Bitcoin Mining GPU Performance Comparison @ [H]

Now, I've read about this bitcoin before and still don't quite understand it fully. Maybe someone can give me a quick summary about it.

1.) How do you earn money? Or, more specifically, where does the money come from?
I know you can't create money from nothing yet I see it's possible to convert bitcoin to real currency.

Yes, pretty much into any currency you want from a variety of exchanges. See here
2.) I see that bitcoin is similar to folding, is that entirely true?

Somewhat I guess

3.) Importantly, is this even legal? Taxable even?

Yes it's legal, NO it's not taxable. EDIT: Ehh short answer I guess is not if you leave it at bitcoins. See this post from someone who seems to know more about it.

4.) Can you purchase with bitcoin alone?

Yes, check this list of sites that accept bitcoin as payment. There were more than I expected the first time I looked at it.

5.) How does this even affect video card pricing? I've seen a few posts here on [H] mention that bitcoin crashed/hurt the video card market or something along those lines.

As has been said, driven up some video card prices.

Thank you.

My answers in green.
 
Last edited:
Pretty sure his answer to 4 is BS though. I don't see "lots" of stores accepting bitcoins, and I don't know what "bitcoin related items" even means. They might be destined for black market internet currency, but don't expect to be using them on Amazon ever.

See my answer to 4. A decent amount of sites that will accept bitcoins purely for payment.
 
Pst, your fanboy is showing.

In gaming, both camps are pretty equal... so I'm not really sure how I'm a fanboy. I use whatever gives me the best bang for my buck.

And now AMD's allowing me to get some bucks back? Sign me up! Not saying nvidia doesn't, it's just not as efficent.

Oh, and my last card was a GTX465, I now have HD5830CFX.
 
Very interesting article. Not interested in Bitcoins at all, but the wattage/performance graph did help me in considering how big a PSU I might need for my upcoming rig...
 
Ok, how can I pull a Bernie Madoff with Bitcoin? J/K
This shows some superior OpenCL performance of GPUs but with ATI in particular.
I wonder if Nvidia can address it with a new driver?
Still wanting to see how BD does with this compaired to i7 2600k.
 
Ok, how can I pull a Bernie Madoff with Bitcoin? J/K
This shows some superior OpenCL performance of GPUs but with ATI in particular.
I wonder if Nvidia can address it with a new driver?
Still wanting to see how BD does with this compaired to i7 2600k.

I don't believe they could do anything significant. It's just because of the way Nvidia does it's architecture, it would have to be completely redesigned pretty much. Nvidias 512 compared to AMD's 1536. With bitcoin mining, efficiency of cores/shaders doesn't matter as much as the amount on the card. ( someone correct me if I"m wrong please)
 
Be careful that you double check what your electricity rates are. Here in Los Angeles, depending on what tier you fall into, you can easily pay .13 cents/kWh which would cut the projected profit in half. $60 minus $27 in power costs for a 5770 using 300 watts from the wall and producing 200mHash/s.

Further south in Orange County, SCE has a tiered pricing rate. The more you use, the higher the price. With the Air conditioner runnign during the summer, it's easy to hit the 28 or 31 cents/kwh rate. With that pricing, you would actually be loosing money.
 
Ok, how can I pull a Bernie Madoff with Bitcoin? J/K
This shows some superior OpenCL performance of GPUs but with ATI in particular.
I wonder if Nvidia can address it with a new driver?
Still wanting to see how BD does with this compaired to i7 2600k.

Its a hardware difference, not software. There is an important instruction that AMD GPUs are able to do in one pass, versus several for NV GPUs when doing SHA-256 cryptographic hash. There are a lot more ALUs in AMD GPUs. Basically. Efficiency does matter. NV can't run their ALU clocks high enough to compete because they are more complex. Only an architectural change will change the result.

