So Phil - you're saying you want to mine a coin whose value is being held artificially high as a sales gimmick?

ASICBOOST is a hardware implementation. I wonder how they built it into the S9 in a way that would work with the typical. Maybe if I get time this weekend I'll go back over the whitepaper and refresh on some concepts. If Bitmain was doing something that wouldn't be surprising in the least; those guys really really like profits and don't seem to mind screwing people to get 'em. At least they always deliver the gear, got that going for 'em, but that's about it.

It should be drawing about 60W from the brick, maybe a shade more. Kill-a-watt reports about 70W into the brick. Mind you that doesn't factor in power for the controller, but it does have a small fan on it.

Also, thanks for the hashrate spam.

It should converge to 132GH. Got my Terminus prototype running the same speed right now.

Regarding cables:

Yes, I've always made cables. Been selling PSU breakout boards for a little over three years and they've always shipped with in-house-manufactured cables.

The cables I made for this build are 24" from EVGA 8-pin PCIe (and CPU) split to two PCIe 6+2. 18AWG wiring is necessary for doubling up in a pin (double 16AWG is too much wire for a solid reliable crimp) but that's still good for about 150W per end, meeting PCIe 8-pin spec.

Not sure if they're in use in this build but I also supplied some 6-pin to 4-Molex for riser power, plugging into the EVGA's SATA/RERP ports.

There's also the consideration that coin prices tend to drift toward the average cost to manufacture and there's no reason your favorite altcoin wouldn't do the same thing. And if you're trading your cheaply-mined alts for bitcoins you also have to consider those coins were generated by a miner somewhere getting the short end of the stick.

But anyways, a question about bitcoin mining hardware and protocols can't really be answered by saying "ignore it and mine something else instead".

Last I checked, it was a means of paralleling multiple midstates under a single nonce. The most efficient means of generating multiple midstates relied on a change in bitcoin protocol, which is using a zero-padding field as an extra data field. If done right it ends up combining another step (or partial step) of the double-SHA calculation where the efficiency increases slightly the more parallel midstates are used per nonce. I forget the specifics because it's been six or eight months since I read the whitepaper.

I know nothing about the extra nonce operation you're talking about. I wasn't really concerned with chip functions back then.

What's the current Bitmain/ASICBoost fiasco? I seem to have missed that news.

Wonder if they'd do a group buy or a bulk buy to a reseller.

Also I have a half-dozen or so S5 controllers if anyone does get their hands on one of these and wants a brain for it.

I test for stable operation at 100MHz, no higher than 1.26V (most are 1.24V). That core should allow up to 150MHz without issues, but 100MHz is about 4W so I recommend cooling but past it you definitely need cooling. As with the Compacs, I don't guarantee anything's overclockability and I also don't guarantee it won't burst into flames if you don't know what you're doing.

The Compacs were tested for stable operation at 200MHz no higher than 660mV.

For any questions about if this stick will mine anything besides bitcoins and other SHA256-algorithm coins, please do a bit of research because the answer is definitely NO but understanding why might lead you to answer a lot of other questions for yourself that people are getting tired of hearing around here.

Also, I honestly don't really care much about BFG so I haven't talked to the developer at all about supporting 2Pacs. If VH wants to send him what he's got for driver code maybe we can get it in there.

The black epoxy won't come off under rubbing alcohol or acetone. The white epoxy is acetone-soluble. I don't know of any epoxy that'll dissolve in rubbing alcohol, nor would I respect it if it did. But that's not really the point, or at least not the entire point.

As has been mentioned a lot of times before, having to heat 50-60 individual heatsinks and pop them off and then pop all the backside heatsinks off and then soak the board in an industrial solvent and scrub off epoxy just to get the boards to a state S5 boards are already, and then you have to strip the chips and then reliably install them on something being careful around the 0.4mm-pitch pads and traces, really REALLY isn't worth the trouble. Don't even argue. It's not going to happen. I already put up more effort than most sane people would to get BM1384 products out there using stripped chips.

So anything newer than BM1384, I'm working on BF16 dev for that because I can actually get new chips instead of wasting buttloads of time jacking around with someone else's crap trying to harvest chips out of a hairball of metal and glue.

Were you cleaning an actual thermal adhesive, or just regular heatsink grease?

Phil did mention sometime a month or so ago seeing if I could build some of those bridges. Probably wouldn't take much, really, if I had one to pattern. I reckon we should see if they work on this new hub first.

Um, that's great. Just so you know, you already sent me 4 messages about it today and also you're not my only customer.

Also, a note to seekers of warranty work - I test and guarantee the sticks at stock specs. A stick that doesn't overclock as well is not grounds for warranty work. For a stick that has obviously been overclocked with inadequate cooling (PCB or chip discoloration is apparent) or abused (physical damage, parts missing and/or badly attempted home repairs) won't be covered under warranty. If you want me to fix it I will require compensation.

A stick that hasn't obviously been mistreated but still fails to perform at stock will probably still be covered.

No offense but I'm not clicking that link. Google scripts aren't allowed on my machines.

From BFL's playbook though, it wouldn't be too hard to make a tube with a fan on one end, or Jalapeno-like box with fan on top.

I wonder about just a folded steel tube with a C cross-section, with a fan on one end, wrapping around the miner with the jacks exposed. Would need a few tabs for screwing into the board, some for the fan. Not too difficult for someone with the right equipment.

Right, except those guys are two of my biggest customers, between the two of them responsible for over half the entire first batch of 2Pacs (and a majority of Compacs ever built). So anyone care to help out with this, instead of criticize needlessly?

Both MacEntyre (bitshopper.de) and CrazyGuy (ASICPuppy.net) desire that these Terminus pods had a case or housing.

I'm figuring on adding screw holes to the PCB to help facilitate this, but I have not the resources to manufacture a case worth selling.

Anyone care to help out with this?

Phil's I replaced the ASIC. Yours I saw no issues with, ran it for a week with no errors.

Done. Come and get 'em.

Even without that, 40A is borderline for half an Avalon. Probably not failing in 30 hours kind of borderline, but certainly pushing it for long-term.

It depends on if your hub has that 0.5A/0.9A limit built in, or if it's more of a limit in the regulators feeding the ports. There's also the issue of wimpy wires in the Y cable melting under loads.

I should have revised prototype PCBs in about ten days. Mind you that's the same time 2Pac batch 2 starts production so I might not get to them right away, but I'll do what I can. Hopefully this time they work as expected. That buck circuit is very disappointing, especially since it was going to be used on the Amita and TypeZero stuff as well. If the one as designed on the Terminus is up to snuff I'll probably carry that design forward to bigger models instead. It's about the same cost but more complex, however it should also handle the heat from pushing high current a bit better which is a plus for larger hungrier boards.