|
If you are serious about this why would you build it on the East coast? Build it in Washington state or Oregon where you have access to cheap hydro power. Just because you are located in Maine doesn't mean Maine is the ideal location for your startup.
Thinking about a 1M BTC funded startup is a fun thought exercise but I don't think the market could fund such a thing at this point. I hate to tell you to aim lower but, without sourcing traditional VC capital AND having a solid business plan, $5M USD is a SERIOUS sum of cash. You might have better luck in the $100k USD to $500k USD, prove the concept works and go for a second round of funding. This isn't like you are pitching a web startup in a VC environment with a proven track record of successful exits. Here, any VCs interested in funding a mining operation has a year or two of experience and no previous examples to judge future performance on..
|
|
|
|
I wasn't suggesting Merrick1 for Bitcoin for those reasons. The cut down Merrick6 is a better board initially to talk about. Doing a derivate design for Bitcoin it could be possible to up the power supplies to support all devices by simply say removing DDR chips to make space. The 12A regulator circuit we use could fit into the space taken by 2 DDR chips and we could end up with 6/7 FPGAs each with 12A of core voltage.
That's all viable to do and that sort of design change could be done in a few days maybe a week. A prototype could be turned in about another week if we pushed although that fast gets expensive.
That would be an interesting option. If possible, a 4-pin Molex or SATA power connector might be more useful. I know a lot of my PSUs have 1 floppy power connector but a dozen SATA and 4-pin connectors. |
|
|
|
We are based in the UK but we already ship more or less worldwide with a few exceptions. Principally the exceptions are places where we are not allowed to sell due to export restrictions. For most of Europe and N America delivery is usually 24-48 hrs for anything in stock. Other places can take a bit longer.
In terms of deliveries we running this as a professional business and it's rare for us not to meet our delivery promises. Usually if we are out it's because one of our suppliers let us down and beyond our control. We have been in business for 23 years and building FPGA boards for the last 9 years so we have this pretty hooked. We are building boards every week and this is a normal task for us so I hope we know what we doing by now.
Do have a look at our 101 FPGA board Merrick1. Current FPGAs might be a little small for Bitcoiners but they could be Spartan-6 LX150 if we did the design again. Current Merrick3 is our biggest S6 board launched with a total of 26 S6 on board. It's only got 72 amps for those and would need a haircut, or more power supply, to make a better balance for Bitcoin.
I saw the Merrick1. At that price point you are competing with the BFL Mini-rig and LargeCoin C200. Those are custom turnkey solutions with the space to dissipate the heat. A Merrick1 with 100 Spartan6 LX150s would dissipate around 1 kW and generate around 20 GH/s @ maybe $25k USD given the performance/price of other Spartan6 LX150 based devices. Impressive, but I think you'd sell larger overall volume with a slightly smaller device. Perhaps create a poll to see what gaps there are in the market? We've got a lot of 1-2 FPGA devices. ZTEX just launched a 4 FPGA device. Lancelot is expected to be a 4 FPGA device. The next step is the BFL Mini-rig and that's a massive jump disclaimer: I'm eagerly waiting for someone to release something in the 1-2 GH/s range with competitive MH/$ and MH/w ratios. So, I'm biased when I talk about what the market needs  |
|
|
|
Can they handle more than 5-6 GPU's? I heard that a Windows Cat/Forceware Driver cant handle more than 5 ore 6 GPU's, not sure.
I wonder if I should consolidate all the information into a new topic, since no one ever reads the posts in this one. Threads of this length become hard to parse and maintain interest through the entire thread. Especially with the last few pages consisting of the same "how will you run that many GPUs" without any progress on that front. Might be worth closing this thread, linking at the last post to a new thread with updates. I guess I'm probably taking it a bit more personally than I should, since this is my baby. But perhaps I will start a new thread when I get the new host board. Understandable. Keep in mind though that it is a serious time investment to read a 20 page thread if you haven't been following this from the start. Keep at it, we want to see this thing completed. |
|
|
|
On the power Merrick6 uses a disk drive connector for higher power applications so it won't burn out the PCIe connector.
At the moment the full product is a build to order and the reason for that is cost of silicon sitting on the shelf. That's already being considered to become a stocking board and a cut down would be easier to justify stocking especially if we get steady orders. The 4-6 weeks is a standard time we quote and usually if we have parts, which is usual on Spartan-6 based boards, it can be a lot less. It's more a case of not disrupting our assembly line schedule for say 1 board. If we have a 6 week window we can make a nice slot and maybe build a proper batch size at one time. These are all to do with the economics of building boards.
Merrick6 can run stand alone if needed and there are some communication options here if you are not using the PCIe.
If there is a serious interest we could do designs based on our Raggedstone2 and Lamachan2 boards to offer a PCIe board with 4 or 8 XC6SLX150 FPGAs that would come in cheaper per FPGA. That would need some numbers to make this a worthwhile project.
I'm just a guy interested in FPGAs, I don't have the knowledge to understand what chips would be ideal and what not. What I do know is that there are a number of FPGA mining device manufacturers. While they might not be interested in sharing their sales numbers but a few hours of searching on just this forum could give you an idea of a fraction of sales they are achieving. I'm fairly certain the majority of miners have seen the light and realize that FPGAs will eventually price GPUs out of the mining industry. Someone who could deliver a hashing dense FPGA device with an appropriate turn around time could swoop in and snap up orders from the likes of BFL and LargeCoin. I'm not sure what the setup/design costs would be for you guys to design a PCIe card pushing 1-2 GH/s but I'd be thrilled to see another competitor in the manufacturing market. |
|
|
|
I think u should build a board with at least 4 FPGAs. Two Spartans seem to be mainstream now  Look at Ztex shiny new 1.15y Board (850MH/s @ ca. 1188€ with tax). And please try not to do something like BFL to us. 4-6 Weeks BTO is ok if u can hold that promise  BTW where its the company located ? The UK -> http://enterpoint.co.uk/contact-us/ |
|
|
|
Can they handle more than 5-6 GPU's? I heard that a Windows Cat/Forceware Driver cant handle more than 5 ore 6 GPU's, not sure.
I wonder if I should consolidate all the information into a new topic, since no one ever reads the posts in this one. Threads of this length become hard to parse and maintain interest through the entire thread. Especially with the last few pages consisting of the same "how will you run that many GPUs" without any progress on that front. Might be worth closing this thread, linking at the last post to a new thread with updates. |
|
|
|
If I understand correctly, the cut down Merrick6 would have 2 XC6SLX150 FPGAs for mining purposes and a XC6SLX150T for PCIe connectivity and some mining for $1000. If that is the case, that is damn expensive per MH/s but perhaps some people would pay for the form factor. For the industrial motherboards with no fan directly above the board it would be possible to use every slot. That does need very good case fans then. Our solution in development will also allow every slot to be used.
Sweet. In my application, I need power density. Have you considered a design that will allow you to cram a larger number on a board? The reason I ask is because most boards with 2 Spartan 6 FPGAs are no more than $500, and having more of them on a board could lower the per-chip cost. A six FPGA unit a 1200 US$/950 EUR would be really sweet, you need to have a motherboard to use them, but they would replace GPUs in current rigs very nicely. spiccioli. I'd be all over a $1.2k USD 6 FPGA board that fit in a 2 slot PCIe form-factor. For such a card I'd prefer the ability to feed it a PCIe cable to prevent from burning out a mobo. I'd basically swap out my GPUs for FPGAs. I doubt however, unless they want to re-engineer their product, that we will see such a thing. |
|
|
|
|
I wouldn't even bother with cleaning it up. If a component went up in smoke, RMA it as is. Hopefully you didn't void the warranty by cracking it open.
I hope you were running that thing at 70+C on the cores. Chances are the VRMs were roasting and one of the fets blew if this is the case.
|
|
|
|
Hmm some of the shipment times are going to be off. It might appear that they are suddenly shipping faster due to the fact people report the day they ship out vs the day received. I happen to live in a state adjacent to BFL so perhaps that's why I go them so quickly. I wonder if I can walk in and buy some off the shelf? That would be awesome...but I know they wouldn't do that.
If that was possible I'd drive down over a weekend to pick some up. |
|
|
|
So $100 per PCIe slot.
$50 Sempron 145
$10 2gb ram
$174 MSI 890FXA-GD70(I bought a number of them @ $125 open box)
$200-$250 your flavor of 1200w PSU
$25 5 PCIe extender cables
$5 4gb USB stick + BAMT/Ubuntu
Comes out to ~80/PCIe slot but you have 1 machine to manage instead of 6. The resale of hardware will also be higher.
With only 7 computers I wish I could combine them all in to one machine to manage. I couldn't imagine doing this with ~35 machines.
I was thinking 4 cards per rig until I thought... what happens when 1 board or PSU goes up in smoke? 4 cards are down. If the op was thinking of these... http://www.newegg.com/Product/Product.aspx?Item=N82E16813157209. However, as others have said, I wouldn't necessarily go for the cheapest PSU. Cut the 16gb down to 4, get a used 24 10/100 switch for $10 or $20 bucks. The other thing is your power inefficiency goes down when you have 1 underloaded PSU per board. Let me know how your nightmare goes with dozens of computers running to get any appreciable amount of hashing power |
|
|
|
Ok so I signed up to GPUMax... is there like a waiting list to get in or something?
Yup. Signed up a month ago, still waiting. Get in line, punk |
|
|
|
The compressor can run with 100% duty cycles. Refrigerators are typically not equipped with adequate condensers to reject the heat that is removed from the cold chamber and from the compressor inefficiencies. Swap an appropriate condenser and you'll be good to go.
My guess was indeed that the compressor was overheating. But I'm not sure that the compressor itself is meant to be cooled down by the condenser: the condenser is after the compressor in the cycle, so the compressor's input (warm vapor) should always have roughly the same temperature range (unless the whole fridge is overheating and mine wasn't). My guess is that (at least in my case) compressors have sufficient thermal inertia to not overheat during their common working periods (I'd guess 5 minutes at most) and then they are cooled by natural convection during their sleeping periods. I guess that if they have to work continuously, the natural convection isn't sufficient and they begin to overheat (lubricant is slowly cooked, the motor mechanism eventually becomes stuck). You are partially correct. On mini-fridges the condenser coils are embedded in the walls of the unit close to the exterior. The mass of the walls are used to dump heat and that is radiated away over time. Some full-sized refrigerators use this same principle and others use a dedicated condenser with a fan. These are still sized for small loads as fridges normally don't have things inside of them that generate heat(except when your cat hops in the fridge). I had a twin rotary cascade that ran for 8-12 hours at a time cooling a Xeon w3570 down to -110C. I built it for continuous runtime and it was able to reject all its heat through the 1st stage condenser and 2nd stage desuperheater. If you build them correctly, refrigeration systems will run until their compressors wear out(which can take a long time). |
|
|
|
|
So $100 per PCIe slot.
$50 Sempron 145
$10 2gb ram
$174 MSI 890FXA-GD70(I bought a number of them @ $125 open box)
$200-$250 your flavor of 1200w PSU
$25 5 PCIe extender cables
$5 4gb USB stick + BAMT/Ubuntu
Comes out to ~80/PCIe slot but you have 1 machine to manage instead of 6. The resale of hardware will also be higher.
With only 7 computers I wish I could combine them all in to one machine to manage. I couldn't imagine doing this with ~35 machines.
|
|
|
|
yes i know this but the system needs it to run stable,or am i wrong ?
Needs what to run stable? Clean power, adequate ventilation, appropriate GPU core voltages and frequencies and you are ready to mine. |
|
|
|
Condensation may not be a problem. I wouldn't be sure until a test is done (20 minutes isn't enough) because air is warmed locally by the board and could eventually reach a cold surface just above the board were condensation would slowly build up if it hasn't time to cool down before. Seems unlikely to happen but I'll check such a setup regularly and wouldn't change what I place above the boards in the fridge without additional checks...
But that's the least of your problems. The compressor of a refrigerator isn't designed to work continuously and will break if it is forced to do so. I learned this painfully when the switch which should have cut the light in the refrigerator stopped working. After several weeks/months (I'm not sure when the switch failed exactly), the refrigerator was dead.
The light bulb in the refrigerator was only ~20W...
The compressor can run with 100% duty cycles. Refrigerators are typically not equipped with adequate condensers to reject the heat that is removed from the cold chamber and from the compressor inefficiencies. Swap an appropriate condenser and you'll be good to go. |
|
|
|
so you don't have to connect your cards with crossfire bridges ? what the name of this monster? has a sempron cpu enough power to run 6 hd 5870 cards ? CPU power is not needed for mining. You only need PCIe connectivity for a card to mine. A PCIe x1 slot will suffice. |
|
|
|
|
CFX only scales to 4 cores. You do not need to run CrossFire to mine on more than 4 cores. I have two rigs with 8 cores each for reference.
|
|
|
|
|
eBay for PCIe cables. I just bought a box full at $4/cable and I got them 3 days after I ordered.
|
|
|
|
|
Very interesting. This might line up with another plan of mine.
|
|
|
|
|
Show Posts