Read - https://en.bitcoin.it/wiki/Why_a_GP...CPU#Why_are_AMD_GPUs_faster_than_Nvidia_GPUs?
 
i have a 6950 with a q6600 i wouldn't mind joining , does hardocp have a group ? I'd trade in the points for visa gift cards and use it to buy a new rig. I currently pay nothing during the day to power my pc
 
i have a 6950 with a q6600 i wouldn't mind joining , does hardocp have a group ? I'd trade in the points for visa gift cards and use it to buy a new rig. I currently pay nothing during the day to power my pc

I don't believe so. This is a pretty good list of pools currently and the different ways they work. If you do hop into it, trading directly into newegg would probably be better than getting a visa, strictly for hardware reasons. Unless obviously your buying elsewhere :D
 
I started mining because of the thread here on [H]ardforum a few weeks ago about 5870 prices still being in the $300 range on ebay because miners were buying them up.

Well I purchased five Gigabyte 5870s from Newegg last week for $239 each when they miraculously had some in stock (must have been an old shipment or cache?) and overclocked them all to 1000MHz core. They are doing 430-440 mhash/s.

I am combining my system with a friend's. We'll have eight 5870s by the end of this week which equates to around 3.44 ghash/s at a GPU cost of under $2000.
 
yay I love having a gtx 285 lol oh well guess since I barely use it I'll go at the steady 60 mhashes/s lol. Would be nice to get something better for mining, but at the same time retaining performance for games.
 
Ok, how can I pull a Bernie Madoff with Bitcoin? J/K

In a way, Bitcoin IS a Bernie Madoff-ish scheme. It's a system inherently advantageous to the creators and early adopters, who could farm coins when the effort per coin was much lower. It will crash because 90% of people are farming them for this reason:
Stick to selling them for money.


That is the #1 example of how to spot a bubble. In real life terms, think back a few years when million of people were trying to flip houses for profit. They even had TV shows about it. People didn't bother to understand the underlying market and economics, and while a lot of people made money, those still holding property when the game of musical chairs ended got screwed.

So...have fun with this. Make a few bucks if you want to, but be prepared for it to evaporate at any moment. Don't make long term decisions on it. It doesn't have the foundation to sustain you.
 
. A lot of people pulled out after MTgox had people's money disappearing as it hacked or he did it himself and came up with the story to cover for it, anyway some one sold off 4000 bitcoins which drove the price down to pennies but was brought back up by the people buying it while the price was low figuring the price would go back up.

No When MTGox figured out what happened they stopped all trading and eventually reversed every ones transactions and the market to before the hack. When gox opened trading again coins were still at $17. The price was not driven up by buying.
 
Something to realize here - the more people jump on the bitcoin mining bandwagon, the more processing power it will take to generate each bitcoin. The network is adaptive that way. If you're a miner and you're encouraging others to mine, you're hurting your own profitability. Ideally, you'd want more people using the currency, but not mining it, to maximize your individual profit.

It's fairly apparent that a lot of people don't understand this.
 
With the Air conditioner runnign during the summer, it's easy to hit the 28 or 31 cents/kwh rate. With that pricing, you would actually be loosing money.

5.98 cents per kWh
BC Hydro FTW :D
 
Great article on Bitcoin mining! Considering that i'm still a bit miffed that Stanford won't give us anything on the ATI side that makes it worthwhile to turn my dual 5850's on, at least there is some use in the form of Bitcoin mining. Of course I have no desire to join this craze, but it is definately neat to see as this is a real life social experiment that i'm sure many Economists are observing.

Something to realize here - the more people jump on the bitcoin mining bandwagon, the more processing power it will take to generate each bitcoin. The network is adaptive that way. If you're a miner and you're encouraging others to mine, you're hurting your own profitability. Ideally, you'd want more people using the currency, but not mining it, to maximize your individual profit.

It's fairly apparent that a lot of people don't understand this.

You pretty much nailed it as i'm not sure everyone understands this. You want demand for the currency to go up if you are going to turn a profit. The more people mine, the quicker the difficulty will increase at which point it becomes harder for the existing pool of miners to obtain bitcoins. The interesting thing though is that there will be a set number of bitcoins ever in existance which will prove to be an interesting experiment.
 
