OFFICIAL CGMINER mining software thread for linux/win/osx/mips/arm/r-pi 4.11.1

[ancow]

The CPU mining component of cgminer is annoying me as random bugs keep coming up, I have no incentive to work on it, and I am quite against CPU mining as a concept any more. Unless someone gives me a good reason to work on it any more, I'm going to recommend people go back to the original cpuminer code or ufasoft, and I will code out the cpu mining component of cgminer. Comments please!

I'd also prefer it if it didn't go. I'm not sure how many people are affected by my scenario, but I have a few computers that, when I switch them on, run for quite a while with barely any load, just sitting there and wasting energy. If I use the linux cpu governors' ignore_nice_load option, these machines will at least do something useful while running without using a whole lot more energy than if they just sit idle.

Being able to use the same configuration files across these computers as well as the ones I GPU mine on is a major attraction of cgminer, as far as I'm concerned.

[Meatball]

. The main argument against CPU mining (electricity costs make it
  an unprofitable undertaking) is, IMO, not a very good one. Mining
  is not just about making money, it's also very much about sustaining
  the bitcoin ecosystem. I personally mine on CPUs because I have a
  bunch of idle boxes in remote data centers that I want to use to
  support bitcoin.

I'm all for people CPU mining if they have one or two boxes that they just want to throw into the mix to support bitcoin and are not concerned about the profit side of things, but if you are CPU mining on anything more than a handful of machines you're throwing money out the window.

You can pick up a Radeon 5770 for $100 that will easily do 200-210 MH/s.  It only uses 100 watts at full load.  The most you can likely get out of even a high end CPU is about 20 MH/s and that'll use about the same amount of wattage running the GPU.  So, let's calculate it out using approximate numbers.  Assuming 24x7 running, you'd need 10 of those CPU's to equal the Radeon 5770 and would use about 1000W compared to the 100W on the 5770.  We'll also assume a pretty good electric rate of 10 cents/kwh.  

To run the 5770 for a month (24x7 for 30 days), we're looking at 72 Kwh of power or $7.20 a month.

To run the 10 CPU's for a month (24x7 for 30 days), we're looking at 720 kwh of power, or $72 a month.

I understand people hate to 'waste' idle CPU's sitting there doing nothing, but I don't understand why they're willing to 'waste' all that electricity in the process.  Seriously, grab a single 5770 and let it mine, you'll pay it off in power savings in like 6 weeks and not waste all that electricity.

[ovidiusoft]

Second, you don't seem to have much experience with datacenters.
If you did, you'd realize that plonking a GPU in a rackmounted
server is slightly more complicated than doing it on the gaming
PC sitting under your desk.

I will also add to this that I pay a fixed amount for colocating my servers in a datacenter here in Bucharest. I don't pay for electricity, I don't pay for traffic. So running a CPU miner on low priority as to not interact with the normal services on those servers makes perfect sense for me. As for GPU mining, as soon as someone fits a 6990 in a 1U server, let me know, I'm very interested

[The LT]

First, thanks for rehashing the deeply obvious.

Second, you don't seem to have much experience with datacenters.

If you did, you'd realize that plonking a GPU in a rackmounted
server is slightly more complicated than doing it on the gaming
PC sitting under your desk.

Seriously, are you saying that your average DC gets its electricity for free? At a discount, yes, but not for free. Also, are you implying that it's a good idea to bog a rackmounted server down with mining calculations? From where I come from, datacenter servers do more useful work rather than waste cpu cycles mining. And we all know how efficient mining is on CPU's.

[xcooling]

it would be stupid to cpu mine on a webserver.

[ancow]

[I understand people hate to 'waste' idle CPU's sitting there doing nothing, but I don't understand why they're willing to 'waste' all that electricity in the process.  Seriously, grab a single 5770 and let it mine, you'll pay it off in power savings in like 6 weeks and not waste all that electricity.

This isn't really about machines meant for mining. These are machines that already sit around, using said electricity. Adding anything, even a graphics card, will make them use more electricity, which is not an option.

This really isn't about making money, it's about giving them something to do while they're standing there and supporting something I like while I'm about it. (AAMOF, if ckolivas decides to remove CPU mining capabilities, I'll probably just use them for boinc/FAH/...)

