Tried different pool? 
Maybe something with initial diff of shares?!
Maybe something with initial diff of shares?!
Which one do you recommend?
From time to time I also get a lot's "(Share is above target)" rejects
Topic: [ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX] - page 223. (Read 3426936 times)
July 25, 2014, 06:38:18 AM
Which one do you recommend?
From time to time I also get a lot's "(Share is above target)" rejects
July 25, 2014, 06:33:30 AM
i've added chaincoin to cudamining.cc, was pretty easy 
forked from djm x17, my personal version use accepted and rejected instead of yay and boo, i like them more lol
forked from djm x17, my personal version use accepted and rejected instead of yay and boo, i like them more lol
July 25, 2014, 05:36:26 AM
I come from the Demoscene. My school is old school
you are working on global optimization? pushing x11 to x2 or more, would be really impressive
pushing x11 means pushing pretty much all other algo on ccminer. I think he is optimizing base algo which is used by all other algo.
July 25, 2014, 05:35:43 AM
I'm pretty sure the output of the last algorithm is used as the input of the next one for X11, precisely so you can't do that.
Yes you can. If each thread is working on a different hash.
If this works as expected I bet we'll see the power consumption and heat production go up on the cards. Everyone used to talk about how Xwhatever was lower power and heat than algos such as Scrypt. That likely was the result of current implementations not being able to fully drive the GPU. Once the power and heat levels go up to what is seen with Scrypt that will be a sign that things are working right.
July 25, 2014, 05:16:36 AM
I come from the Demoscene. My school is old school
you are working on global optimization? pushing x11 to x2 or more, would be really impressive
July 25, 2014, 05:15:04 AM
What's your preferred cl file with ccMiner: whirlcoin.cl, whirlcoin_old.cl or wirlpool.cl?
Is it common with the whirlcoin algo to get after a new start of ccMiner at the beginning about 55 rejects before first MHs gets accepted?
My Nvidia GTX 750 Ti settings:
BTW I get about 40% rejects with the newest ccMiner and about 25% with the newest SPH-SGMiner.
THX for every hint
Tried different pool? Is it common with the whirlcoin algo to get after a new start of ccMiner at the beginning about 55 rejects before first MHs gets accepted?
My Nvidia GTX 750 Ti settings:
Code:
#!/bin/sh
export DISPLAY=:0
export GPU_MAX_ALLOC_PERCENT=100
export GPU_USE_SYNC_OBJECTS=1
cd /home/x/ccminer-djm34-0723
./ccminer -t 1 -a whirlcoin -o stratum+tcp://www.hashharder.com:9649 -u XXX -p X
export DISPLAY=:0
export GPU_MAX_ALLOC_PERCENT=100
export GPU_USE_SYNC_OBJECTS=1
cd /home/x/ccminer-djm34-0723
./ccminer -t 1 -a whirlcoin -o stratum+tcp://www.hashharder.com:9649 -u XXX -p X
BTW I get about 40% rejects with the newest ccMiner and about 25% with the newest SPH-SGMiner.
THX for every hint
Maybe something with initial diff of shares?!
July 25, 2014, 05:14:09 AM
IS there how-to/changelog for latest tsiv miner?
July 25, 2014, 05:04:11 AM
What's your preferred cl file with ccMiner: whirlcoin.cl, whirlcoin_old.cl or wirlpool.cl?
Is it common with the whirlcoin algo to get after a new start of ccMiner at the beginning about 55 rejects before first MHs gets accepted?
My Nvidia GTX 750 Ti settings:
BTW I get about 40% rejects with the newest ccMiner and about 25% with the newest SPH-SGMiner.
THX for every hint
Is it common with the whirlcoin algo to get after a new start of ccMiner at the beginning about 55 rejects before first MHs gets accepted?
My Nvidia GTX 750 Ti settings:
Code:
#!/bin/sh
export DISPLAY=:0
export GPU_MAX_ALLOC_PERCENT=100
export GPU_USE_SYNC_OBJECTS=1
cd /home/x/ccminer-djm34-0723
./ccminer -t 1 -a whirlcoin -o stratum+tcp://www.hashharder.com:9649 -u XXX -p X
export DISPLAY=:0
export GPU_MAX_ALLOC_PERCENT=100
export GPU_USE_SYNC_OBJECTS=1
cd /home/x/ccminer-djm34-0723
./ccminer -t 1 -a whirlcoin -o stratum+tcp://www.hashharder.com:9649 -u XXX -p X
BTW I get about 40% rejects with the newest ccMiner and about 25% with the newest SPH-SGMiner.
THX for every hint
July 25, 2014, 03:09:54 AM
Welp. Managed to split the most offensive part of the kernel into four parallel threads per hash, result is spectacularly unimpressive. The best I've come up with breaks even with the current single thread per hash implementation. Well, almost. It's actually a percent slower AND loses compute 2.0 compatibility due to using shuffle. On the other hands it performs a lot more reasonably with various launch configurations, 15 blocks of 32 threads works our equally well as the original 8x60 magic bullet for 750 Ti.
At this point I'm starting to think I'll just forget about that part and start looking if there's something else to be improved. I'm still curious as to how it runs on other hardware, so if a couple of gents on Win boxes with something else than a 750 Ti in would be willing to take it for a spin, I'd appreciate it. I've added the number for SMX/SMM/Whateverthingmabobs into the miner thread start-up info, you'll probably find your card performing best when the block count is a multiple of the SMX count and the number of threads a power of 2. 4/8/16/32/64 are the best bets.
https://github.com/tsiv/ccminer-cryptonight/releases/download/v0.15-rc1/ccminer-cryptonight_20140723_exp.zip
At this point I'm starting to think I'll just forget about that part and start looking if there's something else to be improved. I'm still curious as to how it runs on other hardware, so if a couple of gents on Win boxes with something else than a 750 Ti in would be willing to take it for a spin, I'd appreciate it. I've added the number for SMX/SMM/Whateverthingmabobs into the miner thread start-up info, you'll probably find your card performing best when the block count is a multiple of the SMX count and the number of threads a power of 2. 4/8/16/32/64 are the best bets.
https://github.com/tsiv/ccminer-cryptonight/releases/download/v0.15-rc1/ccminer-cryptonight_20140723_exp.zip
Also, any chances for this code to get released already? Or are you competing against Wolf0
It works like a charm, 220H/s for GTX760, before it was 190. GTX750TIs seem unchanged.
I get 270H(peaks of 297H with -l 8x50) with this release and a GTX 760 overclocked -->v0.15-rc1 ccminer-cryptonight_20140723
Thanks for that launch setting
306H/s (MSI gaming, +180core, +500mem). Still have to test what's the most stable, but thanks for giving me a start 
Ooh damn, you've released that a looong time ago, tsiv. Should've noticed ^^"
EDIT: 320H/s with +222core, +666mem
I'm waiting anxiously for a driver crash Fantastic Bombadill...im on +180 core +300 Memory
If you find any better launch configs please post it
I also asked in the other thread if there are binaries for wolf nvidia xmr miner
dunno but i can't oc my cards at least one of them keeps crashing if i do so, you changed the power limit in the bios?
No i left the power limit at 100% but every card behaves different in overclock(some better some worse)
Currently i am at +210core/+500 memory stable [asus 760gtx and the drivers djm posted 340.46 with the new cuda]
Also Tsiv thanks for the new binaries i got a good boost
July 25, 2014, 02:19:27 AM
The main GPU is not working at 100%, what can be done?
Use onboard graphics (as primary, for your monitor) instead of your 750Ti if you want it to work like the rest.
Thank you very much, it helped =))
July 25, 2014, 01:53:58 AM
anybody got ideas on how to reduce heat / absorb heat / remove heat from rig ?
1 - avoid over clocking
2 - use fans and open rigs (if possible). Could also use water cooling but I distrust those.
3 - live in a cold country
4 - add more fans
5 - use heat extractors. Basically a fan that will suck the heat out.
That's all I can see...
Thanks.