Does anybody else see this as a massive waste of computing power? Why not award bitcoins for doing something of worth, like Folding......
 
Does anybody else see this as a massive waste of computing power? Why not award bitcoins for doing something of worth, like Folding......

Bitcoins are a digital currency and wouldn't exist without miners supporting the network because that network is required to send/receive a bitcoin transaction.
 
Bitcoins are a digital currency and wouldn't exist without miners supporting the network because that network is required to send/receive a bitcoin transaction.

Ya but your computer has to do equations to 'mine' these bitcoins. Why couldnt the equations actually be folding algorithms?
 
Ya but your computer has to do equations to 'mine' these bitcoins. Why couldnt the equations actually be folding algorithms?
Because it is verifying the hash of recent and ALL past transactions. Without this verification, someone could easily fake transactions.

When you send coins to an address, they don't show up immediately because it hasn't been verified in a transactions block, which is then distributed across the network and hashed continuously by miners.

My first transaction has been verified over 3,000 times so far.
 
Can someone tell me what the fuck is a MHASH?
And what are/is the GPU actually calculating, and for what purpose?
Aren't you concerned you may be running programs to hack into something?


foil hat placed..........;)

Since no one answered you concisely yet... :)
Mhash is mega-hash, it's the hash rate that your GPU "mines" at to verify blocks of transactions.

The calculations are constant verifications of past transactions, and new transactions are put into large blocks and their hashes are verified, and once verified, they show up in the other user's Bitcoin wallet.

It's been verified by a lot of people what it's doing, but someone DID just get banned from BTCGuild's pool because they were using a large botnet of computers they hacked into to mine for them!
 
Pretty sure his answer to 4 is BS though. I don't see "lots" of stores accepting bitcoins, and I don't know what "bitcoin related items" even means. They might be destined for black market internet currency, but don't expect to be using them on Amazon ever.

You didn't see any stores accepting bitcoins because you didin't look and you don't know what bitcoin related items are because you lack basic common sense.
 
If I would have known what bitcoins were 6 months ago I might have gotten 4 GPUs and did this but at the current difficulty and market prices I just don't know. It definately is not worth is as a nighttime+work crunching rig with my 6970 at 1Ghz.
 
Now if ATI could use that massive power for physics in Batman Arkham City I will be overjoyed.
 
I've been mining with the systems in my specs for little more than six weeks now. -v -w128 -f10 in GUIMiner, doing about 1.2 Ghash/sec, all being fed to Slush's Pool.

I've ground out about 60 BTC so far but I'm down to barely a single BTC a day now. Difficulty being what it is, pooled mining is the only way to go. You need something like 5-10 Ghash/sec to go solo efficiently, and that number goes up every time the difficulty does. Might still find a block but it'd be like winning the lottery.

I'm gonna shut it down when I get to 100 BTC or so, which'd pay off my new rig at current prices... no idea if they'll be worth more than that someday. They're honestly only worth the amount of electricity I put into them. Which is about $2 a BTC at my electricity pricing of 0.08 kw/h, with market pressure driving them the rest of the way up. It'll be a while before mining BTC using GPUs isn't profitable, though.
 
(Revised)

Stock clocks/voltage.


bitcoins2.png

This is how brent had guiminer setup. (basic opencl + flags)

bitcoins.png

This is how I had it running.
phoenix -k phatk VECTORS FASTLOOP=false AGGRESSION=7 BFI_INT WORKSIZE=128
 
Last edited:
hmm bought my 5870 video card used, guess it might have gone up in value since I bought it. So rare that something electronic goes up in value.
 
In a way, Bitcoin IS a Bernie Madoff-ish scheme. It's a system inherently advantageous to the creators and early adopters, who could farm coins when the effort per coin was much lower. It will crash because 90% of people are farming them for this reason:

That is the #1 example of how to spot a bubble. In real life terms, think back a few years when million of people were trying to flip houses for profit. They even had TV shows about it. People didn't bother to understand the underlying market and economics, and while a lot of people made money, those still holding property when the game of musical chairs ended got screwed.

