Oh I just set up a thread on litecointalk.io where I can fix these things.

https://litecointalk.io/t/fixing-and-hacking-knc-titans-because-they-break-from-time-to-time/1708/2

Posting here so people there can come here, see this, and know that thread is mine.

C

Well, what happened? Last I heard Antland brought up 30% more new systems that needed formware updates in their data hall which had the effect of killing segwit. Did they change their minds, was it just a short trap all along, or something else?

Curious.

Who would ever want to trade Bitcoins for dollars?

Actually I am, and I've always been interested in seeing what makes bitmain tick. I'll send over the address in a PM.

Thanks!
C

Oh lord, let's try to stay on topic.

Two problems:
1) That solution will not have enough surface area to keep the main chip cool. The power supplies are the bigger problem, but the best solution is to run them a bit more slowly (60-70mh) and not at a screaming 80. Also someone else made a new bracket for the factory heat sink that allows you to stick heat sinks on the supplies.

2) Ability to dump more heat does not fix the problem that the power plug on the Titans is already over limits. Pull more power and you burn the plug which shorts and destroys the cube in a lot of cases.

I ran a Neptune with liquid cooling, worked great until the pump stopped one night. Small explosion, big mess.

The way to win is to donate it to me so I can figure out what's going wrong :-) Then I can tell everyone else how to fix these things.

Oddly enough I have no problem with using ASICs as a shortcut on the algo (Asic Boost). Every person on the Earth is looking at SHA256 to see if there are places for efficiency and if found they exploit. Technically KNC and others exploited their own type of ASIC boost in order to make Scrypt run on dies (lots of cores with little memory each working on a smaller segment of the problem instead of fewer cores with a monster memory map as is GPUs). So you know, there's that.

The bigger item to note (not a problem, a feature) is there is no regulation in the bitcoin marketplace, thus this type of manipulation is part of the point of the damn thing. We didn't want any of that annoying govt stuff, we're getting some dictators.

What interests me in this particular grab is that litecoin per se isn't the point, the point is to squash development of segwit on litecoin (which if successful would pull stuff from bitcoin, put pressure on bitcoin to adopt it, gather all sorts of happy press, etc). This can be done by adding 25% hashing power to the litecoin network (to forever block it) which would at 90k difficulty levels take only a thousand or three L3+'s.

At roughly $300 each to manufacture in China land and no support costs (they own em, who cares if they blow up) that's a good solid 300k to sink segwit, protect your bitcoin market, and roll along. Pennies if you think about it.

Now if you had other ASIC manufacturers in business you might have a counterweight on this, but BFL, Avalon, AMT, KNC, GAW, CoinTerra, Spoondles, and others are all gone. Done in by bad feedback, crap products, skyrocketing difficulty and the like leaving AM standing.

I often wonder why AM even bothers selling miners at all, just keep them and rule the market. Any competitor shows up either sue them into oblivion, or just sell your castoffs for $10 each.

Hm.

You forgot short the coins in the meantime just because, but that sounds quite possible. I think it's highly unfortunate that there are no other miner manufacturers left (well, Inno), they seem to have been run out of the market.

Hm. No, I'm going back to bed.

Not to drop a rock on anyone's parade, but I just took a look at Segwit's support on the LTC network: It has dropped from ~80% a few days ago (est) to 63.7% support in the last 24 hours. No, 63.19%.

http://litecoinblockhalf.com/segwit.php

Segwit miner support within the last 24 hours (last 576 blocks)   364 (63.19%)
Segwit miner support (percentage of segwit blocks signaling within the current activation period)   1750 (76.29%)

What's happening? Not sure, but apparently a big boost in hashing power is coming online from somewhere.



Hm.

(Edited since I don't know where things are mining to, this is all pretty new on the LTC network)

Blah. I just had to send in my taxes for all my bitcoin work. That's a *lot* of recordkeeping to do to keep track of the work, payments, cost of parts, and time. And now I have to deal with the fact that most of those coins have doubled or quadrupled in value which totally screws up my basis points for when I sell them for LTCG.

Oi.

Sweet. Is that 4 troy oz, and how many btc does it go for?

P2pool has been purring along well as of late. One of the guys there has fixed some of the minor nits (mining less than full blocks), they seem to be good eggs.

Not sure how someone could help WK, any thoughts on what someone could do (I'm of the old Yourdin school of PM that says adding programmers to a late project makes it later).

Let's see. If nothing is connected to the controller, the light on the side (white, super bright) comes on, then off, then on then off, etc and does not go green then you have a bad FPGA on the board. Remove it with your usual SMD rework tools (flux, preheat, heat, pull straight up with a good picker once it just moves a bit when touched with a tweezers), remove all of the solder from the pads without hitting any of the capacitors, cool, clean off the flux with 95% isopropyl alcohol.

Then test the board. Fire it up with a bbb running the neptune code (a titan will try to reboot), log in and type io-pwr init as root. See if it comes back with OK (the TPS chip is good) or with a bunch of errors (you blew the TPS chip, replace that too, pain in the ass). Need to do this because there's nothing as fun as swapping the FPGA to find the TPS is bad as well.

If TPS is bad, replace it, then test board again to make sure it comes up ok. If not, keep working on getting the tps working. Soldering QFN is a pain IMO.

Then replace the FPGA making absolutely sure to line the balls up perfectly with the pads. This is why you removed all the solder, needs to be perfectly flat.

I normally do this for about .2btc+shipping (was .3 but bitcoin prices are up as are litecoin so .2 is fair). You can blow the FPGA by unplugging the 10 pin while powered, plugging it in wrong so only 5 pins are on the titan, it's a fragile interface because they don't have optoisolators or buffer drivers. Oh well.

Then follow my post on testing titans one at a time. Look for shorts on pins 4,6 to ground especially, those are the worst.

Or just put the Titan Pi and BB on the Jupiter board and fire it up. Success is white light on then off, then green with 2.00 or better Titan code.

C

Thank you LongandShort for the litecoin donation, glad the BeagleBone code is working for you!

As they say, you don't pay a hooker to come over, or to stay, but to leave. So yes, 2.5m transaction confirmations would be golden.

True, however most bitcoin ATMs for example can be easily programmed for Litecoin. One big problem with BTC is if you're trying to pay a hooker in Vegas do you really want to wait 10 minutes for your transaction to maybe confirm? You want a nice quick settlement so you can go out and get a burger at the Luxor.

It should be in the thread a few posts back. Everything that I figure out I have written in the thread, so the part numbers (Digikey) should be there as well.

Took me forever to find/figure out though.

buffer chips on the upper left side have probably blown out from plugging it into the 10 pin improperly. Signals can get to the power supplies but not the clocks inside the hashing chips (clock is provided via SPI signals.)

C

So if you put cube 1 and 2 on a controller, fire it up, the light comes on then off then green on the controller and the display works, right? If so they can probably be fixed.

C

No problem. Feel free to donate to the repair miners cause :-) The goal was to help the small miners keep in the game.