[Meatball]

There's options out there if you really want to stick GPU's into a rack mountable server.  Supermicro has a ton of machines for that, though I'm sure they're probably not cheap.

[The00Dustin]

This isn't really about machines meant for mining. These are machines that already sit around, using said electricity. Adding anything, even a graphics card, will make them use more electricity, which is not an option.

This really isn't about making money, it's about giving them something to do while they're standing there and supporting something I like while I'm about it. (AAMOF, if ckolivas decides to remove CPU mining capabilities, I'll probably just use them for boinc/FAH/...)

OK, you're missing the point the anti-cpu-mining posters are making.  Using you CPU for mining will use MORE electricity than adding a GPU AND using it for mining.  The CPU mining will also put off more heat for the same amount of work.  Idle equipment draws less power and puts out less heat (therefore again drawing less power on cooling).  As such, your "adding a graphics card will make it use more electricity, which is not an option" argument implies that you should NOT be CPU mining.

Outside of the nonsense that is the same arguments being made in the same looping fashion (yes, if you don't care about electricity/making money and want to support the project in an ethical way, fine, there is nothing wrong with CPU mining) in thread after thread:

Not that my 2 cents should count for much, but I would like to see cpu mining support maintained, fixed, and improved (without interfering with any potential gpu mining maintenance and improvements, of course).

[Meatball]

You're still not getting it.

This is not about mining and making a profit, but taking a boatload
of existing machines that aren't easily upgraded with a GPU and aren't
doing much CPU-wise and getting them to participate in the bitcoin effort.

Much in the same way people donate their CPU and time (and - shock ! -
their electricity) to things like FAH and Seti.

Whatever man.  Like I said, I'm all for chipping in a few CPU's to support BTC, but wasting hundreds or thousands of dollars a month of someone else's money for electricity on your "boatload" of CPU mining machines for that is just silly when you can 'support' BTC the same amount with a few cheap cards and not waste all that money/power.  I just hope all that warm and fuzzy feeling you get from supporting bitcoin makes you feel better when you get tossed from the data center or fired from your job for using someone's else's gear/electric.  Whether or not it's you, someone has to pay for all that power.

I assume we're not going to see eye to eye on this, so why don't we just drop it and let people keep using this thread for actual CGMiner talk

[ancow]

Using you CPU for mining will use MORE electricity than adding a GPU AND using it for mining.  The CPU mining will also put off more heat for the same amount of work.  Idle equipment draws less power and puts out less heat (therefore again drawing less power on cooling).

Welcome to computing. From your claims I assume that you only joined the community a couple of years ago, since you apparently haven't been around for the frequency battles and only know about the very recent "green computing" vibe.

Seriously, though, my old desktop with the 3GHz P4 draws pretty much the same amount of energy as well as producing similar amounts of heat whether under load or not. Now, a new graphics card (which can't be too modern since I only have AGP available) would definitely add to the idle heat production and power consumption.

(I can't believe I just had to give a concrete example - the thing's only 7 years old, FFS!)

Just so you don't make that mistake again, try adding the prefix "modern" to your claims about computers in the future. Some of these old birds have some life in them yet...

[d3m0n1q_733rz]

Umm, if I may interject, CPUs are far more energy efficient today than they were years ago.  My Core2 Quad is rated at 95W max.  However, I have found that it is still steady at a much lower voltage while being overclocked allowing me to do ~15MHash/sec with about as much power as a single lightbulb (not the cool curly kind, the generic).  Now, if you're wanting to get into a debate of power usage, let's not forget that a CPU core is required to assist in running the GPU.  However, to do the same amount of work, the HD5450 states 19.1W on spec. for "typical" usage.  But I figure if we're maxing out a core on the CPU anyway, why not just go ahead and max them all?
Now, with the newer OpenCL stream intrinsics, there is the ability to automatically determine if certain math calculations should be run on the CPU or the GPU.  This allows the two to work better TOGETHER by having the CPU precalculate some of the earlier rounds before things start getting tough and then pass it to the GPU to complete and possibly increase the hashing rate significantly by reducing the overhead of minor calculations on GPU load and offloading the huge calculations from the CPU.

My point is, I think we may have been looking at this the wrong way.  We're thinking of it as a matter of one or the other while maybe we should be thinking of it as one AND the other.

[-ck]

Just a data point for those not aware, since I am a linux kernel CPU scheduler programmer as well. A modern CPU when idle uses a LOT LESS power than when it's running computation EVEN AT THE SAME FREQUENCY. All modern CPUs adjust power according to their requirements, even within the same frequency profile. It's a misconception to think that just because a CPU is running at low frequency that it's using the same amount of power whether idle or mining.

[kripz]

When does cgminer disable and re-enable a pool?

I just checked on my miner and the pool was disabled for some reason. The pool came back up and cgminer didn't re enable it straight away.

[The00Dustin]

Using you CPU for mining will use MORE electricity than adding a GPU AND using it for mining.  The CPU mining will also put off more heat for the same amount of work.  Idle equipment draws less power and puts out less heat (therefore again drawing less power on cooling).

Welcome to computing. From your claims I assume that you only joined the community a couple of years ago, since you apparently haven't been around for the frequency battles and only know about the very recent "green computing" vibe.

Seriously, though, my old desktop with the 3GHz P4 draws pretty much the same amount of energy as well as producing similar amounts of heat whether under load or not. Now, a new graphics card (which can't be too modern since I only have AGP available) would definitely add to the idle heat production and power consumption.

(I can't believe I just had to give a concrete example - the thing's only 7 years old, FFS!)

Just so you don't make that mistake again, try adding the prefix "modern" to your claims about computers in the future. Some of these old birds have some life in them yet...

Seriously?  You're going to claim that your view has been proven and not cite anything?  You're also going to suggest that a 386 is worth mining with (even a 486 could run idle cycles to save power)?  I'm running a Pentium D 3.0 overclocked to 3.75, it runs at 60C idle overclocked that high (with 20C room temp).  If I start mining on it, it will get to 80C in minutes.  It would get well under 10MH/s if I bothered to do this.  I'm also running a Radeon HD 5830 overclocked to 999MHz at performance level 2.  At idle the Radeon runs around 32C (with 22C room temp).  When I mine with it, it eventually gets up to just over 65C.  I'm getting 320MH/s there with cgminer 1.5.6 (was getting 318.7MH/s at 1005MHz with cgminer 1.5.3).  Both of these devices have very similar TDP (they draw a similar wattage at full load); they also run at similar voltages.  The higher temperature of the processor is directly related to it wasting more of its energy.  However, the Radeon runs underclocked and undervolted when not in use, hardly drawing any power at all, and presumably it also has idle cycles to lower that draw even more, much like any processor capable of running an operating system written in the last 16 years.  Prior to idle cycles, your argument would make sense since the processor ran at full power all the time, but since we'd be talking about a 386 or something running an OS from prior to the second half of the 90s (or something newer if it still didn’t tell the processor to idle), this point is moot.  I suppose it is arguable that my "AND using it for mining" comment is invalid in this example, but the Pentium D isn't an efficient processor by any means.  Regardless of your definition of "modern" equipment, my point from before you chopped it out of my post stands (if adding a video card is unacceptable because it will draw more power, then CPU mining is more unacceptable [this is especially true if a video card could be underclocked to the point where it draws less power than the additional power for a CPU under full load since it would still perform better in mining]).  Moreover, the newer a processor the less true my argument because newer equipment is going to be more efficient, so prefixing my comments with modern would be backwards.

EDIT1:  Also, refer to meatball's post above.  EDIT 3: I don't believe ancow is a troll, our discussion continues in PM.

EDIT 2:  Also, note that Pentium D and Pentium 4 are the same family and see Phil21's post below.  And for the record, my first computer was an Amstrad XT that ran at 1.5 MHz; I haven't missed anything you refer to.

[The00Dustin]

When does cgminer disable and re-enable a pool?

I just checked on my miner and the pool was disabled for some reason. The pool came back up and cgminer didn't re enable it straight away.

In my experience, it recovers pretty quickly when a pool is down and recovers in a short period of time.  I haven't had a pool down for a long period of time.  However, also in my experience, when a pool is down when you start it, it might remain disabled permanently if you don't manually enable it.  Are you sure it didn't detect the pool as down during startup?

[Phil21]

Having operated hundreds of servers with the Netburst architecture (e.g. Pentium 4 based single cores, dual cores, and xeons) I can assure you load tdp is FAR different than idle tdp

[-ck]

When does cgminer disable and re-enable a pool?

I just checked on my miner and the pool was disabled for some reason. The pool came back up and cgminer didn't re enable it straight away.

In my experience, it recovers pretty quickly when a pool is down and recovers in a short period of time.  I haven't had a pool down for a long period of time.  However, also in my experience, when a pool is down when you start it, it might remain disabled permanently if you don't manually enable it.  Are you sure it didn't detect the pool as down during startup?

Exactly. cgminer never disables a pool itself unless it's not responding when you first start it. The reason it disables a pool that doesn't respond is that you may have simply passed it invalid credentials and there's no point flooding the network repeatedly trying to communicate with a server under those circumstances. If at any time a pool goes down that IS enabled, it diverts the load elsewhere, but does NOT disable it. It then checks it at one minute intervals and if it comes back to life will divert the load back. It never disables an enabled pool by itself.

[The00Dustin]

Now, if you're wanting to get into a debate of power usage, let's not forget that a CPU core is required to assist in running the GPU.

I already stated above that I agree with meatball and don't want to get into a debate on this here, but out of curiosity, are you running Windows?  IIRC, DiabloMiner is the only miner I have run on Linux that required anywhere near 100% of one core (and that is only because Java sucks).  cgminer uses 100% of one core and then some in Windows, but not in Linux (all GPU miners I have run in Linux draw less than 10% of one core [I don't remember anything more specific than that for the Linux ones]).  puddinpop's GPU miners are the only ones I have seen that don't waste CPU in Windows (I rarely see 1% of one core from them).  ufasoft's miner might not; I have never used it for GPU mining.  I am only guessing here, but the CPU usage in Windows on so many miners might be due to the porting of things like python and ncurses and curl.  It might not even be that the CPU is being used; Windows might requires the software to send an idle signal that isn't there because Linux doesn't, I've no clue.

[-ck]

Actually most of the CPU usage with GPU mining is related to driver stupidity. On Linux, you get close to zero CPU usage if you're using the 11.6 catalyst driver. However, they've ported the windows CPU usage to 11.7 and 11.8 drivers. So even on Linux if you use 11.7+, you get a lot of CPU usage. (lesson: roll back to 11.6). For those curious about -why- this is the case, they stupidly use an infinite number of sched_yield() calls to keep the cpu busy for a hot return back to the GPU. While this potentially can yield speed-ups through all sorts of means (during gaming for example), in mining it just chews up power and doesn't increase the hash rate in my testing.

[-ck]

Hi there...
I configured the program with ./configure CFLAGS="-I/usr/local/include" LDFLAGS="-L/usr/local/lib" on FreeBSD.
But when I type make next, it showed up like this:

make  all-recursive
Making all in lib
make  all-recursive
Making all in compat
Making all in jansson
Making all in ccan
gcc -DHAVE_CONFIG_H -I. -pthread -fno-strict-aliasing -I./compat/jansson -I./lib -I./lib   -I/usr/local/include -MT cgminer-util.o -MD -MP -MF .deps/cgminer-util.Tpo -c -o cgminer-util.o `test -f 'util.c' || echo './'`util.c
In file included from util.c:35:
miner.h:33:1: warning: "alloca" redefined
In file included from util.c:16:
/usr/include/stdlib.h:237:1: warning: this is the location of the previous definition
util.c: In function 'json_rpc_call_sockopt_cb':
util.c:268: error: 'SOL_TCP' undeclared (first use in this function)
util.c:268: error: (Each undeclared identifier is reported only once
util.c:268: error: for each function it appears in.)
util.c:268: error: 'TCP_KEEPCNT' undeclared (first use in this function)
util.c:271: error: 'TCP_KEEPIDLE' undeclared (first use in this function)
util.c:274: error: 'TCP_KEEPINTVL' undeclared (first use in this function)
*** Error code 1

Stop in /usr/home/users/xxxxxxx/xxxxxxxx/cgminer-1.5.6.
*** Error code 1

Stop in /usr/home/users/xxxxxxx/xxxxxxxx/cgminer-1.5.6.
*** Error code 1

Stop in /usr/home/users/xxxxxxx/xxxxxxxx/cgminer-1.5.6.

Could somebody please help?
Thanks

No one's ever tried to compile it for freebsd before as far as I'm aware. It looks like freebsd will need some special defines to get it working. While I can generically put some in in the next release in the hope it works, I can't guarantee it will be fixed.