So...have fun with this. Make a few bucks if you want to, but be prepared for it to evaporate at any moment. Don't make long term decisions on it. It doesn't have the foundation to sustain you.

I agree.

Bitcoin is a digital currency, backed by nothing. There is no inherent value in bitcoins, there is no government giving it value. The only value it has is based on the fact that the supply is limited... except that the supply ISN'T limited. Anyone can make a Bitcoin 2, or a Bytecoin or an E-Coin. Your new bitcoin system can churn out 1 million coins a day instead of the 7000 or so which the current bitcoin network produces.

Bitcoin is a speculative bubble. It is barely being used as a currency at all, and nearly all the buyers are speculators who are hoping the price will continue going up exponentially, as it has for the last year (until it levelled off in the last month).

There are two possibilities here:

1. Bitcoin use continues to spread until finally many major retailers accept it as currency, and it becomes a legitimate accepted global currency.

2. Bitcoins stop going up in value for long enough that the speculators give up and stop buying. As they're the only buyers, the price collapses down to nearly nothing. The miners give up, and bitcoin is no more.

Number 2 is much more likely than number 1. I would advise any of you miners to sell all your bitcoins as soon as you get them, as the long term value of bitcoins is very probably zero.
 
Thank you Brent (and certainly Kyle of course) for publishing what is perhaps the most original piece on GPUs I recall reading in years...! It is such a breath of fresh air that I had to go back and read it again to make sure I wasn't thrashing around unconsciously inside some unresolved 1980's technical daydream of mine....;) It's nice to see this kind of thinking "outside of the box" going on these days, but it is also frightening to realize just how little of this kind of thinking remains front and center in the public venue today.

To clear up the first possible ambiguity here: I don't especially see any intrinsic value inside a Bitcoin. The real value to me in comparisons like these is the disposition of differing kinds of work accomplished by competing GPUs under what are clearly non 3d-rasterizing workloads and conditions. IE, the "GPCPU" hybrid performance projections that have been so grossly misstated and hyped over the last couple of years--as if, well, as if literally it was more important how much non-rasterizing, non-3d work a GPU can do than how much 3d rasterizing work the same GPU can do in a similar time frame!

This article highlights what should have been up to this point in time the very obvious notion that 3d GPUs manufactured by different, competing companies are going to have some very different performance characteristics concerning the various tasks that the respective manufacturers have resolved important enough to design their architectures to be able to facilitate. In this particular work scenario, we see AMD GPUs walking all over nVidia GPUs of the same general class and properties in these non-rasterizing non-3d, so-called "GPCPU" workloads.

However, and this is where it gets interesting, there are other non-rasterizing, non-3d rendering GPU workloads in which the nVidia architecture and driver set seem to clearly indicate that as a GPCPU workhorse, the nVidia GPUs are clearly the superior choice. (Like Tess Mark's synthetic tessellation benchmarks, for instance.)

What's interesting is the fact that even in 2011 we are still very much slaves to the synthetic benchmarks and other marketing paradigms published abroad by self-interested IHVs to create appearances of positive market conditions so as to upwardly stimulate sales--even when such sometimes desperately clearly spurious marketing data is thoroughly revealed to have little or no impact on an end-user's use of those products in his daily software and hardware environs and regimens. It's interesting that in 2011 we can still be manipulated into providing the very general assertions the IHV wants publicized. This isn't the fault of the Internet journalists in the tech sites for the most part, the ones working on up-to-date product reviews--it's the fault of the hardware IHV who will do and say practically anything to have his products publicly promoted at the expense of his competitors. Truth and fiction in this marketing exploitation are hazards to be skillfully navigated as opposed to ethical practices.

