Topic: Supposed ASIC Scrypt Miner | Scrypt ASIC International - page 4. (Read 13030 times)

I'm going to go for it but only order one when batch number 1 opens up. Nothing to lose other that a few Bitcoin. Better than my 25GH's BFL miner that I have yet to see.

I still do not believe it will be anywhere near as efficient as the BTC ASIC or FPGA. It requires larges ammounts of specific memory hardware in order to increase hashing speeds.

The one thing they could have done to increase hash rates is generate a large table of the needed lookups. You can read a little about it here in this post: http://bitcoin.stackexchange.com/a/1307

I was also wondering perhaps there is some way to partition the RAM in order to use less since scrypt does not fully use it all. That would greatly reduce the total RAM needed.

I would never have believed that SHA ASIC's would see the light of day but they have.

I think a lot of people have been in the background working hard and maybe SAI is full of it but I have a feeling others will follow, even if they fail to deliver.

Just a matter if time.

'koolio' just did a UK meetup and announced that he is working on a FPGA Scrypt miner and had a 'proof of concept' present is what I was told today.

This SAI is from UK as well, would be funny if it was his website.

Well if you want to throw away BTC that easily, why not give them to me? I'll take good care of them

It'll be a better than investing in a little black box that could magically transform itself so it could hash SHA-256 instead of Scrypt.

Yeah that idiotic statement reduced the chance of it being legit from 0.0001% to 0.0000% even.  No sense in even wasting time.

A Scrypt ASIC chip cannot just magically switch to SHA and vice versa, this company is telling porky pies, don't send them any money. Bitcoins are valuable why waste them on something as unprofessional and scammy as this dog and pony show.

I just got an email back from them, they have added a forum section now to answer any questions, of which I have loads.

They even have a project running of a miner that can mine scrypt but then with a push of a button mine SHA-256.

This could be just a wind up but if it true this could change a lot of things.

I mean if that wanted to scam, wouldn't they be asking for money right now?

I am tempted to throw some Bitcoins at it and see what happens, its only Bitcoins a few Bitcoins after all.

I have been watching this for a while now. They have updated post now and it explains a little bit more but they are really holding their cards to there chest.

I would like to see a newletter signup section added though.

If there was a buy button on there already I would say 100% scam and stay well away but they seem to be holding off right now.

No, it wouldn't need even remotely near that much. Your GPU's mining isn't limited by its available memory, it's limited mostly by its processing speed and number of registers. Those gigabytes of memory are barely touched.

For ASICs, however, even the relatively small amount it needs (quoted as 128kB per core earlier in this thread; I'm not sure that's accurate, my impression was it's only 16kB, based on 16 byte hash size * 1024 scrypt parameter, but I haven't looked into it that closely so could easily be wrong) is a lot of space vs. a 16 byte SHA256 used by Bitcoin.

That, in turn, means a lot less parallelism in a given mm2 of die, meaning you'd have to make a much larger and more complex chip (or have many of them) to get a rate like 50 MH/s.

Of course, that would use more power. But power use can be combatted by using a smaller lithography size, at the price of exponentially more expensive manufacturing equipment.

So yes, in theory, it may be possible to build a 50 MH/s for 500W of power. Possibly. But you'd need multi-billion-dollar chip fab equipment, and the large chips would both be expensive to produce and have a very high percentage of them with errors, needing to be thrown out.

End result, nobody is going to make that in practice. Not arguing against it being obviously fake.

clearly fake & scam, as said before, a wordpress template?
No contact info?
Stolen images?

Do we really need to go on?

In theory, there's nothing to prevent someone from making an ASIC Litecoin miner. The scrypt function is only resistant to it, not immune. What it does do is force the chip designers to eat up die space with all that memory, instead of using it for more parallelism.

Their claimed rates are... high, but it would probably be possible to at least approach those numbers with current lithography technology. However, the cost of designing, tooling, and manufacturing a chip that powerful would be so huge, it would result in units being sold for at least 7 digits, and nobody is going to pay that for a Litecoin miner. It wouldn't be even remotely worth the cost.

So far as I can see they aren't giving a price yet, but most likely they'll try to match the Bitcoin ASIC designers at low-to-mid 4 digits. If they do, that'll be solid proof it's fiction.

Buy buy buy....    

lol, so true. This smells of scam; the specs are not even kind of plausible. BFL wanted to but they figured that their current scam operation was better

DaT > *

I was just using numbers based on 7950 stats.

How much memory do you think it would need to hash at that rate? You seem more knowledgeable than most at requirements.

Do you think it is actually feasible?

Um it is certainly a scam but what does 230 GB of GDDR5 have to do with anything?  You do realize Script uses none of the main memory on a GPU.  The main memory has far too much latency (latency for loading from GPU main memory is about 200-300 clock cycles).

A GPU which needed to use main memory would work for 10-20 clock cycles and then up to 300 (totally idle) and then work 10-20 and then wait up to 300, etc.  That is the whole point of scrypt.  A GPU which needs to use main memory on GPU would be beyond useless for hashing however the scrypt used by Litecoin (and others) is intentionally weakened to only use 128KB about 1% of what is recommended for LOW SECURITY (i.e. realtime logins) applications.  That's right Kilobytes of memory.  It uses the L1 cache located on the GPU processor itself (much like it uses L1,L2,L3 cache on a CPU).

To be honest, if I would have the capabilities and the funds to produce ASIC's, I probably wouldn't bother with designing and producing a fancy case.
It would only drive up the costs and people would buy it anyhow, I would just look for a case that fits and that's it. Don't think any miner would say "Cool a Scrypt ASIC, but I'm not buying it without an exclusive case!"  

lol I was just pointing that out in case someone didn't notice.

It is a lot less efficient to use a prebuilt case as you'll have to design your boards to their specifications.

Anyways, I sent them a message about why their dimensions & weight are off.

Should be fun reading the reply 

Not commenting on that. That is just a blatant fail on their part. Even more proof they are scammers.

You just wrote some Sherlock response to something blatantly known, it was a CM case.