July 25, 2014, 01:52:30 AM
maybe u can leave out crytonight if its taking too much time. Majority of miner wouldn't be using crytonight i suppose. If they really do want to use , they can always use latest version from tsiv. For me, all the "x" is the "core" algo on miner and crytonight is the "side" algo. Core should be well taken care of as priority (e.g increase speed , performance , etc) before anything else.
Why do i say "x" is core ? Because i based on number of coins that uses "x" algo. If u add them all together (x11,x13,x14,x15,17) , is like 80% of coins out there.
That's the rub. If the results I'm seeing are valid then it WILL BE the algo you want to mine as it will be most profitable (for kepler).
I'm not sure if my "test code" is actually faster or an illusion. I just haven't had time to try and see what hash rate a pool is seeing vs what the miner is showing. I'll be checking this tomorrow of course.
Carlo
yup. Accepted hashrate and hashrate display on miner can be very different sometimes (at least for the case of AMD "WU" and displayed hashrate)
July 25, 2014, 01:26:56 AM
Hey, hey, it's not me that's running cudamining.cc That's bigjme ^^"
I like the sound of Bombidil though (jk)
That's bigjme ^^"I like the sound of Bombidil though (jk)
Oops, I know this but foo bared it. Sorry about that guys,
Your both well respected so it was an easy mix-up.
That's my story and I'm sticking to it!
Carlo
PS: actually I'm just a dumb-ass but I like my previous explanation better. Afterall, I din't get you guys confused with Chritian or djm34.
July 25, 2014, 01:23:13 AM
maybe u can leave out crytonight if its taking too much time. Majority of miner wouldn't be using crytonight i suppose. If they really do want to use , they can always use latest version from tsiv. For me, all the "x" is the "core" algo on miner and crytonight is the "side" algo. Core should be well taken care of as priority (e.g increase speed , performance , etc) before anything else.
Why do i say "x" is core ? Because i based on number of coins that uses "x" algo. If u add them all together (x11,x13,x14,x15,17) , is like 80% of coins out there.
That's the rub. If the results I'm seeing are valid then it WILL BE the algo you want to mine as it will be most profitable (for kepler).
I'm not sure if my "test code" is actually faster or an illusion. I just haven't had time to try and see what hash rate a pool is seeing vs what the miner is showing. I'll be checking this tomorrow of course.
Carlo
July 25, 2014, 01:23:00 AM
The main GPU is not working at 100%, what can be done?
You are clearly using that GPU while mining so the hash-rate drops. A stated above, use the onboard graphics.
As a side note, I am getting lower hash-rates and up to this day no one has been able to help me. Basically I am getting ~300Kh less on every card.
Are you overclocking? I get similar results with stock cards (assuming x11).
No. All stock...
July 25, 2014, 01:21:52 AM
anybody got ideas on how to reduce heat / absorb heat / remove heat from rig ?
1 - avoid over clocking
2 - use fans and open rigs (if possible). Could also use water cooling but I distrust those.
3 - live in a cold country
4 - add more fans
5 - use heat extractors. Basically a fan that will suck the heat out.
That's all I can see...
July 25, 2014, 01:21:04 AM
Hey, hey, it's not me that's running cudamining.cc That's bigjme ^^"
I like the sound of Bombidil though (jk)
That's bigjme ^^"I like the sound of Bombidil though (jk)
July 25, 2014, 01:14:29 AM
Concerning source code to nvMiner.
In reality, this is Chris's, djm34's and tsiv's code (at least 99% of it).
I've made some changes here and there to integrate it all together in a way that allows easier integration of new releases from these devs.
In some cases where they have far different functions instead of trying to integrate them which makes future changes harder (and or prone to bugs) I've added "tsiv" or "djm34" to the function name and then created a "master function that calls their functions based on what algo that is running. Simpler changes are rolled into the same functions.
Not complex at all once you understand what I did. (easy to grasp by anyone who has done programming). Of course many of these changes go far into the code (header, cu, c and cpp files). Examples include duplicate functions in hefty and cryptonight that exist in hefty.cu and cryptonight.cu. And if you don't test all algos after a rebuild some things can bite you in the but.
Anyway, I still plan on posting the code after this next release. I originally wanted to also modify the make files, etc for Linux builds but I'm now thinking this might be best left to someone who is better at Linux builds then myself (I'm not a Linux guys per say). I can "talk you through the changes".
I'm sure the VS solution and project files could also use revamping a bit. So in the interest of moving this project forward, if anyone familiar with programming would like to help me with the initial cleanup and possible CryptoNight debugging mentioned in the last message I posted let me know via PM. I can make the source available to certain individuals who I think can help. All I'll ask is that you don't give it to anyone else (yet) and send any changes back to me to coordinate. I just don't want the source out in the open yet until it will compile on both Windows and Linux and I get the speed up on CN. I also don't want anyone posting EXEs from this source code before it's published to github as it may cause confusion as to the source. Otherwise I'll have a ton of PMs and support requests for compiling help which I want to avoid for now until it's officially published on github. This should not take long at all.
I'll say this now. I won't send code to anyone I'm not familiar with from the forum (yet). It will be posted soon publicly but not yet. However, anyone who has been around this forum and has shown any type of programming contribution is welcome to PM me for the code.
I hope it's clear the code will be available shortly for EVERYONE but I just want to control the source and changes a few days longer before posting to github to make sure it's cleaned up appropriately and working under both Windows and Linux. I don't know how else to say it but I want the source available but don't want undue stress from doing so.
I'm also going to ask anyone interested in doing the Linux side of things not to bother unless they are prepared to put together a well documented set of directions to configure/compile/make this. This person must be willing to start with a fresh install of Ubunto 14.10 and document all command lines needed to compile. This documentation will be forwarded to Bombidil for posting on cudaming.cc as well as included in a UbuntoReadmeCompile.txt type file included with the source.
We have have been missing this as a community (how to easily compile under Linux) so I want it to be clear enough for a Windows users who has successfully installed Ubunto 14.10 to take over following a readme file/directions. EVERY instruction included. I'll be the judge of what does or does not work as I'm not a Linux guru and if I can do it then anyone can. Well I'll play dumb and make sure any windows user can follow the directions under Linux.
Carlo
In reality, this is Chris's, djm34's and tsiv's code (at least 99% of it).
I've made some changes here and there to integrate it all together in a way that allows easier integration of new releases from these devs.
In some cases where they have far different functions instead of trying to integrate them which makes future changes harder (and or prone to bugs) I've added "tsiv" or "djm34" to the function name and then created a "master function that calls their functions based on what algo that is running. Simpler changes are rolled into the same functions.
Not complex at all once you understand what I did. (easy to grasp by anyone who has done programming). Of course many of these changes go far into the code (header, cu, c and cpp files). Examples include duplicate functions in hefty and cryptonight that exist in hefty.cu and cryptonight.cu. And if you don't test all algos after a rebuild some things can bite you in the but.
Anyway, I still plan on posting the code after this next release. I originally wanted to also modify the make files, etc for Linux builds but I'm now thinking this might be best left to someone who is better at Linux builds then myself (I'm not a Linux guys per say). I can "talk you through the changes".
I'm sure the VS solution and project files could also use revamping a bit. So in the interest of moving this project forward, if anyone familiar with programming would like to help me with the initial cleanup and possible CryptoNight debugging mentioned in the last message I posted let me know via PM. I can make the source available to certain individuals who I think can help. All I'll ask is that you don't give it to anyone else (yet) and send any changes back to me to coordinate. I just don't want the source out in the open yet until it will compile on both Windows and Linux and I get the speed up on CN. I also don't want anyone posting EXEs from this source code before it's published to github as it may cause confusion as to the source. Otherwise I'll have a ton of PMs and support requests for compiling help which I want to avoid for now until it's officially published on github. This should not take long at all.
I'll say this now. I won't send code to anyone I'm not familiar with from the forum (yet). It will be posted soon publicly but not yet. However, anyone who has been around this forum and has shown any type of programming contribution is welcome to PM me for the code.
I hope it's clear the code will be available shortly for EVERYONE but I just want to control the source and changes a few days longer before posting to github to make sure it's cleaned up appropriately and working under both Windows and Linux. I don't know how else to say it but I want the source available but don't want undue stress from doing so.
