Topic: 1GH/s, 20w, $500 — Butterflylabs, is it a scam? - page 35. (Read 123112 times)

There is no significant market for that level of SHA-256 hashing outside of Bitcoin (except maybe SHA-256 brute force attacking).  

Think about it for a second.  Even a crappy CPU can perform tens of millions of hashes per second (remember a Bitcoin hash is almost 2 complete SHA-256 hashes).  While SHA is used for things like keeping passwords secure even Google doesn't need the ability to process 2+ billion logins every second.   If ever single human on the planet had a gmail account and these boards had enough bandwidth it would only require 3 boards to ensure every the entire planet could be authenticated every single second.


If 3 boards could (which they can't due to other limitations like bandwidth) handle 6 billion simultaneous logins how large would you imagine the market is for these boards?

SHA-256 has other uses (hashing for signatures, resource organization, etc) but likewise there are no applications (military, financial, or otherwise) that need 1 billion hashes per second.  TLS & SSH uses SHA but an individual server is working on the scale of thousands of session or less not billions.  SHA is used for hashing a document in order to digitally sign it but even for the US government an entry level CPU has enough computation power to hash thousand page document in a fraction of a second.  For stream hashing (hashing large amounts of data) the bottleneck is the ability to get data TO the CPU/FPGA/ASIC.  

For example this board is able to process 2 billion hashes per second based on 512 byte blocks.  Thats roughly 1 trillion bytes per second.  So to hash documents at that speed would require an I/O bus (and main memory, or hard drives) capable of 1 TB per second.   No hashing engine can hash faster than data can be delivered to it.

This ( only good for SHA256 algo when FPGA should be able to flash FW and make it good for all algos ) sounds like they are using some existing SHA256 processor and not expensive FPGA from Xilinx with crap performance and high wattage and high pricing. Maybe this is real indeed and those targets can be achieved with a processor designed to do only one thing and one thing only at the best of its ability : BTC mining. Anybody know of an existing CPU for SHA256 in particular ? Maybe they are using military technology

I am especially curious for the rigbox.

I would like to get some additional technical details here.

As the number of cards (32) used does not correlate with the proposed hashrate (54,4 Gh/s) altough it was stated that there are the same cards used.
It would be interesting wich system is employed here.

Also i'd like to know wich price range can be expected here and if it is possible to start just with the rig box including one or two cards to limit startup costs.

Our processors are designed for SHA256 specific applications, 1 to 8 level multi step processes only.

Regards,
BFL

Says it has upgradable firmware and that they will be shipping in waves as they finish the fw for each application.

That's not saying I believe it's real...

There's a fundamental misunderstanding here about how and why FPGAs are so good for Bitcoin mining that I don't even have the time to begin addressing, but I'll say this much: Only two algorithms have been chosen thus far for mining Bitcoin and its associated forks: SHA256 and sCrypt. sCrypt is used for "CPU friendly" currencies and is CPU friendly explicitly because it uses more RAM than is typically available in on-chip cache for both GPUs and FPGAs, but not more than is typically available in L1/L2 cache on your CPU. FPGAs are a poor choice for sCrypt mining for the same reason that GPUs are: they require too much on-chip cache.

That said, if the BFL boxes turn out not to be vaporware there's only three options for what's inside: an FPGA, an ASIC or some kind of existing SHA256 hardware. In all cases but FPGA the device will likely only be good for the one purpose of its design. Even if they are FPGAs it's unlikely you'll be able to accelerate sCrypt with them since sCrypt was explicitly built to avoid such acceleration.

Thanks for the answers. As I mentioned before, you can put all the scammer speculation to rest by inviting Inaba to come take a quick tour of your facilities in KC. Will you please invite him over this week?

*Awaits public demo that will change the face of BTC mining FOREVER*

Can your product also be used for things like :

-WPA and rainbow tables
-vanity BTC address generation
-integer dependent general custom computation ( e.g. is there any way to make those FPGA universally compatible with OpenCL and other code like GPUs are etc. so they can also be used to mine say SolidCoin not only Bitcoin etc. )

Thanks !

Good luck, And dont forget, Goat is paying 10%  of your costs. In his words ""I will pay 10% of any ones costs to buy any of this guys products just for the fucking lulz. ". Chances of him honoring that seem vastly slimmer than BFL coming up with a product however.

Or you could do both.

All you that were betting with each other, screw that, I'll bet by placing a pre-order. If I win, I come out more ahead than if I won a silly bet.

We plan on demonstrating our BitForce Single product before accepting pre-orders for our Rig Box system on the 25th of this month.  The date will be published here once it's fixed.

BitForce Single PCB's are used in the Rig Box platform.  However, the Rig Box also employs parts unique to itself.  Why would we go to the trouble for such a small market?  We didn't.  Our product design is influenced by other factors beyond hash mining.

Regards,
BFL

Amen!

Is there any news on initial delivery or demonstration dates?
Also the rig box, is it using the same board as the singles, and if so, how come 32 of those get ~54x the throughput?

So why does your website state your company name is Butterfly Labs Inc.  


A name which isn't registered in any state.

Are you Butterfly Labs Inc?  If so then why is your WY registration BF Labs Inc?
Are you BF Labs Inc? If so then does your website state you are Butterfly Labs Inc?

Close enough, or abbreviated isn't sufficient for the legal name of a company.  Also there is no fictitious business name license issues in either state (which is another legal way for a company to operate under multiple names i.e BF Labs & Butterfly Labs Inc.).

Hi guys.

I see there were some well intentioned questions.  Here are the answers.

1.  Corporate registry - The company is a 'C' corp registered in the state of Wyoming as BF Labs Inc.  Why Wyoming when you're not based there?  Mostly due to it's favorable tax and legal climate.  Operating in the same state of incorporation exposes the company to additional tax liabilities.  You might also find it interesting to know that the majority of US domiciled fortune 500 companies are either Delaware or Wyoming corporations for these same reasons.  Intel is a good example.  

2.  Drivers - We're currently working on driver support for Win / Linux / Mac.  I can't say we'll have all three done by release.  However, at this point it's very likely.  Windows, then Linux, then Mac is our dev priority order.

3.  USB - Someone asked how many units can be hosted by one computer.  127 maximum units can chain from each of your computer's USB ports.  Since the actual data passing through the USB chain is very small, there is no exaustion.  However, you'ill need a very large USB hub to pull this off, and the hubs themselves consume a USB id, so in practical terms the total number will be something less.

If you have any additional good natured questions, I'll be happy to answer them.  On the other hand, if the thread degenerates, I'll be happy to leave you wait for our public demo.

Kind regards,
BFL

PS.  Sorry to be crabby, but you guys deserve it.

Waiting for the scam to arrive ?

4-8 weeks probably means 8+ weeks. If they meant 4 weeks, they would have said 2-4 weeks.
Marketing 101!

-rph

Well the thread was started 18 Oct. BFL has a 4 to 8 week delivery time for pre-order units.  So we are approaching the delivery time zone.  Some time (unless we hear otherwise) on or around (hopefully before) 18 Dec, someone will receive a single.

Let's see what happens.

However, if they were ASICs, they could be designed with fuses and then have no redesign cost when they want to increase the output. The idea was Gmaxwell's, not mine

You are far too optimistic. Better cooling will not cause a 70% leap in efficiency.