So let me say again--kudos and props to Brent & Kyle for pulling these product reviews together into a sensible package wherein common sense is the virtue and hubris is the flaw. But I want to end on the sort of sad note that [H]'s competitor sites seemed to have just caved in on the investigatory side of the product mix, and these days seem content to bask in the warming sunshine of Mac IHV somnolence.

Yet, all around us we see a truly growing and alarming frequency of IHV sleight-of-hand in their respective marketing battlegrounds. This is not a time for Internet pundit laziness and capitulation, it's time for brutal honesty and calling a spade a spade. That something is going on in a big way is surely plain to see. Recently, AMD, VIA and nVidia, reportedly, resigned en masse from BapCo after decades of partner support because they felt and so stated the benchmark (Sysmark 2011) had ceased even a pretense of objectivity, with key portions of the bench weighted so far towards the strengths of Intel products and the weaknesses of AMD, VIA, and nVidia products that Intel inevitably and inexorably carried many benches away even though it had actually scored lower in the preponderance of the tests.

Interestingly, the FUD & damage control FutureMark sought to counter with indicated that AMD's reasoning for quitting the bench was "sour grapes" because BullDozer wasn't going to ship in a condition remotely competitive with Sandy Bridge. If true, then AMD's position is both understandable as well as logical, if reprehensible. However, this spin on events from FutureMark does nothing at explaining nVidia's problem or VIA's. OK, long day today...;) If I don't sign off I'll be carried off....:D .
 
Seems like a good candidate for oil submerged cooling, and just stick it outside or in your garage.

Students living in dorms have a very unfair advantage in getting free electricity!
 
shocking difference between the fastest cpu, and the cheapest gpu. I knew there would be a huge gap many times over, but such an exponential advantage was not expected. this really justifies those applications like gpu based physics, for all those physx haters out there.
It doesn't justify anything. Different workloads require different coding techniques, and work more efficiently depending on the hardware. You can't make a blanket statement that because Bitcoin mining runs much faster on the GPU, that in game physics are automatically going to run by the same factor using the same GPU. They may if all you are doing is physics processing, but load the GPU down with other rendering tasks, and the scenario changes quickly. The whole point of physics is to use whatever hardware/processor cycles are available, this is what makes DirectCompute and OpenCL so interesting, it allows this.
QFMFT

I always laugh at people who think the CPU can handle decent physics. It fakes decent physics for sure, and the reason it seems like the GPU doesn't offer any better physics is because of two reasons
Again, you can't look at the speed of Bitcoin mining and extrapolate the same advantage for physics calculations. The GPU is ONLY calculating the hashes, nothing else. Game workloads are much more complex and diverse. Good coding uses all processing power available, it doesn't shoehorn it into a specific piece of hardware. Plus Bitcoin mining just happens to work extremely well on the GPU.
Does anybody else see this as a massive waste of computing power? Why not award bitcoins for doing something of worth, like Folding......
Can you point me to advances and real world benefits that have occurred from folding applications? I keep hearing how great it is, but I never see any tangible results.
 
amazing review guys, hardOCP u all rock! everything I coulda ever wanted in a bitcoin hardware review. this site never ceases to amaze me :D
 
phoenix -k phatk VECTORS FASTLOOP=false AGGRESSION=7 BFI_INT WORKSIZE=128

Hey, I don't like arguments, and why are people using me anyway? :p

All kidding aside, this whole bitcoin thing is strange to me, but seeing how the AMD GPU's are outperforming Nvidia's in this it makes me wonder what other differences between the GPU's will show up in the future. I would hate to see the development focus shift from game performance to alternative function performance, though I could see potential for improvements in one area to affect both.
 
I just checked and its $.058/kw where I live. I might have to setup a machine with one card and update as the money arrives. If it doesnt work out, then I can use the machine for something else. I guess the key to starting out is building it cheaply (using as many spare parts as possible) I have a leftover machine that can supply the case and I have a spare PSU. The board wont work (AGP slot) Dont worry, the PSU is a new SeaSonic. If it doesnt work out, the machine would end a decent gamer!
 
Back
Top