@wolf0
I think this community needs you, and some might pay you also.
https://bitcointalksearch.org/topic/m.11362643
Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 1093. (Read 2347641 times)
It is also possible to mod the kernals so that they use less power...
Look at my whirlpoolx mod. The Threads per block was reduced from 1024 to 512, and instead I computed more than one hash per thread.
When I reduced the number of threads, I got a lower power consumption and a higher hashrate...
doing that with sgminer is not trivial.Look at my whirlpoolx mod. The Threads per block was reduced from 1024 to 512, and instead I computed more than one hash per thread.
When I reduced the number of threads, I got a lower power consumption and a higher hashrate...
maybe with a custom miner ;-)
My Whirlpoolx implementation runs at around 0.5watt per MHASH on the 750ti. (40Watt/80MHASH)
If smolen's smelter miner does 600MHASH with 300watt this meens that my slow AES implementation is as fast if you measure POWER/HASHRATE
It is also possible to mod the kernals so that they use less power...
Look at my whirlpoolx mod. The Threads per block was reduced from 1024 to 512, and instead I computed more than one hash per thread.
When I reduced the number of threads, I got a lower power consumption and a higher hashrate...
Look at my whirlpoolx mod. The Threads per block was reduced from 1024 to 512, and instead I computed more than one hash per thread.
When I reduced the number of threads, I got a lower power consumption and a higher hashrate...
doing that with sgminer is not trivial.
maybe with a custom miner ;-)
I'm working on it!
wolf - pm'd you ...
#crysx
It is also possible to mod the kernals so that they use less power...
Look at my whirlpoolx mod. The Threads per block was reduced from 1024 to 512, and instead I computed more than one hash per thread.
When I reduced the number of threads, I got a lower power consumption and a higher hashrate...
Look at my whirlpoolx mod. The Threads per block was reduced from 1024 to 512, and instead I computed more than one hash per thread.
When I reduced the number of threads, I got a lower power consumption and a higher hashrate...
doing that with sgminer is not trivial.
maybe with a custom miner ;-)
It is also possible to mod the kernals so that they use less power...
Look at my whirlpoolx mod. The Threads per block was reduced from 1024 to 512, and instead I computed more than one hash per thread.
When I reduced the number of threads, I got a lower power consumption and a higher hashrate...
Look at my whirlpoolx mod. The Threads per block was reduced from 1024 to 512, and instead I computed more than one hash per thread.
When I reduced the number of threads, I got a lower power consumption and a higher hashrate...
that would be great ...
would it be possible to have fine grained control of gpuclock and memclock ( like that in sgminer ) ? ...
#crysx
would it be possible to have fine grained control of gpuclock and memclock ( like that in sgminer ) ? ...
#crysx
Yes. This is the plan. --gpu-memspeed and --gpu-memclock. The NVIDIA API supports this, just need to code it. But the miner will need elevated rights..
The gigabyte cards you have with a 6 pins adapter will run x11 up to 3,4 MHASH with overclocking. but the powerusage will increase around 15W
that would be fantastic sp ...
there are only about 16 of the powered ones ... the rest are non-powered ...
but if the control is within the commandline - that would give us such a massive boost ...
tanx again ...
reading the pm now ...
#crysx
Why the new account? Did your old one finally get banned?
No, the password was compromised, but it's back in my hands now. Lots of random posts on it so I just use my new one.
hehehe...Guys...too funny, but true.
Now that ccminer is supposed to report a little truer hashrate, what kind of speed is everyone getting for a single 750Ti mining quark? Not that important...just curious. I show around 6.3+Mgh miner side. Also, I'm using MinerControl (finally) and want to know if the version of ccminer that comes with it has been changed just for that program or is it 'stock'?
Thank you
GREAT job sp and friends!!
Now that ccminer is supposed to report a little truer hashrate, what kind of speed is everyone getting for a single 750Ti mining quark? Not that important...just curious. I show around 6.3+Mgh miner side. Also, I'm using MinerControl (finally) and want to know if the version of ccminer that comes with it has been changed just for that program or is it 'stock'?
Thank you
GREAT job sp and friends!!
Damn. With 280X doing 11.4MH/s on Quark, it's not quite good enough. Need to up that.
Yeah and at 250w. I should've switched to maxwell when it first came out, but no one knew bitcoin was going to bottom out.
these devs are doing awesome jobs - and if they help us mine better - they deserve to do deals ...
You missed the part where some of them sold custom private builds behind the scenes (last fall) to big miners without telling anyone they were doing that for a few months and made everyone with pub miners obsolete, forcing them out of the market. I suppose sucking up is a good way of getting those out of them though (not talking about sp btw). It's good to see them practicing more professionally now days.
