Topic: [ActiveMining] The Official Active Mining Discussion Thread - page 111. (Read 479317 times)

There are open-source FPGA implementations they could have used.

It's a pretty simple business model: You take an open-source FPGA hasher implementation, hand it over to eASIC, and they generate wiring patterns for their nextreme (or whatever) configurable ASIC cores, and send those files off to the fab for production. Pretty easy even at 28nm, but a lot more inefficient then a standard-cell or custom design.

Even if they're not using the open-source FPGA code, writing an SHA256 hasher in verilog (or whatever) probably isn't a lot of work.

Interesting. Does anyone have any idea on what the potential status is on the avalon chips? I guess we are just speculating. Are they in transit? Stuck in customs? Still being manufactured? What's the deal?

Here's ActM's chance to get those coins back and try again!

Yeah but...

ActiveMining doesn't have existing FPGA designs that need converting to ASIC. From what I understand, ACTM will be ASICs from the get go... maybe I'm not getting it?

Yes, as opposed to designing something from the ground up as an ASIC.

An FPGA works by having a grid of logic units, and then using switches to create links between them.
These FPGA->ASIC designs work by getting rid of all the switching stuff and replacing them with with metal (i.e. wiring them together).

The result is a microchip that does the exact same thing as the FPGA did, but requires way less space.

This is obviously much cheaper and easier to do then designing a new ASIC from scratch. But it's less efficient then doing something like a standard-cell design like what KnC is doing.

A further level up from that is doing a 'full custom' design where you actually specify the exact images you want printed on the chip layers. This is the most difficult but you can maybe take advantage of various electrical properties to make things more energy or space efficient. This is what HashFast (and I think cointerra) is claiming to do at 28nm, and what Labcoin is claiming to do at 130nm.

Sooo.... if ActiveMining is using the 'easicopy' stuff, or whatever they call it, then their chips are not actually going to be that good compared to the competition.

avalon and asicminer have the worst chips in existence. what horrible failures they are.

Yes, as opposed to designing something from the ground up as an ASIC.

An FPGA works by having a grid of logic units, and then using switches to create links between them.
These FPGA->ASIC designs work by getting rid of all the switching stuff and replacing them with with metal (i.e. wiring them together).

The result is a microchip that does the exact same thing as the FPGA did, but requires way less space.

This is obviously much cheaper and easier to do then designing a new ASIC from scratch. But it's less efficient then doing something like a standard-cell design like what KnC is doing.

A further level up from that is doing a 'full custom' design where you actually specify the exact images you want printed on the chip layers. This is the most difficult but you can maybe take advantage of various electrical properties to make things more energy or space efficient. This is what HashFast (and I think cointerra) is claiming to do at 28nm, and what Labcoin is claiming to do at 130nm.

Sooo.... if ActiveMining is using the 'easicopy' stuff, or whatever they call it, then their chips are not actually going to be that good compared to the competition.

Correct me if I'm wrong - total layman here: It is my understanding that the above process you mention is for customers that have existing FPGA chips in their product line and want to convert, fast and easy, over to ASIC.

I see some links about eASIC. But one of the links is this: http://www.easic.com/high-speed-transceivers-low-cost-power-fpga-nre-asic-45nm-easic-nextreme-2/easic-nextreme-2-fast-turnaround-asics-manufacturing/

which is their 'fast turnaround' process that actually produces slower, less efficient chips per mm^2. If they are using that process it would explain why their specs are so much lower then their competition's. Unless the dies are a lot smaller, that's going to mean much higher costs per chip.

I understand that eASIC is good at what they do. But if they are using their automatic fpga->ASIC process their chips are going to be slower and less efficient then their competitors.

___

Is the main reason you think ActiveMining will come out on top is because they are an eASIC customer? Or is there any other reason?

It's true, but again they make it up in turnaround (which is much more important in bitcoin terms)!

produces slower, less efficient chips per mm^2

I suggest you read all of this, and click links: https://bitcointalksearch.org/topic/activemining-overview-and-speculation-thread-254930. There are a number of reasons it is a strong competitor. Many of us wondered what you are now asking you have the benefit of reading most of the goods in one location.

Sorry... who is ORSoC working with then?  They normally work with eASIC, and I don't recall them ever saying they weren't working with them this time.  I suppose there are other options they could be going with. Maybe they haven't said.

I don't know how exactly the financial relationship between ORSoC, eASIC and KnC works.  But ORSoC employees sit on the KnC board, and ORSoC announced their deal with KnC on their website. You can see their announcement here  


But basically the people at KnC are good friends with the people at ORSoC and some of them sit on their board of directors.

Also, from what I understand when you submit a chip design the people you submit it too do actually check it to make sure it does what it's supposed to do. So people from eASIC are going to at least be looking at the design, even if they don't actually do the design work.

I find HashFast kind of shady, and it seems like they're way over-charging their customers.  However, they also have a close relationship with Uniquify.  Uniquify put out that letter talking about their chip design for HashFast and HashFast's offices are right inside Uniquify's building.

I don't know much about the relationship between Labcoin and SZICC, their website is all in Chinese.  But labcoin is designing their own chip, and you can even see low res images of the layout on their site (that info is a little out of date, but you can scroll down and see the chip layout images).  They're chips are pretty low res anyway so probably won't be that difficult or expensive to design.

And of course the people at cointerra have a lot of experience designing high-resolution ICs.

Soo... I don't really get why ActiveMining's chips are going to be better then the competition because they're working with eASIC when some of their competition is also working with eASIC, or competitors like Uniquify, while also having their own designers or working closely much more closely with a design company.

No no, you see that's wrong and was already called out as such. No one has any other information pertaining to a deal between eASIC and any other bitcoin mining operation.

their competition is also working with eASIC

Of course I'm not at any of the companies and only they would know, but -right now- eASIC aren't involved in KnC as far as anyone is saying (in any of the camps).

In future, who knows.