ive never built the bitaxe before so whats the power draw on there to get say around 700gh? I know on the s17+ you can get it to around 370gh per chip. yes bitdeer is the offshoot of bitmain. we dont do any manufacturing only cloud hosting, this is just a personal project. the heatsink issue is confounding me and I really am just tempted to use thermal pads but I know how hot these things run and just dont think that will cut it unless youre at like 100mhz, thermal paste would work but you run the risk of not putting enough on and it making contact same with thermal adhesive. The way the 17+ is laid out it has seperate heatsinks per asic but thats also in a machine with a 6000 rpm fan which im looking for this to sit on the corner in the floor and you basically forget about it, maybe get lucky with a block, or just hope pricing goes back up and by then even the little bit youve mined would be worth something. its really frustrating that you either can only do a hobby project with just a few chips or you have to go for it and basically just clone a hashboard. cause I can now see that my design wouldnt really work anyway because if its drawing that much power per two chips in series then me having 6 sets would make it impossible to feed, which refers back to at that point you basically just have to clone a board just with less chips.

Quote with image tags.

It would honestly be great if you could export those EasyEDA files to KiCAD and then push that KiCAD project to GitHub & continue developing it on KiCAD. I switched myself, too, following Skot's advice and it's not only been easy to switch, but this software is also sooo much better. World of difference.

Nowadays most power goes to 12V. For instance, this 1600W premium unit from Seasonic only delivers 82.5W of those 1600 on the 3.3V rail.
https://seasonic.com/prime-tx#specification

Lower specced PSU's still deliver 20A, but that's pretty wasteful. You're looking at using 66W out of a 650W unit.

You were faster!

I am right now running 6 BM1397 in the R909 that Skot linked to you earlier, producing 2.133Th/s at around 100W. So you need 2 chips at 350GH/s each and those will pull about 30W together. Giving them more juice just makes them run too hot and getting less efficient.

Thermal paste would still give you contact in some areas. It is just meant to fill tiny imperfections in the metal of heatsink & chip, not to create an electrically insulating layer between them.

Maybe look into graphite thermal pads. They have very high thermal conductivity (like paste) and I believe low electrical conductivity.

You could buy yourself an R909 and open it up (I won't do it with mine); I see 2 beefy heatsinks, one per side of the PCB. It has 6 chips. No idea how sidehack did it.

So very cool!!! Going to make a really neat case for it.

You should join us at Discord. Just a fun place to collaborate on tiny bitcoin mining gadgets and gizmos . Knowledge base and trying to get some tutorials on there as well. Collaborate, share knowledge and have some fun.

You should be able to scale it up without problem quality loss. I actually didn't even check what scale it was imported as and simply resized the heatsink to fit the PCB. Probably should have been the other way round, but you don't see that in a render..

All knowledge in the discord is open source. We should not have issues with scammers. Seeing everything is free. None of the mining projects associated with the discord or the open source Bitcoin hardware initiative sell anything. They give away all knowledge gained. All their time  and energy is spent from what I have gathered. For personal pleasure and to enrich the Bitcoin community with open source knowledge. Something that has been void in the Bitcoin mining hardware development community for far to long.  Most mining hardware companies, go to great lengths to keep their information closed source.

These developers evolved in these emerging open source Bitcoin mining hardware projects are truly spectacular. They are selfless and have the original  vision of Bitcoin “open source” close to the chest. Nothing to scam when everything is FREE. We will however watch closely.

What is really awesome is they are actually trying to make it even easier for people to get with PCB manufactures and order these prototype boards themselves “tutorials”. A hug daunting task for the average person is to actually order these prototype boards. A goal is to make tutorials for anyone with zero hardware knowledge to order these boards and slap on an ASIC chip. Done and Done.

Many people who have zero hardware development skill. Can actually gain first time hardware knowledge by the love of mining Bitcoin and  through the knowledge provided by these open source hardware projects. Really fun concept of people learning Bitcoin through building their own miner.

At the same time Mining hardware companies will still thrive. People will always want boxed, plug and play miners. For people wanting a more hands on deeper understanding of  Bitcoin and the hardware that mines it. They have options like Bitaxe and Voodoo. To name a few. Sorry for the long reply. Was able to get out all my thoughts out via your reply. Thank you.

I just checked on some used BM1397AG I have; there is about 30Ω between the bottom GND pad and the top. I also checked on my last remaining new BM1397AG and it's more like 90Ω.

It is my understanding that the S17 has individual heatsinks for each chip because of this problem.

I have heard from a guy on our discord (you should join!) that in some cases they sand off the copper top.. I didn't believe it, but he has some pictures.

I think they could also use a electrically insulating but thermally conductive layer too.

i think the "sanding" is to flatten the prior solder remains, when retrofitting the mono heatsinks, then termal paste is used for isolation

Unless they are on the same ground. Then it would work.

I just had a closer look at my R909 and noticed it actually has 4 individual heatsinks; 2 per side.
Here's a well-lit picture from a review; the standoff in the center makes it difficult to see, but you can tell that the spacing between regular finns and between the two most center ones is different. Also notice the gap between standoff and bottom right heatsink. That's not a shadow.


You can check out the beginning of that review for more angles (he moves it around a bit): https://www.youtube.com/watch?v=fgU1fUfEd3Y

I'm not sure if it makes sense to have chips on both sides of the PCB or if GekkoScience cools the chips from the die side and from the PCB side.
In any case, they do seem to group chips together; in case one side was just PCB cooling, then 3 chips wold be sharing one heatsink.

---

Edit: Mystery solved. Thermal pads seem to work fine thermally and insulate the chips electrically, at the same time.