Topic: We discuss the prospects of an LTC based FPGA. (Read 1300 times)

7950's have never taken 80 watt man.  At best it will take about 230w overclocked.

I thought a bare 7950 uses 80 watt at std settings? (Mine 7970 says 80 amp @ 1.1 volt - which is slightly above but not that much)

 (not including the rest of the system)

Anyway - following this thread

Oh and as an example if you have a 7950 that runs at 250w for 600khs that would be about  2.4khs/w so even if we achieve halfway (which we already beat) we already achieved superior efficiency.

Minimum is 400khs. No price estimates yet, but most likely I will give a perk to the people that buy into first batch since they are going to cost more. Then when the follow up batches are optimized and implemented then we will offer those to the batch 1 customers first.

And when working an fpga this complex, I am not sure if you could safely compare it to what is out there. This is not something like the asics that were created here on the forums or other fpgas people made on a shoestring budget. Our FPGA is a very complex device and will likely be 8-10 layers at final production.

What is the minimum goal you're hoping to achieve (hash rate) by producing these cards, and any price estimates?

No final numbers yet. But we are aiming to beat 10khs/w and so far we are winning that battle

Its been a while ....any rough ball park figures on what the performance specs may be

If you can ...it would be good to start doing the mental excel spreadsheet stuff

Currently have ~ 50m and hit the capacity in regards to my solar setup ... any more would be at 100% power costs and thats not where I want to be going

GDDR5 is not necessary for the fpga to be capable of mining scrypt efficiently. Any fpga we produce or anyone else for that matter will likely have ddr3 on their boards.

With what Damnsammit said, that makes a lot of sense. However the key to it being a hit would be its ability to have an ROI worth investing. I wish these boards could just easily integrate some GDDR5 and we'd be set.

We shall see soon enough. The intent is to save vs energy consumption as you pointed out. Then continue optimization further after that.

That's what I have read, too.

There was a thread on either this forum or the Litecoin forum that a person with FPGA experience had succesfully built and tested an FPGA for Scrypt using .  The result was something like 3KH/s and from everything that I read, he wasn't very optimistic about anything much better than that being acheived.  Factor in electricity and it *might* be beneficial, but it just depends on how much the initial cost is...

Simple economics.  Fixed costs + Variable costs.  A more energy efficient miner (FPGA) will be beneficial in the long-run only if it saves more electricity/hash (variable) than the initial purchase (fixed).  Another thing that I think must be included when talking about Scrypt mining, is that if a GPU can produce 2KH/s/USD (HD7950 @ 600KH/s @ $300) then without factoring the benefit of electricity, an FPGA would need to produce more than that.  Although an FPGA is reprogrammable, it likely will not retain the same value as a high-end GPU if for some reason Scrypt mining because completely unprofitable, either through waning interest or some ungodly ASIC development.

But, if someone could make an FPGA that produced >2KH/s/USD then I think it would be a hit.  As it is, the Bitcoin USB Miners are a hit because of their sexiness and the fact that they are cheaper than a graphics card that will produce the same hashes.  Even if reaching a breakeven ROI is a fantasy, it is a cool thing for a hobbyist to have.  I think if that can be acheived with Scrypt mining through an FPGA or an ASIC then it would be met with much applause.  

I highly doubt that you will ever see an ASIC for Scrypt that will hash at >50MH/s and be as compact as the 50GH/s+ ASICs for BTC.

I think this is the board that was used in the thread I read, but I didn't save all of the information because it didn't seem like a feasible option.  

http://www.dinigroup.com/new/DNK7_F5PCIe.php

FPGA's or ASIC's for LTC will never be as close to the power of bitcoin ASIC's, or in the price. The differences will be enermous.

Well, I think if said technology were to exist, alt currencies would merely just re-position itself to accommodate.

Example Scenario:

ASIC units developed to hash scrypt based coins.

Litecoin becomes maxed out at 84 million, Feathercoin (LTC clone) becomes the new 'litecoin' weighing in at 384 million coins.

New alt coin comes out with a far larger maximum capacity. (1 billion?)

-------------

Alternative situation:

Something else other than scrypt comes along and becomes the new mainstream technical way to mine.

Thoughts?

In the long term, more specialized hardware will always win out if not for hashrate, just for their power efficiency.

If these projects are to become actual currencies ASICs will be the endgame for all of them.

Nobody is going to bother with FPGAs this time around.

64-128K per thread doesn't seem that expensive ASICs might very well be possible.

bump

Personally, I'm very skeptical that this will ever happen, and if it does, we're looking at least 3 years.

Anyone have any opinions on LTC based FPGA's or Asics? Will it happen soon? Long time from now? Never?

This guy seems to think he can pull it off:

https://forum.litecoin.net/index.php/topic,2702.0.html

Opinions and thoughts?