15 pounds of aluminum.
Is that a 1.3 amp fan? I want to buy one, but that would become the loudest thing in the house.
It is impressive to see how so many PCBs can work together flawlessly like that, but seems way over-engineered. I think more material cost is going to the aluminum blob in the middle than the actual mining hardware.
Rockxie, you were on the right track when you designed the "ROCKMINER_BLADE_V1.2" component that makes up the RK-Box. I hope you can put together a simpler unit, perhaps that exact part (the blade) without a Raspberry Pi, USB Hub, giant center heatsink, fan, cables, etc. Think ultra-simple.
This simple:
Just a PCB with 32 chips on it with an ethernet jack. If you used that exact part but replaced the chips with gen3 ones, it would be a very powerful unit and leave the cooling fan setup to the end user. It's nice to have a finished product that they can simply plug in, but after watching the video I don't think this is as simple to use as it could be.
An alternative to a redesign, I propose to offer the sale of the 1.2 blade (seems like it would be 112-120 GH/s per blade) seperately, with just one heatsink as a seperate product to order. It would cost nothing to offer the sale of the individual component, it would be very cheap to ship, and down the road if someone's RocketBox was failing they could order the replacement part.
There is also a gap in the product lineup, it goes from 37 to 450 GH/s. You need a ~100 GH/s device to fill that void, and I think flooding the market with those blades is perfect. The blade is already designed and everything.
A 'new' blade using BE200 chips would be really nice to see and have. 100 GH/s would be equal to 37 Gh/s in a few weeks anyway due to difficulty increases.
Difficulty increases are going to drive us all insane anyways. I'm trying to think about the cost of making one of those gen1 (blue) blades... the heatsink, PCB, ethernet connector, green power connector, the 32 chips, and the rest of the circuits and components that are on there that make up the power module and ethernet controller. These all have costs associated with them, but the RK-Box comes with a Raspberry Pi. That's... probably close to the cost of the entire 1st gen blade. Plus you need two RK-Box's to get a Pi. It's such a confusing invention, 10 USB cables for two RK-Boxes and 2 USB hubs, gobs and gobs of aluminum, it seems more like a prototype than anything.
I just had to come put in my 2 cents, the first and second blades were fantastic and very successful. The USB sticks were also extremely successful, being super simple in design, they are going to be christmas tree ornaments and keychains for decades to come, I even still have half a dozen running at home.
The BE Cube is where it all started to fall apart. Having 5 blades on a backplane in a little box, although the concept was superb and it looks fantastic (mine is running for ages now without interruption), the tiniest flaws caused a lot of pain. I had to take apart hundreds of BE Cubes one at a time, tighten every screw, and re-align the blades due to movement during shipping. The odds of getting a perfect Cube were not in your favor, often you had to determine which of the five mini-Blades were unable to run on high clock, and I effectively had to tune each blade and grade them manually, taking hours.
When I have an RK-Box in hand, my opinion may change, and there's not much I can do from a user's standpoint but deal with what is developed. However it is the simpler designs which are the most successful and I hope we can encourage Rockxie to go further down that path.
Basically like the DataTank immersion blade, with a heatsink and a price tag beside it.
It would be nice to see 1TH on a blade that allowed you to scale to 10 blades.
I think 1 TH is too much. That's like 80-90 chips? I think 32 would be the perfect amount. 280-350 GH/s per blade, maybe 3 TH/s on a backplane... Even make it compatible with the existing AM backplanes we have lying around for a quick drop-in replacement?