Search content
Sort by
Showing 20 of 15,656 results by DeathAndTaxes
My heating unit mounted on the wall. 3.5GHash @ 1700W
I like this! I think I should mount my FPGAs on the wall, too, but they'll be a terrible heating unit.
100 of them lined up in a custom case would make a nice base board radiator.
PXE/NFS diskless
We need to talk.
Once I get some of my hardware setup I would like to talk about your PXE setup.
Since everyone is in the sharing mood today.
As a GPU miner I saw this ...
is that weird?
@ The-Real-Link
I can somewhat understand why someone would want to use cases.. while it costs money and increases temps, it definitely improves Feng Shui and makes moving things around a bit easier. But why on earth would you equip every machine with a keyboard and monitor? And a HDD?
Even if you are new to linux, you really want to check out BAMT:
https://bitcointalk.org/index.php?topic=28967.0
I can somewhat understand why someone would want to use cases.. while it costs money and increases temps, it definitely improves Feng Shui and makes moving things around a bit easier. But why on earth would you equip every machine with a keyboard and monitor? And a HDD?
Even if you are new to linux, you really want to check out BAMT:
https://bitcointalk.org/index.php?topic=28967.0
Heard BAMT is pretty neat so may look into that. Even though all the systems are together now, you have to keep in mind I built one at a time and I couldn't guarantee they'd be near each other in the early stages so I'd have to monitor each one on it's own and to set up the OS and such. Again, complete rigs that could be used for things outside of mining. Guess I'm also a bit of a tidy person...not neat freak - just tidy
. The monitors aren't super expensive though so no biggie. Maybe in the future if I do more rigs I'll have just the systems but again if they're used in different places at once or by anyone outside of us techies here, they'd probaby like a monitor IMO.Couldn't you just buy a monitor, keyboard, mouse, extra ram, aftermarket CPU cooler, etc if/when it is no longer used as a rig? Still they look very nice.
While the monitors may be "cheap". Cheap" monitor, case, keyboard, mouse, aftermarket CPU cooler, extra RAM, higher end CPU, HDD, etc = lot of non-hashing cost. Could have built 4 3x5970 rigs for that price. That is why they are called "rigs"
vs
It's almost fully covered... too much heat will stay inside.
Force air cooling. If you force cold air IN unless you trapped a blackhole inside the hot air will be forced out. Now you are right if you just slapped together some shelves and cards that would be bad. It would require some "ducting".
Don't laugh at my ASCII "art".
Code:
hot air out pulled out
^
| |------------------------|
| <- bank of GPUs <- |
| |----------------------| |
| <- bank of GPUs <- |
|------------------------| |
| |
^ (cold air forced in)
All the ^ are fans indicating direction of airflow.
Only two banks shown but more could be used.
Hooking all the fans to a fan controller with some temperature probes in each bank could be used to optimize cooling. Upper bank would need higher airflow because air is lazy.
Sorry, don't mean to show off... ok maybe a little
Yeah your rigs are the inspiration for the ideas I got in my brain. I was thinking of having the twist of mounting MB "standing up" so the GPU "exhaust" (I use 5970s exclusively) points "up". Best way to describe it would be blade server meets mining rigs.
Cold air would flow under the bottom of MB, be pulled into GPU intake, warmed up, and blown out the top where it would have nowhere to go except the exhaut out. I wanted to makes some changes to my garage anyways and now my wife has seen other people have "organized" rigs it might happen soon rather than later.
Why add a second powered riser? I've been curios as to what these are used for.
I'm really curious to see what would happen in the long run.Touch the single 12v wire going to your motherboard connector to see if it's hot,
It may be providing ~45 watt per GPU, or 30 amp on one wire. (rated for ~8amp)
(45w according to http://blog.zorinaq.com/?e=42 using 5970)
I personally use powered extender when going past 5 gpus.
I'll check it out when I get home, you've got me curious now. This board does allow me to run 2, 8 pin 12V ATX power. I don't have the second plugged in right now, never thought about it
Nice dual 8pin EPS plugs. If your powersupply has dual 8pin EPS connector for the motherboard then I would use it. More connectors = lower amperage = higher efficiency = less heat.
Yeah like I said may the best capitalist win. I hope you can improve the throughput that would only make your product superior. Still on an investment basis that isn't a risk I am willing to consider. I am interested in the design that has lowest cost per MH based on a $4000 or $10,000 investment. If that is your board then I am interested.
catfish: to compare to ztex, our dual FPGA boards are $560 each, or $280 per FPGA. His boards are $460 each with one FPGA. That's a big difference, eh?
It really just comes down to the design of the board that you prefer, though. I think all these boards look awesome, and it's great that people have some options, depending on their needs!
