Pages:
Author

Topic: What's the best use of my time/energy? (FPGA/Control Tech) - page 2. (Read 2482 times)

c_k
donator
Activity: 242
Merit: 100
tbh, we're better off having someone make http://www.raspberrypi.org/ work perfectly for connecting mining hardware to Smiley
rjk
sr. member
Activity: 448
Merit: 250
1ngldh
We don't really need even more standards do we? What's wrong with USB?
It does too many things, and none of them well at all. Protocols such as 1wire are used in industrial facilities for a good reason - they are cheap as hell, and extremely robust. They are long distance (none of the USB-6-foot bullshit, sure you can make a longer USB cable but no guarantee it will work), and a simple daisy-chain topology instead of the rats-nest inducing star topology with hubs and whatever else.
hero member
Activity: 742
Merit: 500
I would love to see a dedicated, open protocol and interface other than USB that could be used for daisy-chaining many devices. If the protocol and interface were cheap and open, it could be possible to apply the same interface across diverse products, and all control them on the same bus.

There are a number of bus technologies that already exist, but they all need to be properly evaluated, and a cost assigned to them, i.e. cost per node in wholesale manufacturing quantities. Often in these discussions Ethernet comes up and is discussed, but is determined to have a cost per node that is too great to be of value.

To accompany this lofty initiative, an inexpensive and useable head controller/marshaling device needs to also be designed. It could be based on something extremely simple such as Atmega/PIC/Basic Stamp/Arduino, or something a bit more powerful based on ARM and/or x86. It would be self-contained - possibly a USB, JTAG, or serial port for programming, and an Ethernet port for communication during normal tasks.

This way, a single device (or more, if N+ redundancy were built in) could control a wide number of devices from many manufacturers, assuming they could be convinced to include the special interface - thus the need for it to be inexpensive.
We don't really need even more standards do we? What's wrong with USB?
rjk
sr. member
Activity: 448
Merit: 250
1ngldh
I would love to see a dedicated, open protocol and interface other than USB that could be used for daisy-chaining many devices. If the protocol and interface were cheap and open, it could be possible to apply the same interface across diverse products, and all control them on the same bus.

There are a number of bus technologies that already exist, but they all need to be properly evaluated, and a cost assigned to them, i.e. cost per node in wholesale manufacturing quantities. Often in these discussions Ethernet comes up and is discussed, but is determined to have a cost per node that is too great to be of value.

To accompany this lofty initiative, an inexpensive and useable head controller/marshaling device needs to also be designed. It could be based on something extremely simple such as Atmega/PIC/Basic Stamp/Arduino, or something a bit more powerful based on ARM and/or x86. It would be self-contained - possibly a USB, JTAG, or serial port for programming, and an Ethernet port for communication during normal tasks.

This way, a single device (or more, if N+ redundancy were built in) could control a wide number of devices from many manufacturers, assuming they could be convinced to include the special interface - thus the need for it to be inexpensive.
member
Activity: 89
Merit: 10

Why don't you try an advance the Mhash/$ by using your skills to fit a whole unrolled chain into a spartan6 LX100.
From Elden Tyrells 3 ring project it seems this should be possible.
Then make a board with 4 of those. This should give a total board price and performance similar to the BFL single.

hero member
Activity: 720
Merit: 528
Option 2:
I say, forget this, I'm just going to use ZTEX, and make a motherboard for ZTEX daughter cards so that they can be controlled without a PC host, and still have a web interface for configuration/data, price ~$100 to handle multiple ZTEX boards (I need to look at more specs to know precisely how many).

First of all, you seem to be forgetting (or do not know) that there are two other FPGA mining boards out there, both at much lower prices than the Ztex board: Icarus and X6500.

As for your question, there already is a motherboard that can run many USB based FPGA miners (all of the FPGA miners are USB based) and it costs ~$100. It's an Atom based mini-ITX board like this one. I run my X6500s on one, and at one time had 24 boards on it (48 FPGAs) with no problems. So, I vote for the first one, and I welcome the competition!

Another option would be to work on software that supports several FPGA miners, like MPBM.
hero member
Activity: 784
Merit: 500
I went with Ztex for now but it would be nice to see a Competitor in that field. Moving with Ztex was the right move fo me because here in Germany all the other FPGA Boards that are aviable in the maket right now are: 1. hard to get.  2. expensive (with taxes and shipping costs) and  3. they only have limited warranty (ore warranty invalid like icarus)

If you habe the money and abillity to do it then why not:D
newbie
Activity: 59
Merit: 0
Firstly let me state that I will be releasing all source code to date, and I will be helping anyone who needs it.
I'd like to contribute to this community.

That said, after a lot of looking around, designing, talking, and pondering, it seems that ZTEX and BFL have some hard-to-beat boards, as some people made very clear in my other thread. 
Starting a new SoPC to mine bitcoin at this point seems to only want to draw ridicule...

So
I'll float an idea...

Option 1:
Were I to develop a board that was priced 300-350 USD that had the following features:
  • Spartan-6 (or similar) @ 200MH/s (exact TBD, using XC6SLX150 Specs)
  • ARM Cortex-M3 (or similar) Serving Ethernet & Control Functionality
  • Optional USB configuration/communication
  • Standalone Architecture (Ethernet only, no USB PC host)
  • Simple Web Interface & Config (HTTP Server w/ statistics, status, config...)
  • Extensible Architecture (Mezzanine [aka stackable] or similar build)

Would that be interesting to anyone?  I know the hashrate, cost, and power consumption are similar to other competing technologies, but the essentially plug-and-play configuration, web interface, and mezzanine extensibility do make it a lot easier to start using.

Option 2:
I say, forget this, I'm just going to use ZTEX, and make a motherboard for ZTEX daughter cards so that they can be controlled without a PC host, and still have a web interface for configuration/data, price ~$100 to handle multiple ZTEX boards (I need to look at more specs to know precisely how many).

Either of these options would be complemented with a full featured web application and configuration utility, so no more Java, no more terminals, no more OS specific utilities, or any of the other technologies crammed together to make all this work.

If people are interested, I'd love to do this.  I don't mind making "third-party" technology.
Now, importantly, I won't be asking for money to start this project.  This will be all on my own time, on my own dime, until I have a working proof of concept that I can put pictures, screenshots, and demos up of.
As always donations will expedite things, but if for some reason I need something specific, I'll mention it.
What that all means is, I just wanna know if it's gonna be worth it.  It's gonna be a lot of work to do either of these options, and I want to know that at the end, there will be people willing to purchase these.  Basically, all I need is a show of hands.

Thanks so much, and I'll continue to keep things updated.  Expect VHDL from Nanominer over the next couple weeks, and more specifications of the proposed project over the next few days.

Thanks so much!
Pages:
Jump to: