 Show Posts
|
|
Pages: [1] 2 »
|
I have CUDA 6.5 installed and mpir 2.7 alpha 10, what should I delete from Makefile.am ?
The edit of Makefile.am just decreases compile time by not bothering with anything other than compute_50. Find any lines with compute_50 and then delete the other options to leave you with just compute_50 like this: $(NVCC) @CFLAGS@ -I . -Xptxas "-abi=no -v" -gencode=arch=compute_50,code=\"sm_50,compute_50\" --maxrregcount=80 --ptxas-options=-v $(JANSSON_INCLUDES) -o $@ -c $< This means it will only work with compute_50 but it will compile very quickly. |
|
|
|
|
I have been away for a while, so just catching up.
Just downloaded the new ccminer code from github.
Installed latest CUDA 6.5
Install latest mpir 2.7 (alpha 10)
Deleted everything except compute_50 from Makefile.am
All very straightforward, and very fast compile.
First indications suggest about 5+% faster on X11 and X15 at least (750Ti)
Thanks djm34 for continued great work
|
|
|
|
I miss cbuchner1
me too...  yeah, sometimes I miss myself. Welcome back Christian! I am struggling with something silly using ccminer with python (on linux) and I am sure you will know the answer: I am using subprocess.Popen to start ccminer, and I redirect stdout to a file, *BUT* I am getting nothing in the file and still getting all the output to the screen :-( My C programming is even worse than my python (and that's bad) but it looks like the applog function is writing to something other than STDOUT? Or at least subprocess.Popen doesn't recognise it as STDOUT? Any clues? Ta When you start the process, try to add "2> /somewhere/ccminer.log", that is how I get the output of my background ccminer in my linux rig  Thanks for the help. Christian dropped me an email with a hint. Basically ccminer is writing to STDERR and not STDOUT. I was only redirecting STDOUT and doing nothing with STDERR. That sorted it. Your comment is exactly right for people using bash shell, but I was doing it within python. For bash 2>&1 would do the trick (redirecting STDERR to STDOUT) All working now. Ta |
|
|
|
If you want quick compilation (for compute_50) modify Makefile.am: change all compute_30/sm_30 into compute50/sm_50 and get rid of compute_35/sm35 strings.
Thanks to that compilation takes ca 15 minutes but ccminer will work only on maxwell cards with compute 5.0.
That's a great tip - so much quicker! Thanks |
|
|
|
I miss cbuchner1
me too... yeah, sometimes I miss myself. Welcome back Christian! I am struggling with something silly using ccminer with python (on linux) and I am sure you will know the answer: I am using subprocess.Popen to start ccminer, and I redirect stdout to a file, *BUT* I am getting nothing in the file and still getting all the output to the screen :-( My C programming is even worse than my python (and that's bad) but it looks like the applog function is writing to something other than STDOUT? Or at least subprocess.Popen doesn't recognise it as STDOUT? Any clues? Ta |
|
|
|
Any ideas on how to monitor hash rate with ccminer running in background on linux?
I have a switching program that moves pools by stopping and starting ccminer. I don't have any display output though (eg; stdin and stderr both go to /dev/null)
Is there way to monitor what the cards are doing in terms of hash rate and temperature?
redirect stdout+stdin to a file instead /dev/null. you can truncate the file periodically to avoid excesive growth. Yes that should work, but I am using Python's Popen to start ccminer and funny things happen with stdout etc. However I will give that a go and see what happens. I was hoping there might be another command somewhere that would report stats :-( Thanks |
|
|
|
|
Any ideas on how to monitor hash rate with ccminer running in background on linux?
I have a switching program that moves pools by stopping and starting ccminer. I don't have any display output though (eg; stdin and stderr both go to /dev/null)
Is there way to monitor what the cards are doing in terms of hash rate and temperature?
|
|
|
|
Hi Guys, I use ccminer on a linux machine with 4 off GTX 750 Ti (no OC) My hash rates are all in the right ball park but always lower than others report. Can someone summarise the best versions of CUDA and drivers to compile new ccminer versions on linux? I am using: cuda_6.0.37 NVIDIA-Linux-x86_64-340.17 ccminer source was taken from djm34's fork on 11th July. # ccminer --version *** ccMiner for nVidia GPUs by Christian Buchner and Christian H. *** This is version 1.2 (beta) based on pooler-cpuminer 2.3.2 (c) 2010 Jeff Garzik, 2012 pooler based on pooler-cpuminer extension for HVC from https://github.com/heavycoin/cpuminer-heavycoin and http://hvc.1gh.com/ Cuda additions Copyright 2014 Christian Buchner, Christian H. LTC donation address: LKS1WDKGED647msBQfLBHV3Ls8sveGncnm BTC donation address: 16hJF5mceSojnTD3ZTUDqdRhDyPJzoRakM YAC donation address: Y87sptDEcpLkLeAuex6qZioDbvy1qXZEj4 ccminer 2014.06.15 libcurl/7.22.0 GnuTLS/2.12.14 zlib/1.2.3.4 libidn/1.23 librtmp/2.3 Thanks |
|
|
|
|
Is everything working right on TradeMyBit at the moment (11:00 BST 20th July)?