Also mining is like a pie, mining with a better kernel doesn't mean more income if everyone is using the same miner. When everyone is on the same build, you're mining exactly the same amount as someone else. When a build is updated with higher hash speeds, everyone upgrades and you get 'more coins' for a little bit until everyone gets on the same build again. So you essentially take coins from someone with a crappier kernel (all else being equal). You can't 'get ahead' unless you're using a private build that isn't distributed to everyone else, which has happened in the past as I mentioned.
The poor density of the 750 Ti (having to buy more mobo/cpu/memory) is probably being offset by having to spend on a bigger PSU on a 970 rig. Unless you use multiple power supplies per rig but I'm not a big fan of that.
Yeah, I agree.
I am currently running two PSU's and I just ordered a bigger (850W) PSU to replace them both in my machine. It should be more efficient and stable, as the two ~500W PSU's I have are really struggling to provide enough stable current on the +12V rails.
I determined that they are lacking because the newer, faster versions (V44 and up) of sp_mod-ccminer really tax the +12V rails of these weaker, cheaper PSU's.
The 850W PSU that I ordered can push 70 Amps to a single +12V rail, so I hope to be pleased with its performance and stability.
Yeah, 750tis take a lot of space, even if they're slightly better, which is why I never bought any. I'm currently using 970s. 980s are just a waste of money as they're too expensive for the little bit extra performance they offers (much like the 290 vs 290x back in the day).
"stock clocks" varies from one brand/version of card to another.
But you can always overclock your cards.
I want to make it possible to overclock over the commandline as a parameter.
The 750ti's has sweat spots gpuspeed/memspead that will improve the hashrate.
Is this also true for the 970 and memory speed? I know AMD cards had a good spot for memory with scrypt, but it didn't matter so much for other algos. Also does efficiency go up or down by moving to that spot?
Curiously what happens to 970 or 750 power efficiency when you OC them a bit? Do they lose a lot of efficiency or does it only go out of spec with additional voltage?
absolutely amazing bensam ...
only a twat like you would say im sucking up ...
i have paid for ALL the private miners that i have - ALL of them ...
i used to code when i was in uni - and that is something i wish i continued ...
however - an imbecile is an imbecile is an imbecile ... so no matter WHAT happens - you will always find a way to poke and prod ... so on the 'bullshit twat' list you go ... no more garbage from you - thank God ...
these devs ( INCLUDING sp ) are ALL trying to make money - ALL of them ... this is why the devs ( INCLUDING sp ) SELL their private miners to people who MINE - not just farms - who are WILLING to pay the price for the miners ...
your drivel is nothing more than an attack on those that are actually trying to make a difference AND some money along the way - there is no 'sucking up' in saying what i have - you imbecile ( yes - look up that word for the exact definition OR look in the mirror - your choice ) ...
you cry and cry and cry - and yet have almost nothing constructive to say ...
the day you actually stop your drivel - you may just have something worth listening to ...
sucking up! ... pffft ... had i the cashflow - i would EMPLOY these amazing devs on salaries someone like you could only dream of ...
and you would STILL bitch about them ...
im done with you - and your bullshit ...
and im sure im not the only one ...
#crysx
that would be great ...
would it be possible to have fine grained control of gpuclock and memclock ( like that in sgminer ) ? ...
#crysx
would it be possible to have fine grained control of gpuclock and memclock ( like that in sgminer ) ? ...
#crysx
Yes. This is the plan. --gpu-memspeed and --gpu-memclock. The NVIDIA API supports this, just need to code it. But the miner will need elevated rights..
The gigabyte cards you have with a 6 pins adapter will run x11 up to 3,4 MHASH with overclocking. but the powerusage will increase around 15W
"stock clocks" varies from one brand/version of card to another.
But you can always overclock your cards.
I want to make it possible to overclock over the commandline as a parameter.
The 750ti's has sweat spots gpuspeed/memspead that will improve the hashrate.
that would be great ...
would it be possible to have fine grained control of gpuclock and memclock ( like that in sgminer ) ? ...
#crysx
Why the new account? Did your old one finally get banned?
No, the password was compromised, but it's back in my hands now. Lots of random posts on it so I just use my new one.
hehehe...Guys...too funny, but true.
Now that ccminer is supposed to report a little truer hashrate, what kind of speed is everyone getting for a single 750Ti mining quark? Not that important...just curious. I show around 6.3+Mgh miner side. Also, I'm using MinerControl (finally) and want to know if the version of ccminer that comes with it has been changed just for that program or is it 'stock'?
Thank you
GREAT job sp and friends!!
Now that ccminer is supposed to report a little truer hashrate, what kind of speed is everyone getting for a single 750Ti mining quark? Not that important...just curious. I show around 6.3+Mgh miner side. Also, I'm using MinerControl (finally) and want to know if the version of ccminer that comes with it has been changed just for that program or is it 'stock'?
