No issues over the past 10 minutes or so; that seems to have worked.

Would it be possible to add rejected/stale stats for the current round in addition to lifetime without incurring too much overhead?

From cgminer:


Seems to last about 20-30 seconds each time. Bandwidth, latency or processing?

I've been following p2pool also and am running a cpuminer on it. Scalability is still concerning; how many mini nodes can branch off and still maintain reasonable propagation? The structure just seems too flat. I could see advanced improvements in network latency alleviating that to some degree.

Donations aren't forced. That's why this would be called a fee.

Getting caught up in semantics is pointless - the issue still remains that the pool has operation costs that need to be covered.

Yes, about 0.63% would cover the 13.5BTC or $270/month, all else being equal. Hmm... I think I'm going to increase my donation...

Hey BT - just go for a 1% fee. It's a good buffer if BTC languishes around the current price level and you can always lower it as the pool grows.

Please feel free to correct me if my assumptions or math are wrong in any way. From the pool stats page, the latest round currently has about 3.86 million shares valued at 0.000029570125 BTC per share and a donation weight of 0.4480.

3860000 * 0.004480 = 17292.8 shares
17292.8 * 0.000029570125 = 0.511350258 BTC

This has been a very slow round, taking over 18 hours to find a new block. If the frequency remains at an average of 1 block every 24 hours, then:

0.511350258 * 30 = 15.3405075 BTC/month

At the recent Mt. Gox exchange low of approximately $13/BTC, that comes out to roughly $200/month.

I don't know what the server costs are, but from the above analysis, 0.5-1% fee implementation seems a very reasonable rate to cover bandwidth and processing expenses, as well as at least some of BT's time spent on the pool.

Aside from that, it would appear that the decline in USD/BTC rate has met with enough demand to suggest a bottom. As the rate rises, the incurred overhead costs should marginalize further, providing a much-deserved increase in return for BT.

This is personal preference of a purist nature and anyone unhappy with double floats is welcome to find another pool. This is along the lines of the 80/20 rule wherein 80%+ of the complaints are raised by 20% or less of the participants.

Other than the recent disconnects, Ars and Eligius are the most consistent pools for me. Thanks for putting in the effort.

I don't suppose former Seattle residents count...

Annoying yes, but well worth it if you get them working properly. The worst issue right now is that there can't be more than a 125 Mhz delta between the core and memory clocks. Pushing the mem clock down from 810/1250 to 810/685 brought the temperature down by over 10C, so the memory is definitely a problem there. Other than that, they're good cards.

I'm using 2x6950 Asus DCUII 1Gb cards; both are set to 850/725 @ 80% fan speed using the following:


You can also use aticonfig, I've just come to like atitweak.


The cards are currently 68C and 71C with no case and only a weak box fan that does a lousy job clearing heated air from the dead space between the cards. Dropping fan speeds down to 60% brings temperatures to 72-76C and 50% goes to 75-78C.

Currently running Catalyst 11.6/APP 2.4SDK and Phoenix 1.50 using the latest phatk kernel with options "VECTORS BFI_INT AGGRESSION=12". They're pushing ~345 MH/s with those settings. During cooler times of day, they get kicked up to 880/755 @ 80% which gives ~360 MH/s and temps of 70-74C.

I have yet to attempt flashing the BIOS or changing voltage. Even with all else being stock, I can push them to about 920 Mhz and brush up against 400 MH/s, although the temp gets into the 80s and they aren't completely stable. With voltage control and the ability to underclock mem more than 125 Mhz below core, I'm fairly confident I could approach 1 Ghz and easily exceed 400 MH/s while still keeping the temp around 85C or below.

It's tempting, but I don't want to install Windows (outside of a VM to run RBE) to make those gains. I'll be testing Hashkill soon to see if that offers any improvement. Otherwise BIOS flashing seems to be in order - and maybe some Arctic Silver. At least that can be done with a FreeDOS USB drive and a screwdriver.