The algorithm profits don't seem to be changing at all? eg; stuck on X11?
Best has been X11 on 0.00027 or so for well over an hour. That is unusual.
Is there an API call to get all coin profits so that I can check myself?
|
|
|
|
Hey guys, Have you made a change to the bestalgo API call??? [{"algo":"x11","score":"0.00327663","actual":"0.000595751"},{"algo":"nist5","score":"0.00266068","actual":"0.000161253"},{"algo":"x13","score":"0.00215897","actual":"0.000583506"},{"algo":"nscrypt","score":"0.0012064","actual":"0.00256681"},{"algo":"scrypt","score":"0.000672299","actual":"0.000672299"},{"algo":"x15","score":"0.000350805","actual":"0.000103178"}]
That's the normalised score and the actual score? Great!! Thanks very much How about an API call for raw (non normalized) BTC/Day/Mhash earnings per algorithm?
I'm planning to extend my autoswitcher (currently works with NiceHash) to support TradeMyBit. To make it work correctly I need the non-normalized earnings per algorithm as I factor in hash rates myself to get the best bang for the hash. The "bestalgo" call has the format I want, but it currently has normalized values. If another key/value pair could be added per algorithm for the non-normalized values this would work great.
+1 vote for this request |
|
|
|
|
Can some explain the details of the proportional payout system that is used on TMB please?
With the use of the multi-algorithm miners (or in my case my home-made API based switcher) on the TMB pools, how does a miner get rewarded when swapping between algorithm pools?
Is the payout based only on shares within a block regardless of time? Some PPS variants use a time element to discourage pool hopping - eg; shares in last N seconds or similar or so called PPLNS.
If the difficulty is relatively low within the block and the number of shares is low, then the payout per share at that stage will be better than average for the block? Conversely if the difficulty is higher and the number of shares is high, then the payout per share is worse than average for the block. This is the whole point of the 'pool-hopping' problem isn't it?
How does the multi-algorithm hopping take account of this? Does it take this into account when switching?
For people using their own API-switching code, are you looking at this when making the switching decision? eg; number of shares and difficulty
I am doing a simple switch based only on the reported profitability every 5 minutes. If the profitability of a different algorithm is +5% then I switch. If not I stick. However I feel I should be taking more into account before switching.
Thoughts?
|
|
|
|
Not that I can see either. All closed, and binaries for Windows only :-( Looking forward to trying something when (if?) the source does get released though. This weekend I am going to try KopiemTu 1.4 and see what that is like - reading good things about it, and I like the idea of trying some tweaking of the card on linux. https://litecointalk.org/index.php?topic=16800.0No sourcecode available in nvminer.zip.
|
|
|
|
I would love to try that but I am running on linux (ubuntu) and so can't easily overclock as far as I know. Unless people know otherwise?
2 x 750 (non ti) give me 4.1-4.2 on X11. Overclocked (see my other post).
planning to replace my 40 280x with 66 Ti's , will save 1/2 of power costs.
|
|
|
|
3 cards giving you about 3,600 khash/s? I have 4 cards giving me about 9,500 khash/s for X11 so your numbers look a little low to me. I am running the very latest ccminer from djm34 compiled with CUDA 6.0 ( https://github.com/djm34/ccminer) is this value is normal for gtx 750 ti for x11 mining fractal
[2014-07-16 17:22:38] Stratum detected new block
[2014-07-16 17:22:38] thread 0: 23003137 hashes, 1193 khash/s
[2014-07-16 17:22:38] thread 2: 25624577 hashes, 1220 khash/s
[2014-07-16 17:22:38] thread 1: 26542081 hashes, 1261 khash/s
[2014-07-16 17:22:44] thread 2: 7303333 hashes, 1189 khash/s
[2014-07-16 17:22:44] accepted: 374/374 (100.00%), 3643 khash/s (yay!!!)
[2014-07-16 17:23:01] thread 0: 27686087 hashes, 1200 khash/s
[2014-07-16 17:23:01] accepted: 375/375 (100.00%), 3650 khash/s (yay!!!)