Thank you
GREAT job sp and friends!!
Damn. With 280X doing 11.4MH/s on Quark, it's not quite good enough. Need to up that.
Yeah and at 250w. I should've switched to maxwell when it first came out, but no one knew bitcoin was going to bottom out.
these devs are doing awesome jobs - and if they help us mine better - they deserve to do deals ...
You missed the part where some of them sold custom private builds behind the scenes (last fall) to big miners without telling anyone they were doing that for a few months and made everyone with pub miners obsolete, forcing them out of the market. I suppose sucking up is a good way of getting those out of them though (not talking about sp btw). It's good to see them practicing more professionally now days.
Also mining is like a pie, mining with a better kernel doesn't mean more income if everyone is using the same miner. When everyone is on the same build, you're mining exactly the same amount as someone else. When a build is updated with higher hash speeds, everyone upgrades and you get 'more coins' for a little bit until everyone gets on the same build again. So you essentially take coins from someone with a crappier kernel (all else being equal). You can't 'get ahead' unless you're using a private build that isn't distributed to everyone else, which has happened in the past as I mentioned.
The poor density of the 750 Ti (having to buy more mobo/cpu/memory) is probably being offset by having to spend on a bigger PSU on a 970 rig. Unless you use multiple power supplies per rig but I'm not a big fan of that.
Yeah, I agree.
I am currently running two PSU's and I just ordered a bigger (850W) PSU to replace them both in my machine. It should be more efficient and stable, as the two ~500W PSU's I have are really struggling to provide enough stable current on the +12V rails.
I determined that they are lacking because the newer, faster versions (V44 and up) of sp_mod-ccminer really tax the +12V rails of these weaker, cheaper PSU's.
The 850W PSU that I ordered can push 70 Amps to a single +12V rail, so I hope to be pleased with its performance and stability.
Yeah, 750tis take a lot of space, even if they're slightly better, which is why I never bought any. I'm currently using 970s. 980s are just a waste of money as they're too expensive for the little bit extra performance they offers (much like the 290 vs 290x back in the day).
"stock clocks" varies from one brand/version of card to another.
But you can always overclock your cards.
I want to make it possible to overclock over the commandline as a parameter.
The 750ti's has sweat spots gpuspeed/memspead that will improve the hashrate.
Is this also true for the 970 and memory speed? I know AMD cards had a good spot for memory with scrypt, but it didn't matter so much for other algos. Also does efficiency go up or down by moving to that spot?
Curiously what happens to 970 or 750 power efficiency when you OC them a bit? Do they lose a lot of efficiency or does it only go out of spec with additional voltage?
"stock clocks" varies from one brand/version of card to another.
But you can always overclock your cards.
I want to make it possible to overclock over the commandline as a parameter.
The 750ti's has sweat spots gpuspeed/memspead that will improve the hashrate.
The GTX 750 has a tdp of just 38 watt. Many of the early 750ti's had the same limitation Some cards are hashing at fullspeed @40watt.
Thanks sambiohazard. I wanted to know how I was doing. Not sure what a 6pin is? I have this one OCed and am using the switches. Guess I'm running good. I do get 6.4 but, once in a while, it crashes the driver. Pretty cool tho, MinerControl keeps going. Of course, you only mine at ~1/5th of the hash.
Thanks again
edit; musta missed something. Don't think those last 2 posts were meant for me. I don't think...
Thanks again
edit; musta missed something. Don't think those last 2 posts were meant for me. I don't think...
by 6 pin i meant the PCIe power connector that feeds additional power to the card, w/o it & with stock BIOS my card only uses 45W and is throttled after that. So i get low hashrate. Although i am fine with it. I guess with high temps i wouldn't want to push cards too much. hash/watt remain good.
yup - the powered ones NEED the connector ...
the lp versions dont have one at all ...
#crysx
"stock clocks" varies from one brand/version of card to another.
exactly ...
the gigabyte oc versions are actually factory overclocked - but not by much
...#crysx
@crysx, what are your settings for 750ti to get 6.0 MH/s at stock clocks? Mine are only doing ~5.5.
i was completely wrong about the hashrate ...
when it first starts it hits around the 6MH but slowly settles on a lower rate ...
i try and push the intensity levels up to the highest it can go ( which adds up to a huge amount of cuda threads ) - and ends up killing the miner ...
im assuming this is for quark? ...
ill setup the test system ( this one only has 2 x gigabyte 750ti oc ( 1 x oc and 1 x lp ) ) right now and see what levels we get standard ... it will be mining on the sp donation link ( stratum+tcp://donate.granitecoinx.com:7919/ ) so you can see the progress here - https://www.westhash.com/index.jsp?p=miners&a=12&addr=1CTiNJyoUmbdMRACtteRWXhGqtSETYd6Vd ... or here - http://yaamp.com/?address=1CTiNJyoUmbdMRACtteRWXhGqtSETYd6Vd ... in the case of a failover ...
westhash is placed first - then if it fails - the link initiates the failover to yaamp ... soon all donation links will have a failover in the order of ...