It really just comes down to the design of the board that you prefer, though. I think all these boards look awesome, and it's great that people have some options, depending on their needs!
Well no it all comes down to MH/$. While you may have 2 FPGA you are getting what 240MH/s combined.
Ztek gets 190MH on a single FPGA.
Also he offers significant bulk pricing. I wish he had a 4x FPGA design though. Building a 10GH cluster with 50 boards is kinda repetitive. Anyone looking to buy FPGA is looking to buy big. Whoever make the highest capacity board w/ lowest price per MH will be the one who gets the lions share of the business.
Good luck to everyone. May the best capitalist win.
Dexter you have forced me to upgrade my 10GH/s sawhorse juryRIG monstrosity.
I thought your rig was so cool that I showed it to .... wait for it .... my wife.
Her response was "Why don't you do that?". DOH!
One question. What are the specs on your AC unit? I assume it is a dedicated unit not used for the rest of the residence. I mean if you are pulling 12KW at the wall that is >40K BTU/hr which requires some "real" heat transfer.
I thought your rig was so cool that I showed it to .... wait for it .... my wife.
Her response was "Why don't you do that?". DOH!
One question. What are the specs on your AC unit? I assume it is a dedicated unit not used for the rest of the residence. I mean if you are pulling 12KW at the wall that is >40K BTU/hr which requires some "real" heat transfer.
Ah gotcha. Thanks for the suggestions. Again guess I'm just used to complete systems, I'm afraid.
The extra hardware (not counting the palty discount on RAM and not factoring in HDD) for all the monitors, CPUs, KB/mice, and CPU coolers only amounts to about $420 over the baseline as-low-as-you-can-go-on-Newegg prices - at this rate, the price of one 5970. Well, again, as low as *I* wanted to go on prices. Now, aside from the fact that at the moment, the only system that has a UD7 that could take a 3rd 5970 in it, I was debating on having to swap the LEPA G900 for one of the 1,200w units (looking at about 700w load for the system and 2 cards already so a third would put that wattage very close to, if not over, the limit).
Now CPU-wise ok I might be $120 over from getting cheapo single cores. When I bought an X3 and saw that the two cards pegged it to 100%, I wasn't sure if it was just a bug compared to the CPU being powerful enough to keep the cards fed. I think now it is rather the former and if a single core is really good enough to mine with then that's good to know. But why spend 66% of the cost of a quad CPU on something that would suck for pretty much every other task except mining? Can't BOINC on a single, nor render or anything for any decent efficiency. Maybe my thought is just flawed but then again I thought I was building pretty "cheap" compared to previous systems I've had. This thread still has shown me that no, you can go cheaper (much cheaper). Oh well, live and learn.
Again these were originally intended to run at home where I could barely keep temps at decent levels (IMO, mid-70s and below) with two cards, let alone 3. Now that they're in a cooler location I could probably spend a bit of time to swap things out if I had to. I have all 5970s working away at the moment though so until I can do more builds or get more cards, I might be losing 200MH due to temps overall at the moment (ideal VS current setup).
The extra hardware (not counting the palty discount on RAM and not factoring in HDD) for all the monitors, CPUs, KB/mice, and CPU coolers only amounts to about $420 over the baseline as-low-as-you-can-go-on-Newegg prices - at this rate, the price of one 5970. Well, again, as low as *I* wanted to go on prices. Now, aside from the fact that at the moment, the only system that has a UD7 that could take a 3rd 5970 in it, I was debating on having to swap the LEPA G900 for one of the 1,200w units (looking at about 700w load for the system and 2 cards already so a third would put that wattage very close to, if not over, the limit).
Now CPU-wise ok I might be $120 over from getting cheapo single cores. When I bought an X3 and saw that the two cards pegged it to 100%, I wasn't sure if it was just a bug compared to the CPU being powerful enough to keep the cards fed. I think now it is rather the former and if a single core is really good enough to mine with then that's good to know. But why spend 66% of the cost of a quad CPU on something that would suck for pretty much every other task except mining? Can't BOINC on a single, nor render or anything for any decent efficiency. Maybe my thought is just flawed but then again I thought I was building pretty "cheap" compared to previous systems I've had. This thread still has shown me that no, you can go cheaper (much cheaper)
. Oh well, live and learn.Again these were originally intended to run at home where I could barely keep temps at decent levels (IMO, mid-70s and below) with two cards, let alone 3. Now that they're in a cooler location I could probably spend a bit of time to swap things out if I had to. I have all 5970s working away at the moment though so until I can do more builds or get more cards, I might be losing 200MH due to temps overall at the moment (ideal VS current setup).
Well you learned stuff so that is always a good thing.
Buying a more powerful CPU "just in case" is not good economic sense. I can see the desire to have cased rigs and going w/ dual 5970s is a good choice in that respect.
BTW the goal isn't "cheap". My 2.2 GH/s rigs have a $260 PSU. Hardly "cheap". It is efficient.
As in what combination of equipment gives me the optimal MH/W, MH/$ and uptime.
Great, I love you to start out FPGA clusters, there is only one possible criticism, it is wasteful to use a USB cable when two cables are only useful around the connector, I mean the input and output data, it may reduce the cost with the use of two simple normal pins and wires between the PCB and the hub.
Off the shelf is usually cheaper than custom.
If you create a custom wiring interface then you need a custom built hub. It is unlikely to be cheaper than a usb hub and usb cables. A 12 port hub is <$2.00 per port and usb cables in bulk are less <$1 ea. So cost becomes <$3.00 per unit (likely <$2.00 if you look hard enough). Maybe it is wasteful/overkill but it is unlikely any custom solution no matter how spartan will be cheaper than $3.00 per unit. Even if it is you are splitting pennies when you look at total system cost.
Re: Enforcing 'n' signatories & signing order for m-of-n P2SH Vs. Multisig
The only significant limitation is that P2SH are limited to 520 bytes because pushes to the stack of more than 520 bytes are not valid.
The bad news is we likely will have that limitation for a while as changing it would require a hard fork.
Do the current consensus code accepts blocks with pushes over than 520 bytes?The bad news is we likely will have that limitation for a while as changing it would require a hard fork.
(I think yes, but not sure)
No the blocks are invalid because they contain invalid txns. This is not a "IsStandard() check it is an valid vs invalid check. A block containing a txn with a push larger than 520 bytes is no more valid than one containing a coinbase generation of 500,000 new Bitcoins.
Re: Theoretical minimum # of logic operations to perform double iterated SHA256?
As I understand it, (and I could be wrong here) what is actually absolutely required to spend energy for, is the output. IOW, you could at least in theory design a system that answers the one-bit question, "is there a nonce meeting the difficulty target within " by actually spending the energy to write exactly one bit. Everything else can be reversible, so the greater the amount of computation you can do without any external effects required the more of it can be done "free" (albeit at ridiculously high complexity) but no matter what, you have to write the output.
In theory yes but even proponents of reversible computing don't believe leakage will be that low. There is the energy cost required to perfectly isolate the circuit from the outside environment so even if your raw circuit was perfectly reversible the total system energy cost will be much higher.
Also theory is just theory. In theory it is possible for someone to make a miner with 5,000,000 G/J (instead of 1 G/J) using plain boring classical computing. Granted you aren't going to do it with 20nm silicon but in theory it can be done. Now 5,000,0000 PH/s well we know that is not possible without a massive reduction in the "work" needed to complete a single hash. "In theory" is a nice way of saying nobody has proved it impossible.
Re: Theoretical minimum # of logic operations to perform double iterated SHA256?
Indeed. It seems obvious that reversible computing is the future for Bitcoin mining ASICs, and could be the first major commercial application of the technology. I have been looking forward to seeing informed comment on it.
Maybe in 50-80 years. Nobody has successfully implemented a 32 bit adder using reversible computing. If you think quantum computing is in its infancy well reversible computing hasn't even been born yet. Despite the theory being published in 1973 to date there has been pretty much no practical demonstration of implementing the most trivial of reversible circuits.
Part of the problem is that the circuit must be very insulated from the outside environment in order to remain reversible. To date this means a lot of very expensive near zero superconductors but even esoteric problems like a stray cosmic ray striking your circuit can leak to large scale leakage. To say the future of mining obviously involves reversible computing is sort of like suggesting Honda should stop researching hybrids and start researching hyperdrives because obviously on a long enough timeline some form of faster than light travel is an obvious requirement for a spacefaring civilization.
There is a lot of potential improvement in classical computing. We something like a million times less efficient than the thermodynamic limit. Also as Peter points out there is still an energy time tradeoff. If I gave you today a reversible miner which ran on near zero energy but had a hardware cost $10,000 of per GH would you be interested? If the technology is more expensive on an amortized per hash total lifecycle cost than classical computing well it doesn't really matter how little energy it uses.
Now that most P2SH redeem scripts are standard you can create multisig scripts which uses addresses (pubkehash) instead of pubkeys. Client support for signing that script is another thing. So this is more a client implementation detail not a protocol limitation.
The problem is that simply having an IP address that is accessible or completing a low diff POW doesn't mean you are actually a node. Proving you are actually a node especially over an extended period of time is a non-trivial problem.