[2014-07-16 17:23:17] thread 2: 40878269 hashes, 1223 khash/s
[
|
|
|
|
No, you are correct I do not take into account the 2% withdraw fee. To be honest I was just adding the NiceHash numbers to give me a rough idea of how they are comparing. So far TMB is winning (and the gap is more than 2%). I will factor in that 2% fee from now on. Then next thing to do is to actually record earning properly so that we aren't misled by the stats given to us by the pools!! Having said all that, NiceHash has just overtaken TMB!! I need to get that pool switch code working ------------------------------------------
2014-07-16 13:47:51 Current Rate: nist5 0.0075
2014-07-16 13:47:51 NiceHash Best Rate: x13 0.0070
2014-07-16 13:47:51 Best Rate: nist5 0.0075 stay: algo still best
------------------------------------------
2014-07-16 13:52:52 Current Rate: nist5 0.0068
2014-07-16 13:52:52 NiceHash Best Rate: x13 0.0070
2014-07-16 13:52:52 Best Rate: nist5 0.0068 stay: algo still best
------------------------------------------
2014-07-16 13:57:54 Current Rate: nist5 0.0027
2014-07-16 13:57:54 NiceHash Best Rate: x13 0.0070
2014-07-16 13:57:54 Best Rate: x11 0.0066 switch: performance +138.9%
De-normalizing will probably be my workaround as well. Unfortunately it means the users have to update those values every time the pool changes the normalization factors.
When comparing NiceHash to TMB are you also factoring in the 2% withdrawl fee on TMB? I'm curious myself to see how the two services stack up but it will be a week or two before I have everything wired up.
|
|
|
|
Just done the same thing! And I wanted raw earnings as well. All I did was take their settings and divide all values by their normalisation factor. Not so hard, but it is a pain to keep up to date. Raw would be so much better. Simple python dictionary. tmb_n = {}
tmb_n['scrypt'] = 1.000
tmb_n['nscrypt'] = 0.470
tmb_n['x11'] = 5.500
tmb_n['x13'] = 3.800
tmb_n['x15'] = 3.400
tmb_n['nist5'] = 16.500
I check every 5 minutes, then if the difference is more than 5% I will switch. If less than 5% I leave it run. I also added a check against NiceHash with a view to adding code to switch pool if needed but so far NiceHash never beats the reported TMB numbers. For example: ------------------------------------------
2014-07-16 12:22:39 Current Rate: x13 0.0107
2014-07-16 12:22:40 NiceHash Best Rate: x13 0.0063
2014-07-16 12:22:40 Best Rate: x13 0.0107 stay: algo still best
------------------------------------------
2014-07-16 12:27:41 Current Rate: x13 0.0081
2014-07-16 12:27:41 NiceHash Best Rate: x13 0.0064
2014-07-16 12:27:41 Best Rate: x11 0.0083 stay: performance +3.0%
------------------------------------------
2014-07-16 12:32:42 Current Rate: x13 0.0085
2014-07-16 12:32:42 NiceHash Best Rate: x13 0.0065
2014-07-16 12:32:42 Best Rate: x13 0.0085 stay: algo still best
------------------------------------------
2014-07-16 12:37:43 Current Rate: x13 0.0143
2014-07-16 12:37:44 NiceHash Best Rate: x13 0.0065
2014-07-16 12:37:44 Best Rate: x11 0.0226 switch: performance +57.3%
------------------------------------------
2014-07-16 12:42:45 Current Rate: x11 0.0091
2014-07-16 12:42:45 NiceHash Best Rate: x13 0.0065
2014-07-16 12:42:45 Best Rate: x13 0.0098 switch: performance +8.6%
------------------------------------------
2014-07-16 12:47:46 Current Rate: x13 0.0129
2014-07-16 12:47:46 NiceHash Best Rate: x13 0.0065
2014-07-16 12:47:46 Best Rate: x13 0.0129 stay: algo still best
------------------------------------------
2014-07-16 12:52:47 Current Rate: x13 0.0088
2014-07-16 12:52:47 NiceHash Best Rate: x13 0.0065
2014-07-16 12:52:47 Best Rate: x13 0.0088 stay: algo still best
How about an API call for raw (non normalized) BTC/Day/Mhash earnings per algorithm?
I'm planning to extend my autoswitcher (currently works with NiceHash) to support TradeMyBit. To make it work correctly I need the non-normalized earnings per algorithm as I factor in hash rates myself to get the best bang for the hash. The "bestalgo" call has the format I want, but it currently has normalized values. If another key/value pair could be added per algorithm for the non-normalized values this would work great.
|
|
|
|
Hi Guys, I have written a little bit of code to check the different algorithm profits on TradeMyBit and then to switch my ccminer accordingly. I use GTX 750 Ti so I have to use ccminer and so the need to right my own code. I check the profit every 5 minutes, like this. First value is 'normalized' profit from TMB API, second value is my corrected profit taking into account my rig's hash rate and power consumption. nist5 0.00022 0.00589
x13 0.00072 0.00532
x11 0.00051 0.00481
nscrypt 0.00269 0.00129
scrypt 0.00068 0.00068
x15 0.00007 0.00034
2014-07-15 08:54:34 Best Rate: nist5 0.00589 switch
x13 0.00102 0.00754
nist5 0.00016 0.00431
x11 0.00039 0.00372
nscrypt 0.00248 0.00119
scrypt 0.00080 0.00080
x15 0.00007 0.00034
2014-07-15 08:59:35 Best Rate: x13 0.00754 switch
When it says 'switch' it will kill the running ccminer and start a new one to the new port/algorithm. However I think it is switching too frequently as my profits are not so good in testing. Can someone who understands this better than me (wouldn't be hard) comment on some logic on when to switch algorithms to maximise profit. Too quickly and you lose shares? Too slow and you stay in the less profitable algorithm. The profits seem to fluctuate a lot. Maybe I need a rolling average? What about difficulty?? Thanks |
|
|
|
Hi, Nice work on this. I moved off NiceHash last week after all their issues and am using TradeMyBit for the moment. I am a Linux user so can't try your tool :-( I have written a similar tool in python for use on my linux rig (4 x GTX 750 Ti). No graphics of course :-) I am interested in the logic on how often to switch or not. When will switching cause you to lose any work done before being paid? I guess it is down to the details of the pool but have you looked into this for NiceHash? Some pools will calculate the earnings once a block is found and if you are not mining at the time you get nothing? Thoughts? Just for fun here is the output from my tool on linux running against TradeMyBit. First value is profit reported from them, second value is my actual profit, normalised for hash rate and power consumption. nist5 0.00022 0.00589
x13 0.00072 0.00532
x11 0.00051 0.00481
nscrypt 0.00269 0.00129
scrypt 0.00068 0.00068
x15 0.00007 0.00034
2014-07-15 08:54:34 Best Rate: nist5 0.00589 switch
x13 0.00102 0.00754
nist5 0.00016 0.00431
x11 0.00039 0.00372
nscrypt 0.00248 0.00119
scrypt 0.00080 0.00080
x15 0.00007 0.00034
2014-07-15 08:59:35 Best Rate: x13 0.00754 switch
I would like to work out some logic on when to switch and when to wait....any ideas? |
|
|
|
Just for interest here is what I am getting at the moment. On a 750 GTX Ti running djm34's latest fork off ccminer ( https://github.com/djm34/ccminer) compiled with CUDA 6.0. Per card of course, and rounded down. X11 2,300k/s X13 1,800k/s (about 50w) x15 1,200k/s Nist5 7,500k/s (about 54w) I have tried various versions with CUDA 6.0 and 5.5 but its really hard to see any difference. I am sticking on TradeMyBit at the moment with a homemade switcher that is flipping between X13 and Nist5 most of the time I'm using this Mod v1.02-X13-version - 12/06/2014 she works for x11 x13 and x11 doing more in 2.200k / s
I can't recall if v1.02 contains the improved groestl implementation. Give v1.2 a try and you might see an improvement with both x11 and x13. |
|
|
|
How is the Payout for Nicehash compared to a decent sized ordinary multipool like trademybit.com??
Unless there are massive amount of buyer to compete with each other on the bids, the price should always be noticeably lower than a decent sized traditional multipool, given their profit switching are both efficient.
Good question. I moved to trademybit yesterday when nicehash had its problems. I must be misunderstanding something here so can some check my numbers? Nicehash X13 paying 0.5 BTC/day/Ghash this is 0.0005 BTC/day/Mhash (1GHash = 0.001Mhash) TradeMyBit X13 paying 0.00079 BTC/day/Mhash which is 0.79 BTC/day/Ghash (currently via APEX, but I autoexchange to BTC) So doesn't that make TradeMyBit more profitable? 0.79 vs 0.5? Assuming the stats are both inaccurate, then I would say my actual BTC payments are much higher with TMB? Is this because I have just a very small minnow running around 10Mhash/s? Interested in the views of others |
|
|
|
|