1 - westhash
2 - nicehash
3 - yaamp
not much we can do if all of them fail though
... but the system will keep checking each one until one of them comes back ...the parameters for linux ( fedora 20 x64 ) are these - ./ccminer -o stratum+tcp://donate.granitecoinx.com:7919/ -O chrysophylax.donate:x -a quark -r 7 -R 7 -i 24 ...
edit - westhash quark is down for the moment with yaamp having - extranonce2 size=2 errors - on the miner ...
at least the stratum proxy is functioning properly
... westhash is down so it switches to yaamp as a failover while checking to see if westhash will come back up ...nice ...
...the hashrate using the above parameters is getting around the 5700KH ( above and below ) for quark ... but too many errors at the moment ...
-------
[2015-05-13 13:00:03] accepted: 237/519 (45.66%), 11389 khash/s booooo
[2015-05-13 13:00:03] reject reason: Invalid extranonce2 size
-------
i think westhash and yaamp need to reset their backend stratums again ...
#crysx
by 6 pin i meant the PCIe power connector that feeds additional power to the card, w/o it & with stock BIOS my card only uses 45W and is throttled after that. So i get low hashrate. Although i am fine with it. I guess with high temps i wouldn't want to push cards too much. hash/watt remain good.
There are a bunch of different manufacturers of the 750Ti. Some do, some don't. Mine will, but not well.
Thanks sambiohazard. I wanted to know how I was doing. Not sure what a 6pin is? I have this one OCed and am using the switches. Guess I'm running good. I do get 6.4 but, once in a while, it crashes the driver. Pretty cool tho, MinerControl keeps going. Of course, you only mine at ~1/5th of the hash.
Thanks again
edit; musta missed something. Don't think those last 2 posts were meant for me. I don't think...
Thanks again
edit; musta missed something. Don't think those last 2 posts were meant for me. I don't think...
by 6 pin i meant the PCIe power connector that feeds additional power to the card, w/o it & with stock BIOS my card only uses 45W and is throttled after that. So i get low hashrate. Although i am fine with it. I guess with high temps i wouldn't want to push cards too much. hash/watt remain good.
"stock clocks" varies from one brand/version of card to another.
@crysx, what are your settings for 750ti to get 6.0 MH/s at stock clocks? Mine are only doing ~5.5.
is the Slowly degrading percentage of acceptance a normal thing.
when I first start it up it is hitting 100% ([2015-05-12 15:15:37] accepted: 32/32 (100.00%), 31005 khash/s yay!!!)
then after a while it seems to start to slip in percentage and at what percentage is it desirable to shut it down and restart the miner.
what I notice is that i will get a job not found.. the percentage will drop and it will never recover back to 100%
example
[2015-05-12 15:18:55] accepted: 48/48 (100.00%), 31383 khash/s yay!!!
[2015-05-12 15:18:57] GPU #0: GeForce GTX 980, 16497
[2015-05-12 15:18:58] accepted: 48/49 (97.96%), 31396 khash/s booooo
[2015-05-12 15:18:58] reject reason: Job not found.
all subsequent percentages from then on are down.
I can log in remotely and restart the miner but is it possible to initiate an auto reboot if the percentages drop below a certain threshold.
or is it even necessary because the accepted percentage is based on a queue of results that eventually result in the same hash rate regardless ?
when I first start it up it is hitting 100% ([2015-05-12 15:15:37] accepted: 32/32 (100.00%), 31005 khash/s yay!!!)
then after a while it seems to start to slip in percentage and at what percentage is it desirable to shut it down and restart the miner.
what I notice is that i will get a job not found.. the percentage will drop and it will never recover back to 100%
example
[2015-05-12 15:18:55] accepted: 48/48 (100.00%), 31383 khash/s yay!!!
[2015-05-12 15:18:57] GPU #0: GeForce GTX 980, 16497
[2015-05-12 15:18:58] accepted: 48/49 (97.96%), 31396 khash/s booooo
[2015-05-12 15:18:58] reject reason: Job not found.
all subsequent percentages from then on are down.
I can log in remotely and restart the miner but is it possible to initiate an auto reboot if the percentages drop below a certain threshold.
or is it even necessary because the accepted percentage is based on a queue of results that eventually result in the same hash rate regardless ?
Jump to: