small for 6x 750Ti, I will be more confortable with 750W 80plus. , i have that setup with a single sentey 850W PSU, better safe than sorry   but if that PSU is good, and you keep TDP low, it should be enough, my rig uses 300W at the wall mining quark with 6x 750Ti so, 650W PSU, if can really take 650W, would be enough. for example a GX650 Coolermaster is a nice little beast, it will leaf is ass of with only 300W of power consumption hehe.

good that you have a power meter!!

I have/had two 6x 750ti rigs with EVGA 650watt no problems for two years. I also had two 750 watt evga and they were pretty much the same depending on the algo I was mining. Only reason I got the 650's and 750's was rebates or I would have gone with 750. Anyway there really wasn't any performance or efficiency differences, 750 was more efficient on algos like neoscrypt but otherwise they both worked fine.  Also when I started mining Litecoin I maxed out a Corsair RM1000 it was pulling 1150 from the wall and never had an issue. It just reduced efficiency. 80% load should be fine on an ATX PSU as long as it's a reputable brand.

Since most consumer PSU are just rebranded, this is a good site to find the actual manufacturer:
http://www.realhardtechx.com/index_archivos/FSP.htm
FSP supposdely has the lowest RMA in the industry according to that. The EVGA PSU I have are FSP and I have a 1kw and 1200 watt AURUM series PSU and both have been solid. Only complaint is cables are a little shorter than I'd like. Corsair is obviously a quality brand but sometimes you can get nearly identical seasonic cheaper, using that realhardtechx site you can find the actual OEM and then price comparison and reviews ect. It's helped me a lot, along with Newegg which lists if the PSU is active PFC which I can't find on most etailer websites.

Good thing about investing in a PSU is they don't become obsolete when the next gen comes out lol  

[Liquid71]

actually it make sense, when you use more gpu, the motherboard need additional current, so there is a bit of more consumption for that, and the cpu will work a bit harder, we are not talking about crazy vlaue, but those additional 10-20w are there

there was even an article in cryptoblog saying that

for smi, you need to check the last version, it's pretty accurate, you can do it your test, even another user was surprised when he tried again, recently, and anyway like i've said i've tested with the wattmeter to be sure, and they are around the same

also gpu efficiency matter here, i have the 1500i, which has one of the best efficiency out there, so overall the consumption is lower by few watt because of this

there are many variable that simply can not be ignored, it's not so straightforward like many think



55% tdp, but the first gpu crash it seems, it crashed today, intensity is at default, everyhting at default, minuc oc at +150/+150 on afterburner

The extra current is for the additional GPU. Even if  you want to credit system overhead not the GPU for the extra wattage the power is being used for the additional GPU so the 10-20 watts are needed for the GPU.
Also PSU efficiency matters, but the efficiency changes with the load. Most PSU have a chart showing peak efficiency. If you have too low of a load on a PSU it can lower efficiency just as much as too high of a load.  The 1500i has a chart

So it's the most efficient IF you use 55-60% load. That's pretty much the same with most 80 plus rated PSU

I've used dual PSU on one MB, not for efficiency but necessity in the past, but I don't see why it wouldn't increase efficiency by balancing loads on two PSU
http://www.ebay.com/itm/NEW-Dual-PSU-Power-Supply-24-Pin-ATX-Motherboard-Adapter-Cable-Mining-Litecoin-/201587699757
I don't know what load you're using Amph, wasn't directing that to you just throwing my two satoshi's in on PSU efficiency for mining. SP's thread has become a mining tips thread 

[Amph]

actually it make sense, when you use more gpu, the motherboard need additional current, so there is a bit of more consumption for that, and the cpu will work a bit harder, we are not talking about crazy vlaue, but those additional 10-20w are there

there was even an article in cryptoblog saying that

for smi, you need to check the last version, it's pretty accurate, you can do it your test, even another user was surprised when he tried again, recently, and anyway like i've said i've tested with the wattmeter to be sure, and they are around the same

also gpu efficiency matter here, i have the 1500i, which has one of the best efficiency out there, so overall the consumption is lower by few watt because of this

there are many variable that simply can not be ignored, it's not so straightforward like many think



55% tdp, but the first gpu crash it seems, it crashed today, intensity is at default, everyhting at default, minuc oc at +150/+150 on afterburner

The extra current is for the additional GPU. Even if  you want to credit system overhead not the GPU for the extra wattage the power is being used for the additional GPU so the 10-20 watts are needed for the GPU.
Also PSU efficiency matters, but the efficiency changes with the load. Most PSU have a chart showing peak efficiency. If you have too low of a load on a PSU it can lower efficiency just as much as too high of a load.  The 1500i has a chart

So it's the most efficient IF you use 55-60% load. That's pretty much the same with most 80 plus rated PSU

I've used dual PSU on one MB, not for efficiency but necessity in the past, but I don't see why it wouldn't increase efficiency by balancing loads on two PSU
http://www.ebay.com/itm/NEW-Dual-PSU-Power-Supply-24-Pin-ATX-Motherboard-Adapter-Cable-Mining-Litecoin-/201587699757
I don't know what load you're using Amph, wasn't directing that to you just throwing my two satoshi's in on PSU efficiency for mining. SP's thread has become a mining tips thread  

i'm using exatly 50% on my psu, so it has a very good efficiency at that level, i've posted a screen with my righ and wattmeter showing 773w for 5 970, in another topic

i agree on the rest the overload is pretty low, but is still there, we are talking about a slim difference here anyway, 20w more is still something if you are going to show the correct wattage of your gpu

[tbearhere]

actually it make sense, when you use more gpu, the motherboard need additional current, so there is a bit of more consumption for that, and the cpu will work a bit harder, we are not talking about crazy vlaue, but those additional 10-20w are there

there was even an article in cryptoblog saying that

for smi, you need to check the last version, it's pretty accurate, you can do it your test, even another user was surprised when he tried again, recently, and anyway like i've said i've tested with the wattmeter to be sure, and they are around the same

also gpu efficiency matter here, i have the 1500i, which has one of the best efficiency out there, so overall the consumption is lower by few watt because of this

there are many variable that simply can not be ignored, it's not so straightforward like many think



55% tdp, but the first gpu crash it seems, it crashed today, intensity is at default, everyhting at default, minuc oc at +150/+150 on afterburner

The extra current is for the additional GPU. Even if  you want to credit system overhead not the GPU for the extra wattage the power is being used for the additional GPU so the 10-20 watts are needed for the GPU.
Also PSU efficiency matters, but the efficiency changes with the load. Most PSU have a chart showing peak efficiency. If you have too low of a load on a PSU it can lower efficiency just as much as too high of a load.  The 1500i has a chart

So it's the most efficient IF you use 55-60% load. That's pretty much the same with most 80 plus rated PSU

I've used dual PSU on one MB, not for efficiency but necessity in the past, but I don't see why it wouldn't increase efficiency by balancing loads on two PSU
http://www.ebay.com/itm/NEW-Dual-PSU-Power-Supply-24-Pin-ATX-Motherboard-Adapter-Cable-Mining-Litecoin-/201587699757
I don't know what load you're using Amph, wasn't directing that to you just throwing my two satoshi's in on PSU efficiency for mining. SP's thread has become a mining tips thread 

Yes a mining tips thread  its related to the problems I am having with ccminer and psu ect.

[WigitGetIt]

Just want to make miners aware. I know there are not to many X13 coins out there.

New X13 Coin launching at 12:00 (Noon) Eastern Today.

https://bitcointalk.org/index.php?topic=1461341.0

[sp_]

Cuda 8.0RC is out and it produce shit hashrates for the maxwell cards as expected. Anyone want to share their results?

[Velgelm]

no problem SP!
tpruvot 1.6.7 cuda 8.0 build
Results :
970+960 Stock clock:
blake 2370
Blake2s 1700
blakecoin 4120
bmw 1077
c11 11.53
decred 2271
deep 32.80
dmd-gr 35.12
fresh 13.66
fugue256 404
groestl 35.35
heavy 42.16
jackpot 20.50
luffa 262
lyra2 2600
lyra2v2 14.33
mjollnir 42.21
myr-gr 55.50
neoscrypt 515
nist5 38.60
penta 91.50
quark 20.45
qubit 17.25
sib 2568
scrypt -jane 558.17
skein 349
skein2 235
s3 33.28
x11 11.36
x13 9055
x15 7900
x17 7900
vanilla 3964
whirlcoin 21.72
whirlpool 21.56
zr5 4725

[sp_]

Could you run the same test with the cuda 6.5 and 7.5 build please?

[Stein]

no problem SP!
tpruvot 1.6.7 cuda 8.0 build
Results :/snip

Did you just run each algo one after another manually to get those results ? Wow

[sp_]

tpruvot, has made a command to benchmark all the algos. I don't have it in my fork. (1.5.80)