Using aticonfig, my 6950 took the memory underclocking the first time I tested it at 900/300, but only for a brief time. Since then, the best I can do is the same as everyone else - mem no more than 125 below core speed. Underclocking the memory might not do a huge amount for power consumption, but it significantly reduces heat generation.

I have read that it is possible to set and commit the core/mem speeds in Windows, then switch to Linux and use them with the same configuration. I'm not sure whether it will fall apart if any changes are made from within Linux and I do not have easy access to Windows, but it could be a funky workaround.

Other than that, I'm trying some BIOS mods. This is easily done using RBE from within a virtualized XP session. There are some limitations as the virtualization abstracts the video adapter, leading to RBE and GPU-Z not being able to verify certain BIOS changes, particularly shaders. Speaking of which - does anyone know where to get the shader information from within in Linux? One of the ATI tools, maybe?

Also hoping for cooler weather to overclock easier...

Funny, I was looking into this a few years ago when micro-payment ideas were being floated.

Bitcoin lends itself perfectly to this, especially being completely detached from any one region. Micro-payments are very easy to facilitate with BTCs and it should be highly feasible on an economic basis. There might not be much of a payout, perhaps enough to cover electricity costs, but the concept is definitely sound.

If I had the time...

Guaranteed! The community is small enough that most can understand and at least figure out what they're doing. As it grows, that in-depth comprehension won't be as widespread.

You could always make the BIOS flash image a separate supplemental download. It'd be great for an old 128Mb flash drive.

Agreed - couldn't hurt to try 11.6, though.

Impressive for <50Mb, even if it is still alpha quality.

Choosing between these two, I took the Corsair - the difference of 50W wasn't worth the extra cost for the CM.

Cooler Master Silent Pro Gold (1000W)
http://www.newegg.com/Product/Product.aspx?Item=N82E16817171056

Corsair Enthusiast TX950 (950W)
http://www.newegg.com/Product/Product.aspx?Item=N82E16817139013

HTTPS all the way.

Exactly. It's kind of like the difference between a dividend-paying stock and the next high-flying tech darling - not sexy, but very reliable. Consistency and stability will be much more important as the Bitcoin environment develops anyway.

SMPPS +1

Yeah I've been looking at a 2-4x 6950 setup, but have been hesitant to set up a dedicated machine for mining over a system that I can at least accumulate BTCs with. Now that the exchange rate steadily dropping, Mt. Gox either needs to get its sh*t together or holders need to stop selling before I make the plunge. Time will tell if this is a retest of the $10/BTC range or a deeper sell off.

Of course, I'm not as concerned about using BTCs to pay for the hardware as I am with simply getting out of the USD. As (not if) the dollar continues it's accelerating decline, Bitcoin could behave in a similar manner to gold by going up rapidly (just wait for the QE3 announcement...). This will be even more pronounced as the US locks down on financial transactions across borders and more people realize their dollars are becoming relatively worthless.

On that note, using WalletPaperbackup you can print several hundred Kb of scannable data in hard copy for subsequent recovery. No need to move coins out of the country or risk confiscation of flash drives and other electronic media. Just slip a piece of paper into a book or magazine and endure the cavity search.

Bitcoin may be having a rocky start, but I think the above will build demand to a point where it may be one of the only routes for wealth to escape from the American kleptocracy, so simply having Bitcoins is better than not.

Other than having to install the required libraries and JWasm, I had no problems building v0.10 on a 32-bit Ubuntu 10.10 system. Running it on my 64-bit systems was a non-issue with ia32 libraries installed, but building on a 64-bit system took some tweaking.

Comparing the 32-bit Ufasoft bitcoin-miner to the jgarzik minerd using sse2_64, I'm seeing about 10% improvement. On the 32-bit system which did best using the cryptopp algorithm, switching to Ufasoft's miner produced about 30% greater Mh/s. Granted, this is still CPU mining, but the delta is interesting.

Obviously, you'd need the Linux source for the miner.

For a 32-bit Ubuntu 10.04 or later system (gcc 4.4 is default in 10.xx, 4.5 in 11.04), the required 32-bit C++ build libraries are:


If preferred in the above command, you can use either the OpenSSL (libcurl4-openssl-dev) or Mozilla NSS (libcurl4-nss-dev) version of libcurl instead of GnuTLS.

For JWasm, I simply extracted the binary to the build directory, made it executable and added the current working directory to the path:


If it isn't the default and you get a compiler version error, to avoid having to change the entire environment from gcc 4.4 to 4.5, rocksalt's method of specifying the compilers in the command with the CC= and CXX= definitions works best:



With 64-bit systems, instead of creating a chroot environment to build 32-bit binaries, you can just add the necessary dev files and make a few tweaks. First, make sure the 32-bit sources are installed:


This isn't pretty, but it works - in configure, search for JASMFLAGS and add "JASMFLAGS=-elf" immediately after the section shown to make it appear as so:


Make sure the compiler is going to generate a 32-bit binary and run configure:


The real sticking point is the curl header file curlbuild.h - it has a fixed defined value that will prevent a 32-bit compile. We can fix that for now by editing it.

WARNING! This isn't going to hose your system, but it might cause issues with further compiling efforts if you don't undo it after building the 32-bit miner.


In order to reverse this action, simply do the following:


If curlbuild.h isn't there, you either don't have libcurl installed or it's somewhere else. Just search for the file and replace the relevant lines.

Now you can compile successfully.



Out of curiosity, I'd like to find out whether there'd be a statistically significant increase with a native 64-bit implementation. Too bad I haven't touched assembly in years...

I agree; it has been proven. I think economics in general today is too concerned with valuation when it needs to focus as much or more on motivational psychology and global capital flows (to be fair, at higher levels economic studies do spend a lot of time on investor psychology [while most schools ignore the general consumer/participant perspective], but those efforts are currently being used by major banks and governments to game the system in the short-term rather than improve stability and sustainability [it is possible to have stability without sustainability] over the long-term). I'm amazed that last sentence made sense.

A better debate might be on appropriate usage of motivational mechanisms. Motivational factors are different between guaranteed or immediate incentive and potential compensation. Ex post facto socially-based reward can take many forms, all of which can be categorized as different currencies acting as money (where money is the abstract concept and currency is a manifestable form of money). In other words, creativity might not do well when incentivized by money, but that same creativity can be monetarily compensated for after its creation regardless of whether there was an expectation of reward or not.

The end result of this is still that money is not to be faulted; rather, inappropriate or poor utilization of money leads to detrimental effects on society. That also has a lot to do with the contemporary monetary system (such as fractional reserve banking).

The ZM/VP makes no distinction between money and a system that makes use of it when they are distinctly separate components, just as the internet (and even below that, electronic computer technology) is the underlying framework upon which Bitcoin is based. Their argument is like suggesting that computers (money; base concept), and thus the internet (fractional reserve banking; system based on money) and Bitcoin (currency; functional manifestation of money within the framework system) by extension, should be banned because some people have been able to manipulate the internet with nefarious intent. They throw the baby out with the bathwater.

Widespread awareness and education in this respect will largely mitigate the current level of ignorance regarding how money works, trending toward beneficial systems (such as Bitcoin which can both eliminate the fractional reserve banking system and replace the centrally-managed currencies dependent upon that system) instead of negative ones. This is already occurring (Greece and Japan flocking to gold after government ineptitude at preventing and responding to their respective disasters, for example) and will build momentum over time as government-sponsored monetary systems approach outright failure.

Be patient and read some of Martin Armstrong's work on long-term cycles.

Ars reader from the beginning (early 1999) hopping on-board. CPU mining only atm. Nice, simple pool interface - good work, BT.

A couple of questions:

How is the worker activity determined? I have a few that are consistently running, yet show as inactive on my account details. When I'm away from the machines, it'd be nice to see which ones have hit a snag. If it's just due to my painfully low Mhash rate, c'est la vie.

I know it isn't the best option, but what would an ideal laptop GPU be for mining? I haven't seen much discussion along those lines and I value the Ars community opinion higher than most.

Thanks!