|
Mixing AMD & NVidia is a recipe for disaster. You "could" (with custom BIOS, and Linux kernel) use 9+ NVidia GPUs on a single board but they are worthless for mining.
Honestly there isn't that much economic value to going >8.
A motherboard, USB drive, 2GB of RAM, and Sempron are dirt cheap. Maybe $200.
So going 8 GPU per board you are paying $200 overhead per board or $12.50 per GPU. Say you figured out a way to get 16 GPU on a board. Your overhead would be $6.25 per GPU. You "saved" a whole $6.25 per GPU.
Now lets say you built a 100 GPU farm. Your savings would be $600 (minus the cost of more expensive 8 slot boards, and any work requires to get 9+ GPU working).
When you consider the amount of cost, planning, wiring, cooling, maintenance, and troubleshooting goes into a 100 GPU farm that savings is a drop in the bucket.
|
|
|
|
- A minimum fee.
- A % fee for most cases (usually added to the minimum fee, but sometimes not if said % fee exceeds an amount) then...
- A maximum fee (because otherwise their clients would simply try their best to avoid transfers for such quantities and the processing cost for the bank is the same).
I have never been charged a % fee for a wire transfer, ACH, or check. Not in my business or personal accounts. A purely fixed fee is simple and attractive to miners but it would make a sufficiently small transaction too expensive. Therefore this fee should be very small, which is why I'm talking about a minimum fee (say 0.01 or even less). No it wouldn't. No reason fees would need to be larger than the minimum you are proposing. Paypal for example processes ~ 100 tps. At 100 tps and avg fee of $0.01 each that is $31 million paid to miners each year. At something more modest say 0.015 BTC avg ($0.06 per transaction) that is $200M. In today's hardware that buys about 200 TH/s of protection. Basing it on KB would have a similar effect. It must be based on KB to avoid spam. KB is the only true cost to the network. Charging it on anything else is just dumb. So high KB low value transactions (which have high cost) are charged a low fee and low KB high value transactions (which have a low cost) are charged a low fee. When I said obfuscation is not free in terms of resources, I meant that adding outputs means adding KB. Adding just 1 output is probably negligible, but then you're doing a really crap obfuscation job as adding 1 single address more to track is trivial. If you had to obfuscate/launder a significant sum in a big number of said steps, paying a fixed fee you'd be paying through your nose to have a decently sized tree. You always need a change address. Stupidly sending it back to the input address doesn't take any less space. Charging per kb already "solves" all potential complexity problems. Larger = more complex. It doesn't matter if it is because the person is cashing out 20,000 different inputs, sending it to 800 different people, obfuscating via multiple inputs or outputs, or using a very complex non-standard script. The cost is per KB. The price should always be KB. The thing is, if you make it on KB in a trivial way, then most people will make their transactions the simplest they can. Which destroys anonymity as any extra obfuscation would cost you. Obviously this depends on how much do you plan to charge per extra byte, but in the end you might end up with something very similar to what I describe above, in practice ([min,%/some function,max] ranges).
Using same change address takes exactly the same amount of space. Maybe you should read on how the protocol works before making suggestions to fix it? You seemed totally clueless that a change address even existed 4 posts ago. |
|
|
|
Apart from the aforementioned software limitations , you'll run to the problem of having not enough power. How much of a psu and motherboard do you expect to e able to run 8 full fledged gpus?
That is a non-issue. Dual PSUs and powered extenders solve it easily. The driver, OS, and BIOS limitations are essentially unsolvable though. |
|
|
|
|
The newer drivers (11.5+ ?) support 8 GPU in Windows. Older drivers only supported 4 GPUs under Windows.
32bit OS sometimes have problems (even Linux) with the full amount of GPUs. Others have reported problems getting full number of GPU under Windows XP. Using Windows Vista/7 x64 you can get 8 GPU (4x6990) working.
Nobody has gotten 9+ AMD GPUs working on any OS under any conditions and since AMD shows no interest in becoming a real GPGPU competitor likely never will.
|
|
|
|
You would be wrong.  |
|
|
|
Do you want to heavily obfuscate? this is not free in terms of resources, so you pay the max (fixed) fee.
Of course it is "free" (more accurately no more cost). It takes you the same amount of computing power to process the transaction regardless of the amount. % based fees will never be supported by a lot of Bitcoin members. Bitcoin transactions are more like checks (or wire drafts) than credit cards. No bank charges a % of the face value to process checks. They either process them for free, for a flat fee, or (in the case of businesses usually) on a per item basis. Processing a 1 bitcent transaction or a 1 million bitcoin transaction takes the exact amount of processing power, requires the same amount of bandwidth, and has the same amount of lifetime storage cost. The idea that one should cost 10,000,0000x more is stupid. A miner shouldn't be compensated more just because he got "lucky" and solved the block with a massive transaction. There is absolutely no need for % based fees. None. Having to give up obfuscation in order to "gain" something which is utterly unnecessary is a bad trade. THE CRITICAL RESOURCES IS KB. The space in the blockchain determines the cost of the transaction as it will need to be stored and will exist as tens of thousands of copies forever. Charging based on face value misaligns the cost to the network with the compensation gained. |
|
|
|
Is that a behaviour from the standard client or is it enforced by the protocol?
Obviously to make these x% fees possible that would have to go. Or you'd be optionally paying an x% obfuscation fee.
It can be changed but I would never support it. Never. Bitcoin isn't truly anonymous but the change function allows for psuedo-anonymous transactions. Using same address for change return will make transaction tracking trivial. % based fees aren't required and if that is the "price", it is a price too high. |
|
|
|
what about this variation:
- 5% fee on smallest input
- 0.001btc per input if there are more than two inputs.
this would also give an incentive for small transactions (in bytes) which is good for the network
So if (to use the same example above) I am paying 1 BTC from 100 BTC input I pay 5BTC in fees to pay you 1 BTC? Yeah that is going to be popular. just corrected it... please replace input with output  Then you can easily game it by always including a small output back to yourself.  At which point the only fee paid is a flat 0.001 BTC (per KB) which is why I was saying % based fees will not work. Fees need to be per KB not per input as that is the critical resource. |
|
|
|
what about this variation:
- 5% fee on smallest input
- 0.001btc per input if there are more than two inputs.
this would also give an incentive for small transactions (in bytes) which is good for the network
So if (to use the same example above) I am paying 1 BTC from 100 BTC input I pay 5BTC in fees to pay you 1 BTC? Yeah that is going to be popular. |
|
|
|
You are missing something. Bitcoin doesn't send it back to the same address. To obfuscate transactions the "change" is sent to a new address. Take a look at the example provided. http://blockchain.info/tx-index/3486210/f6f4e145a09ad579cacae3d889323054868a3acf10262ec4ae5c2132fd37365eInput address: 1Dv86AccHGjTJwKQtnqoN57T1CwCH4wZ9p 4.79 BTC (a prior transaction output) Output Addresses: 1Kae15Czf2s8FCRhFJ7QoXuRsPqT3a6pD4 1.775 BTC (likely the amount spent/transfered) 1AGK35na7rvgHS9vvG3RyE4VGvUerDbMH1 3.0169 BTC (likely the "change" address) |
|
|
|
The 99 BTC is going back to me. You do understand that you can't spend part of an input right? Scenario: 1) I have a wallet w/ a single address cointaing 100 BTC as a single prior output. 2) Fee is 1% of inputs 3) I owe you 1 BTC. 4) The only possible transaction is one where input = 100 BTC, outputs = 1 BTC (to you) & 99 BTC (returned to me) 5) The fee is 1% of 100 BTC = 1 BTC. My effective fee is 100% of what I am trying to spend. While your right it is technically a 100 BTC transaction no non-nerd is going to see it that way. The avg user will see it as "I have 100 BTC I want to pay muyuu 1 BTC and it will cost be 2 BTC = 100% fee. Fuck Bitcoin it is 50x more expensive than Paypal". A realistic example: http://blockchain.info/tx-index/3486210/f6f4e145a09ad579cacae3d889323054868a3acf10262ec4ae5c2132fd37365eUnless the sender had the need to send exactly 1.775 to one person and 3.0169 to another person, at the same time, and just so happened to have an input worth exactly 4.7924 (not more, not less) the odds are the transaction ISN'T for 4.79 BTC despite it having a 4.79 BTC input. Most likely (based on the amounts) the person is paying someone 1.775 BTC. If the fee was 1% of the input his EFFECTIVE FEE would be 2.7%. If his smallest input was 49 BTC it would be a 27% effective fee. If his smallest input was 490 BTC it would be an 270% effective fee. If his smallest input was 4,900 BTC it would be an 2,700% effective fee. If his smallest input was 49,000 BTC it would be an 27,000% effective fee .... |
|
|
|
blockchain... erm ... bloatchain will be 1000 terabytes by that time, and bitcoin would be dead cause you'd have get 10 empty hard drives just to hold it all...
In 1995 I went to college and bought one of the first 1 Gigabyte consumer hard drives. It costs me about $300 (more like $500 in 2012 dollars). Cost per GB = $300. In 2007 Hitachi released the first 1 Terrabyte hard drive for roughly the same cost. Cost per GB = $0.30 Hard drive capacity growth has slowed mainly due to demand for faster drives (SSD) still today you can get a 4TB drive for <$0.10 per GB. So say the rate of doubling has slowed to every 2.5 years (instead of 1.5 years for the first two decades). That is 16x capacity increase per decade. 4TB today, 64 TB in 2022, 1 PB in 2032. While eventually we will hit limits of magnetic but the only realistic constraint on flash density is cost. 3.5" flash drive is mostly empty space. Even with current technology you could make drives 10x a large (at 10x the cost). Moore's law will drive the cost of flash down very quickly. Flash drives cost ~$1 per GB (for slower models). Every 3 years that will fall by a factor of 4. <$0.06 per GB in 2020. |
|
|
|
|
The only potential scenario would be isolated locations. There have been instances where a country has become disconnected from the rest of the internet. Some smaller poorer countries may only have a few intra-national links.
If one is concerned the client could be modified to connect to a set of heartbeat nodes. Say 1 in Europe, 1 in US, 1 in Hong Kong. If no longer connected to all 3 there is a good chance you are isolated. The client could display a warning.
In western Europe or the US the odds of this happening as essentially zero though.
|
|
|
|
What's wrong with buying them directly on amazon with a credit card?
1) Seller steals credit card. 2) Seller purchases 10x $100 gift codes 3) Seller sells them to you (sucker) for irreversable currency (Bitcoin) 4) Customer reports card stolen. 5) Amazon reverses value of gift code. 6) You reverse Bitcoin transactions ... er wait. For the record I am not indicating or have any knowledge these codes are stolen just pointing out the risk. |
|
|
|
% fees are not possible (or more correctly accurate) because the network has no idea what the amount of the transaction is.
I want to send you 5 BTC. 1% fee = 0.05 BTC. No problem right?
Well my input is 32 BTC. So is it 1% of 32 BTC? =0.32 BTC fee?
The outputs are 5 BTC and 27 BTC (change)? Is it 1% of the larger output, 1% of the smallest output, or 1% of the average output, or 1% of sum of the outputs?
I don't think I understand what the problem is here. Surely x% of every input? the fee would be paid by the entity making the transaction. x% of every input = x% of the sum of the inputs. In other words the outputs would have to add up to at most (100-x)% of the inputs. So I want to pay you 1 BTC. The smallest input I have is 100 BTC. On a 1 BTC payment if the fee was 1% I would pay 1 BTC fee (on 100 BTC input) for a 1 BTC transaction. "cheaper than paypal, except when the fee is 100%"? |
|
|
|
Does this mean each node is setup to house 10 preferred miners at a time?
If so, I should add that to my site.
No. This has nothing to do with miners. |
|
|
|
Yes, but you've been running for awhile. When you ran your p2pool node for the very first time how long did it take?
I'm pretty sure I had 10 in under a minute when I first setup, too. My backup node is much newer and it got connections just as fast as the server that has been up since January. I have a public IP, so theres no UPNP on my primary. I thought that might make a difference at first, but my secondary node has the port firewalled and no UPNP either. It should have a harder time getting connections, but it still gets them just as fast. Maybe try adding " -n p2pool.stitthappens.com:8335" to your command? That should get you connected to me at the very least. Theres a list of public nodes on http://nodes.p2pmine.com/, but that lists the worker port instead of the p2p port. The first 10 are outbound. They are your nodes contacting other nodes. The "slow" part is the 11th+ connection. Your nodes only maintains 10 outbound and only repalces one of those when one is lost so the 11th+ are other nodes choosing yours. It takes a while. There is some luck involved but the network prefers older nodes so it may take some time (sometimes minutes, sometimes hours) before another node "picks" you. |
|
|
|
|
% fees are not possible (or more correctly accurate) because the network has no idea what the amount of the transaction is.
I want to send you 5 BTC. 1% fee = 0.05 BTC. No problem right?
Well my input is 32 BTC. So is it 1% of 32 BTC? =0.32 BTC fee?
The outputs are 5 BTC and 27 BTC (change)? Is it 1% of the larger output, 1% of the smallest output, or 1% of the average output, or 1% of sum of the outputs?
|
|
|
|
Do you read before you claim things are untrue.... Hardware Requirements
Intel® OpenCL SDK 1.5 requires support for the Intel® Streaming SIMD Extensions 4.1 (Intel® SSE 4.1) or higher (Intel SSE 4.2, Intel® Advanced Vector Extensions (Intel® AVX)). SDK works on the CPU's which contain support for the required instruction set extensions:
Mobile and Desktop Products:
2nd Generation Intel® Core™ i7 Extreme Processors
2nd Generation Intel® Core™ i7 Processors
2nd Generation Intel® Core™ i5 Processors
2nd Generation Intel® Core™ i3 Processors
Previous Generation Intel® Core™ i7 Extreme Processors
Previous Generation Intel® Core™ i7 Processors
Previous Generation Intel® Core™ i5 Processors
Previous Generation Intel® Core™ i3 Processors
Intel® Core™ 2 Extreme Processor, 9000 series
Intel® Core™ 2 Quad Processor
Intel® Core™ 2 Duo Processor, 8000 series
Intel® Core™ 2 Duo Processor E7200.
Server Products:
Intel® Xeon® Processors, 7500, 7400 series
Intel® Xeon® Processors, 5500 series
Quad-Core Intel® Xeon® Processor 5400, 3300 series
Dual-Core Intel® Xeon® Processor, 5200, 3100 series. http://software.intel.com/en-us/articles/opencl-release-notes/See any GPU in there? Intel has long supported OpenCL on CPU ONLY. I will wait for your retraction. If you still don't believe me install the SDK and try to run the samples on an Intel GPU. No need for a miner to prove it is possible just use their samples. No intel GPU to date supports OpenCL. No ifs, and, or buts. That may change in the future but given the lackluster performance I doubt Intel wants a direct comparison. They would much rather companies buying $8,000 server CPUs. |
|
|
|
|
Depends on how many? Some % of orphans is simply unavoidable but if your avg is similar to the network avg then your compensation is the same (i.e. you having 0% orphans and network having 0% orphans pays the same as you have 8% orphans and the network having 8% orphans).
If it is significantly more than the average then more connections may help. The 10 is only on outbound connections. If you portforward you can have more than 10. 10 outbound + unlimited inbound.
|
|
|
|
|
Show Posts
