Block Erupter USB
-
[quote name=“Tuck Fheman” post=“5210” timestamp=“1369195377”]
I (as well as everyone else) can see this really expanding crypto’s …[url=https://bitcointalk.org/index.php?topic=212085.0]https://bitcointalk.org/index.php?topic=212085.0[/url]
[/quote]Yeah… I’m not sold. BFL is offering 5 Gh/s at a similar price point. Although you may have to wait another year to get yours.
The other problem is it’s sha256, not scrypt.
-
[quote name=“Kevlar” post=“5213” timestamp=“1369196347”]
The other problem is it’s sha256, not scrypt.
[/quote]It won’t take long I’m sure.
-
[quote name=“Tuck Fheman” post=“5216” timestamp=“1369197003”]
[quote author=Kevlar link=topic=658.msg5213#msg5213 date=1369196347]
The other problem is it’s sha256, not scrypt.
[/quote]It won’t take long I’m sure.
[/quote]I’m of the same opinion, which would negate the entire reason why people think scrypt is superior to sha256.
Personally I think it’s a pointless distinction that’s only relevant for another year at best. All it’s done is create the opportunity for more industry, it hasn’t actually solved the overarching problem in the long run, just put it off by a year. But hey, I’m all for capitalism, and if as a result another industry is born (dedicated scrypt mining hardware) then who am I to complain? It’s just that once it happens, the distinction will be pointless, and we may as well have just stuck with sha256 to begin with… but someone will make a shitton of money in the meantime. I just think that this is a lateral innovation as opposed to progressing the problem. Real progress in this department would look like making the Proof of Work useful beyond converting kilowatts into heat.
-
Yup it will happen for sure, I was talking about this earlier in the year over on the Litecoin forum, after all they are only [b]A[/b]pplication [b]S[/b]pecific [b]I[/b]ntergrated [b]C[/b]ircuits when all is said and done.
-
Fast scrypt mining is a problem even for FPGAs, so don’t expect it soon.
Block Erupter USB is a nice toy to become obsolete in a few months. You can buy a HD7970 for 2 BTC and get better performance. The only downside is electricity cost. On the other hand, the card’s value decreases much slower over time.
-
wow that’s pretty amazing :D
-
If these where priced at .5btc not only would they sell out but they would get a few months preorders. Still its very interesting to see what is being developed to hash out coins.
-
[quote name=“ghostlander” post=“5537” timestamp=“1369253495”]
Fast scrypt mining is a problem even for FPGAs, so don’t expect it soon.
[/quote]It’s a problem because of the memory access required. The design of scrypt mandates lots of memory access. So while you can design a ASIC that does SHA256 in a fairly straightforward manner simply by keeping everything in local registers, doing so for scrypt presents a different challenge: Either you have to design it with an ungodly amount of registers (good luck writing the firmware for THAT beast), or you equip it with some external RAM (another chip on the board), and in doing so create a whole new set of problems for yourself, because now your ASIC has to talk to the RAM through a controller (another chip on the board). The alternative to this approach is to build it all into your ASIC design; such things are usually referred to as Systems on a Chip (SoC). Either way you slice it, the complexity is not insignificant.
But it’s not unachievable either. It’s just not profitable right now with the price of scrypt coins being under $50 dollars. Give it another year.
-
[quote name=“Kevlar” post=“5581” timestamp=“1369262430”]
[quote author=ghostlander link=topic=658.msg5537#msg5537 date=1369253495]
Fast scrypt mining is a problem even for FPGAs, so don’t expect it soon.
[/quote]It’s a problem because of the memory access required. The design of scrypt mandates lots of memory access.
[/quote]I know. Adding external DRAM or SRAM to FPGAs isn’t cost effective usually, though there are FPGAs with built-in DRAM controllers like Xilinx Spartan-6 LX150. There are also FPGAs with large numbers of cells which can be configured as local memory. In fact, you don’t need much memory without parallelisation:
[code=“scrypt.c”]const int scrypt_scratchpad_size = 131583;
char scratchpad[scrypt_scratchpad_size];[/code]But you need it fast. 32-bit interface inside most FPGAs and their low clock speeds lead to bitter failure.
-
Great Post DigitalDoom 8)
-
Avalon tried to prevent a monopoly.
But when they failed to produce barebones units, they created a massive scam opportunity.
No barebones? No buying. Sorry. Is it really so hard to put this stuff on a standard PCIe slot card? So there can be an economy?
Most people in crypto are geeks anyway. Mom and pop do not know what a hash is.
-
So Coblee pretty much gave in to the idea of ASIC for scrypt dominating Litecoin in the not to distant future. He had no comment on doing anything to stop it, he seemed to be very agreeable with the entire notion.
Did a market just open up (well, in a year)?
Can Feathercoin become ASIC-proof, ASIC-unfriendly, however you want to put it?
I’m not positive but I thought I saw a post on the forum concerning this but I could not find it. Zerodrama, was it you posting about it?
-
What’s the problem of buying a scrypt ASIC if it comes out? If there are offers from several sources, the prices are going to be competitive. GPU mining is near end. Give it a few months for SHA-256 and maybe a year or two at most for scrypt.
-
[quote name=“d2” post=“8089” timestamp=“1369923359”]
[quote author=Tuck Fheman link=topic=658.msg7974#msg7974 date=1369876354]
So Coblee pretty much gave in to the idea of ASIC for scrypt dominating Litecoin in the not to distant future. He had no comment on doing anything to stop it, he seemed to be very agreeable with the entire notion.Did a market just open up (well, in a year)?
Can Feathercoin become ASIC-proof, ASIC-unfriendly, however you want to put it?
I’m not positive but I thought I saw a post on the forum concerning this but I could not find it. Zerodrama, was it you posting about it?
[/quote]ASIC-proof? No. The reason is simply because as previously mentioned ASIC is only an acronym for Application Specific Intergrated Circuits. It will be a while I think before any company starts to market them, but read of a few people who are doing at least hypothetical designs for this purpose. It’s certainly not something many will attempt due to hardware costs vs profit, but this will change soon enough as scrypt based alt coins gain in value and then I have no doubt BFL will start pre-orders 2 years before they have a prototype.
[/quote]Youre wrong there man! BFL will announce the preorders of ASICs for scrypt the next month…!!! JAJAJAJA
I dont think that adding 3GB of ram to an ASIC could be that hard … but … where can the GPU miner go then?
-
Good luck with that. You’ll be waiting over a year for BFL to actually start shipping them. ::)
-
wow this is crazy
-
Nice