I'm also going to ask anyone interested in doing the Linux side of things not to bother unless they are prepared to put together a well documented set of directions to configure/compile/make this. This person must be willing to start with a fresh install of Ubunto 14.10 and document all command lines needed to compile. This documentation will be forwarded to Bombidil for posting on cudaming.cc as well as included in a UbuntoReadmeCompile.txt type file included with the source.
We have have been missing this as a community (how to easily compile under Linux) so I want it to be clear enough for a Windows users who has successfully installed Ubunto 14.10 to take over following a readme file/directions. EVERY instruction included. I'll be the judge of what does or does not work as I'm not a Linux guru and if I can do it then anyone can. Well I'll play dumb and make sure any windows user can follow the directions under Linux.
Carlo
July 25, 2014, 01:09:20 AM
anybody got ideas on how to reduce heat / absorb heat / remove heat from rig ?
July 25, 2014, 12:45:13 AM
Update on next release of nvMiner. I'll sum up the rest of this message with one one: FRUSTRATED
I've been working pretty much for the last 24 hours on this.
I have djm34's x17 working. I have tsiv's latest CryptoNight working.
HOWEVER (yea you knew there was a but/however)
I've also been testing/trying/implementing some additional code changes for hash increases. Some are based directly on sp_ and wolf0's mods and some are based on ideas they put forth that I ran with.
Here's the rub. I've increased speed a bit or many algos. Pretty much all of Chris/djm34's are done. Nothing earth shattering by any means but every bit of hash helps long term. tsiv's CryptoNight is another story.
Using pure tsiv code I'm getting slower 750ti hash rates and maybe 10 to 15% faster on kepler. I have another test build of tsiv's code with changes that is getting me 20-25% faster kepler speeds and ever so slightly faster 750ti speeds.
But the rub is that I haven't been able to migrate these changes into nvMiner. It's driving me crazy. I'm doing compares between code and can't find why one EXE runs 20 to 25% faster then the other on kepler.
I now have 3 machines setup doing compiles to figure this out. Main problem is that a full rebuild is around 4 hours. It's a process of elimination and testing. I'll figure it out as I always do but it's time consuming.
So anyway, I didn't want you guys to think I've forgotten about you. I'm working on it as fast as I can but compile times on 5.5 are SLOW regardless of hardware thrown at it.
Carlo
I've been working pretty much for the last 24 hours on this.
I have djm34's x17 working. I have tsiv's latest CryptoNight working.
HOWEVER (yea you knew there was a but/however)
I've also been testing/trying/implementing some additional code changes for hash increases. Some are based directly on sp_ and wolf0's mods and some are based on ideas they put forth that I ran with.
Here's the rub. I've increased speed a bit or many algos. Pretty much all of Chris/djm34's are done. Nothing earth shattering by any means but every bit of hash helps long term. tsiv's CryptoNight is another story.
Using pure tsiv code I'm getting slower 750ti hash rates and maybe 10 to 15% faster on kepler. I have another test build of tsiv's code with changes that is getting me 20-25% faster kepler speeds and ever so slightly faster 750ti speeds.
But the rub is that I haven't been able to migrate these changes into nvMiner. It's driving me crazy. I'm doing compares between code and can't find why one EXE runs 20 to 25% faster then the other on kepler.
I now have 3 machines setup doing compiles to figure this out. Main problem is that a full rebuild is around 4 hours. It's a process of elimination and testing. I'll figure it out as I always do but it's time consuming.
So anyway, I didn't want you guys to think I've forgotten about you. I'm working on it as fast as I can but compile times on 5.5 are SLOW regardless of hardware thrown at it.
Carlo
maybe u can leave out crytonight if its taking too much time. Majority of miner wouldn't be using crytonight i suppose. If they really do want to use , they can always use latest version from tsiv. For me, all the "x" is the "core" algo on miner and crytonight is the "side" algo. Core should be well taken care of as priority (e.g increase speed , performance , etc) before anything else.
Why do i say "x" is core ? Because i based on number of coins that uses "x" algo. If u add them all together (x11,x13,x14,x15,17) , is like 80% of coins out there.
Jump to: