sipa changed the topic of #bitcoin-wizards to: This channel is for discussing theoretical ideas with regard to cryptocurrencies, not about short-term Bitcoin development | http://bitcoin.ninja/ | This channel is logged. | For logs and more information, visit http://bitcoin.ninja
blackwraith has joined #bitcoin-wizards
priidu has quit [Ping timeout: 250 seconds]
dEBRUYNE has joined #bitcoin-wizards
King_Rex has joined #bitcoin-wizards
King_Rex has quit [Remote host closed the connection]
King_Rex has joined #bitcoin-wizards
King_Rex has quit [Remote host closed the connection]
bit2017 has quit [Ping timeout: 258 seconds]
King_Rex has joined #bitcoin-wizards
King_Rex has quit [Remote host closed the connection]
<bsm117532>
Very interesting kanzure. If IC manufacturing could be fully commoditized, this would enable distributed manufacturing of ASIC mining equipment, so that manufacturing didn't represent a 51% attack risk.
AaronvanW has joined #bitcoin-wizards
<kanzure>
azonenberg was working on homebrew transistor fabrication for a long time. the tradeoff is that you have to do low-resolution stuff only, which impacts performance.. and also handle chemicals. *shrug* what's the worst that coud happen.
supasonic has joined #bitcoin-wizards
MoALTz has quit [Quit: Leaving]
<bsm117532>
I'm afraid homebrew will never compete with 14nm billion dollar fabs. Therefore it's best to commoditize the fabs.
AaronvanW has quit [Ping timeout: 276 seconds]
roman__ has joined #bitcoin-wizards
roman__ has quit [Read error: Connection reset by peer]
Noldorin has joined #bitcoin-wizards
laurentmt has joined #bitcoin-wizards
laurentmt has quit [Client Quit]
<vyvojar>
kanzure: is tl;dr: tamper detection through obfuscation?
Noldorin has quit [Ping timeout: 252 seconds]
bustd_soket has quit [Ping timeout: 258 seconds]
PaulCapestany has quit [Quit: .]
PaulCapestany has joined #bitcoin-wizards
Newyorkadam has joined #bitcoin-wizards
bustd_soket has joined #bitcoin-wizards
AaronvanW has joined #bitcoin-wizards
ThomasV has joined #bitcoin-wizards
Noldorin has joined #bitcoin-wizards
AaronvanW has quit [Ping timeout: 276 seconds]
ThomasV has quit [Ping timeout: 240 seconds]
dnaleor has quit [Quit: Leaving]
ThomasV has joined #bitcoin-wizards
Noldorin has quit [Ping timeout: 276 seconds]
supasonic has quit [Ping timeout: 276 seconds]
Dazzix has joined #bitcoin-wizards
dnaleor has joined #bitcoin-wizards
Guyver2 has quit [Quit: :)]
ThomasV has quit [Ping timeout: 260 seconds]
N0S4A2 has quit [Quit: WeeChat 1.5]
N0S4A2 has joined #bitcoin-wizards
unlord has quit [Ping timeout: 264 seconds]
unlord has joined #bitcoin-wizards
<jtimon>
bsm117532: how can you predict what can compete with what? are you claiming that you can predict demand?
<jtimon>
talking about wild precitions: the market will demand non-backdoored hardwardware or humans will become extinct
<bsm117532>
jtimon: I'm not prediction, I'm stating an obvious fact that current fabs cost in the billions.
<jtimon>
bsm117532: ack
<jtimon>
well, no
<bsm117532>
If it could be done on a small scale cheaper, Intel/Samsung/GlobalFoundries/TSMC would be doing it.
<jtimon>
that's if you want to do things "right", you can also do things "cheap"
<jtimon>
exactly, it's about scale
<jtimon>
closed hardware designs greatly benefit big scale players
<bsm117532>
I'd bet a hobbyist could set up a small scale fab with ~micron feature sizes for a few million. Now how are they supposed to compete with the speed and scale of the billion dollar 14nm fabs?
<jtimon>
people could demand free hardware, and that may save us from the 2012 holocaust, or something
<jtimon>
free as in free speach, of course
<bsm117532>
Open ASICs for Bitcoin or other currency would be one of the best things to happen.
<jtimon>
well, I also like those "re-design-the-circuit-without-buying-another-one" things, how were they called?
<bsm117532>
FPGA
<jtimon>
yeah, though, as all swords...
<jtimon>
well, actually not all swords have two sides, but I don't think there's a FPGA design that won't turn against you if you are stupid and let someone else control it
<bsm117532>
Same goes for not reviewing every line of source code you compile though...
Burrito has quit [Quit: Leaving]
<jtimon>
yep, which brings us back to determinisic builds
<kanzure>
oh that's easy, we just need to emulate a billion brains and have them manually review each line of source code, then hash their brain states from every step and then uh... um...
<jtimon>
can we do that in hardware?
<kanzure>
even with deterministic builds you cannot guarantee that a human looked at each line (or that the person who looked at each line was properly evaluating against all known security concerns)
e4xit has quit [Quit: Right I'm out!]
<jtimon>
no, the solution is to reduce the code to a few lines and 10 or so paradigms
<jtimon>
kanzure: you raise a very good point: in the end, you have to trust other people to some extend to trust your own system
<kanzure>
i hope not
<jtimon>
well, I wish not, but that's the state of things for me
<jtimon>
I trust libsecp256k1, but I haven't read it
<jtimon>
just because I trust the people that tell me they have reviewed or written it
<kanzure>
oh weird i thought you would have reviewed libsecp256k1
<jtimon>
kanzure: I have reiviewed some parts of the API
<jtimon>
but not thte guts
<jtimon>
sipa has explained me some of the guts, but many times I demanded higher level for the explanation
<vyvojar>
bsm117532: isnt there something pipelined too? this is low gate.
<jtimon>
I thought I always said clearly that I wasn't reviewing crpyto guts for now
<bsm117532>
vyvojar: Don't know. Link?
<bsm117532>
I'm looking at some of these FPGAs which let you add DDR4 RAM...seems to me that "memory-hard" PoWs will fall to FPGAs soon enough.
<jtimon>
oh, who would have thought that "hardware-will-lose-this-time" wouldn't work...
<jtimon>
hardware is just more efficient software (but the more special the more expensive)
<jtimon>
I say, gamers are special enough, let's use their hardware
<jtimon>
many cores
<vyvojar>
bsm117532: imo not soon enough, but eventually
<vyvojar>
as long the miners are not thread bound on a gpu, but truly only by gddr5 throughput.
<bsm117532>
The major cost of GPUs is not RAM, it's the GPU, and these memory-hard algos keep the GPU pegged. So it seems there's room for an FPGA to be faster than a GPU. Also it's a small incremental cost to keep several copies of the memory-hard data structure.
Newyorkadam has quit [Quit: Newyorkadam]
<vyvojar>
bsm117532: cpus get faster than ram. all. the. time.
<vyvojar>
eventually it decouples
<vyvojar>
especially things like eth or litecoin are doing seem to be bandwidth bound. compute stays idle, all threads stay blocked until next cacheline of random accesses is filled.
MaxSan_ has quit [Ping timeout: 276 seconds]
<bsm117532>
I've always wanted to fool with FPGAs... --> ##fpga
<pigeons>
anyway, as discussed here years ago, the small energy effiency of even an otherwise ineffienct asic, which can of course be made for anything, basically wins