Topic: GekkoScience is now dabbling with 16nm ASICs for new designs - page 20. (Read 77028 times)

I've got a photo of the 2Pac test setup posted. The 2-chip BF16 test board is still a work in progress, nothing worth photographing yet.

Chips tomorrow - is that the Merry Christmas sample chips you're sending so I can prototype pods? Sweet.

I'm really looking forward to having more news for you all.

@ Punin

get us plenty of chips and I will definitely be swapping out ALL of my Bitmain gear for sidehack builds. fed up with losing boards on new Bitmain gear

Fantastic news. Thanks punin. Nice to see someone sticking to their word 

Phew.. took some time to read this thread through Haven't had much time to visit forums lately as things are moving at pretty crazy speeds here at Bitfury, but rest assured I'll do all I can to support community miners such as the pod and the compac also in the future.

P.S. Sidehack show photos of the boards already!!
P.P.S. Chips are going to get shipped tomorrow

Could you kindly provide a link to your tweet? I haven't seen any tweets from you mentioning or directed at me. Also, I don't even know how to block people on twitter as I don't use it that much.

if you already have the pod pcbs why dont you just make a limited run and use them? then make both 1384 and 1385?

Yeah, I'm kicking myself for deciding to pod with BM1385. I made that choice something like 3 weeks ago. If I'd stuck with 1384 I could have that pod PCB on the way alongside the 2Pac and be testing everything all at once. And not have wasted money on pod PCBs for an unused design. Oh well, get to it when I get to it.

makes sense to utilize just two types of chips (old and new), this way it is more streamlined, I agree.

And this is why I don't take in money for products that haven't been successfully prototyped yet, because things can change.

If the roll back of the BM1385 to BM1384 give more time for BF16 Im all for it 

So, Punin sent me a "Merry Christmas" email today offering to send me more samples, enough to build a few more sticks and about half a dozen pods. What a nice guy.

I also am having second thoughts about the BM1385 pod.

The chip specs of the BM1385 are pretty good. Top-clock efficiency is on par with BM1384's best-case efficiency. Estimated specs for the 10x1385 pod are 250GH at 60W DC. That's not bad; a 20x1384 pod would see about 190GH from the same power draw. That 20x1384 pod would have no legroom for undervolting, and you'd only get about 250GH if you pushed it to 100W (versus about 350GH from a 10x1385).

Pros for the 1385 are obvious - better efficiency, higher hashrate.
Cons abound - 0.4mm pin pitch sucks to work with. S7 chip harvesting requires hot air and an acetone bath just to expose the chips. It'll also mean more software dev time for VH to figure the chips out.

Cons for 1384 basically invert pros of 1385 - worse efficiency, lower hashrate.
Pros, however, are numerous - chips will be cheaper to source in bulk, and are easier to harvest. I'll already be doing that for the 2Pac anyway, and have streamlined the process after 2000 Compacs. VH's multi-chip 1384 code is only a few steps from being generalized to n>=1 chips and he's planning on doing that anyway, so by the time 2Pac drivers are rolled out they'd probably already work with the pod natively. I can build off the existing 1385 layout, changing on the chip field and support components, meaning design time will take about 1 day total. That leaves a lot of time for BF16 work.

So, who would be dreadfully upset if I rolled the 1385 pod back to 1384? Since most folks didn't even know I was working on a 1385 pod until a few days ago I'm guessing it's probably okay.

Put me down for a few each.

Rocket Scientist or Rocket Engineer?

 Having known a Rocket Engineer (Doug worked for Rotary for a while, and I think he also had some input into the engines for Starship One) I have a pretty good clue about the difference.

Nice that as I recall you have a technical college in yer area. Ah student labor

I would assemble them for you but I'm waaay over here.

I used to expect 300 cables per week from a cable minion, but with the tools I have now I can do that in one day. I might need a cable minion if I'm already stretched thin, but for that particular need I could probably fill quota just in spare time.

I'll definitely be finding at least one minion to help assemble sticks, and eventually pods. That's for darn sure still a tedious and time-consuming job.

That's exactly what made me think of it. Being a stickler for tidy cables a small link pcie would be a superb addition to the pods. I know it's more work for you but could possibly be outsourced to someone who wants to earn a few bucks.

Seeing as I make cables already, it wouldn't be difficult to keep a stock of something like 8" or whatever would work best.

With daisychaining being built in to the pods would there be a specific daisychain pcie cable available (for an additional cost) so if people were to stack/tower they the cables are tidy and not like a lathargic octopus?

Just a thought for an additional revenue stream where the materials would be relatively cheap and production straightforward.

Constant voltage is fine, just make sure that you never buy constant current stuff to fry ur miners.