LIGHTNINGASIC LA100M,100MHS SCRYPT Miner, USD1999; LA1THS, USD1750.shipped out!

[ISAWHIM]

Thoughts?  Better ideas?

Wattage and cost/watt is needed for those calculations. Since the units are non-standard and wattage changes per operation.

Also, for dual-mode, wattage is altered. (Could do profiles for wattage, but not for cost/wattage, which is a user-specific value.) Since, at some point, there may be a Scrypt-coin that is high, and a SHA-coin that is also high, and mining both could potentially be an option. (Though, that would be a bonus which fades fast.)

Might also want to allow a "Do not mine" list LUT. For times when POS-stalling coins are horribly inaccurate for "reward". Also, for "lotto-style" coins that the calculations for reward are never correct, due to poor calculations of the lotto randomness. (Those are coins with a "?" next to the block-rewards on CoinWarz. There are more losers than winners, but the value is calculated as an average of high/low, which is inaccurate to any actual value, at any point in time. While POS-stalling is when POS blocks boot-out any actual POW block rewards, because the POS causes a block-reset too fast, due to its low difficulty. Thus, cutting reward in half, as it still shows as a valid "block" for that hour, until the formula realizes it isn't a reward, then re-adjusts to a lower diff later. A lower diff that someone-else gets, if you hop off the coin.)

Also, limiting high-diff coins, if you are a solo-miner and not a pool-miner. (Since hopping to a high-diff coin will be zero reward unless you actually stay until at-least one is found, which would be when the value is no-longer rewarding.)

[eightcylinders]

Wattage and cost/watt is needed for those calculations. Since the units are non-standard and wattage changes per operation.

Yeah, I think I have to add at least cost per watt since people have different costs.  The wattage should be calculated from the frequency and mode.  At max stable frequency of 700Mhz, each unit (5 grid chips) consumes: 21watts (BTC mode), 22.8 watts (dual mode) and 2.2 watts (scrypt mode).  So the I think the only thing that will be required is knowing the cost per watt (or Kw) of electricity.  The rest is calculated. 

The main idea is that cgminer could then select which of the 3 modes to use to maximize mining profit, without user intervention.

Might also want to allow a "Do not mine" list LUT. For times when POS-stalling coins are horribly inaccurate for "reward". Also, for "lotto-style" coins that the calculations for reward are never correct, due to poor calculations of the lotto randomness. (Those are coins with a "?" next to the block-rewards on CoinWarz. There are more losers than winners, but the value is calculated as an average of high/low, which is inaccurate to any actual value, at any point in time. While POS-stalling is when POS blocks boot-out any actual POW block rewards, because the POS causes a block-reset too fast, due to its low difficulty. Thus, cutting reward in half, as it still shows as a valid "block" for that hour, until the formula realizes it isn't a reward, then re-adjusts to a lower diff later. A lower diff that someone-else gets, if you hop off the coin.)

I am not sure I am following this, can you give me some more?  Couldn't this be fixed by creating a server/deamon that provides a feed for a fictitious coin?  The fictitious coin would have whatever attributes are required (to account for LOT style coins for example where an "average" may not be representative without taking into account variance).  I was thinking that taking that part out of cgminer would be a good idea so that server feeds for new coins could easily be incoporated without having to mess with the cgminer code any further.  Or am I missing your point? 

Also, limiting high-diff coins, if you are a solo-miner and not a pool-miner. (Since hopping to a high-diff coin will be zero reward unless you actually stay until at-least one is found, which would be when the value is no-longer rewarding.)

I thought about doing something for solo mining, but the coding would be too much for right now; you would have to create a database of all alt-coins and be able to keep price/difficulty curves up to date for all of the coins . it would also make running on a low power embedded device like an rPi a real challenge.  I would love to do this project, but I know my limits!

[eightcylinders]

One thing missing from the original post...

Where to get these... (Prices in USD would be nice too. We can figure-out BTC-rate later, since BTC value is not constant. Or LTC, if accepted.)

asiabtc has quoted (and accepted from myself and at least one other in this thread) $2600 for the 3Mh/s set which consists of a controller, power supply, and qty 10 of the hashing units.  Each hashing unit contains 5 gridseed chips (total of 50 gridseed chips, or 10 hashing units).  Specs for the gridseed chips are here: https://github.com/gridseed/gc3355-doc.  He is also offering express shipping for $150 US.  He accepts BTC or USD at least.  I paid in BTC using preev conversion rate.

He told me he expected shipment by Feb 15.  I asked what the absolute latest ship date would be and he said Feb 20 FWIW.

[asiabtc]

We need more info on the 6MHs and 12 MHs model.

power consumption, and price?

agreed!

there is no clear ETA, price of two new models. still under developing. when its ready,i will announce in forum.

[note235]

We need more info on the 6MHs and 12 MHs model.

power consumption, and price?

agreed!

there is no clear ETA, price of two new models. still under developing. when its ready,i will announce in forum.

is this the 2nd gen model or just 2x of the first?

[ISAWHIM]

I thought about doing something for solo mining, but the coding would be too much for right now; you would have to create a database of all alt-coins and be able to keep price/difficulty curves up to date for all of the coins . it would also make running on a low power embedded device like an rPi a real challenge.  I would love to do this project, but I know my limits!

To keep this from derailing... I will be brief...

For solo-mining, a simple database of "history", would suffice for "actual reward". To compare to any websites "projected reward". (EG, coinchoose data, coinwarz data, where-to-mine data.)

As for the "do not mine"... or "only mine"... All coins have a 3-4 letter ID, easy to identify with any of those sites. That, or just let the user make a list of connections they only want to connect to. (As well as through cryptsy-API, to confirm actual "now" price data.)

For solo-miners, we would have to have a wallet setup per coin. As long as each port was identified with each coins letters, matching should be easy. History would show what we actually get, compared to "estimates", and getting 50% rejects would indicate that the coins value should be 50% less then estimated. (Which is what you see happen in super-low diff coins, POS rushes, and fast block-time coins. You would never see that info in a pool. All shares are accepted, even if the coin is actually producing nothing and the rewards are way off actual values. Thus, why solo-mining is more rewarding for those coins.)

BTW, at certain diffs... rejections are predictable, when also taking the block-time into account. (Eg, diff 0.2 @ 20sec block-time = 50% rejects. diff 0.2 @ 60sec block-times = 25% rejects. diff 0.2 @ 2min block-time = 5% rejects. Add about 50% more rejects if the coin is a POS/POW coin. On the far end... diff 10.0 @ 20sec block-time = 25% rejects. diff 10.0 @ 2min block-time = 2% rejects...)**

** Those are ficticious values, except the 0.2 @ 20sec, which I know is actually worse. (More like 65%) The rest on the high-side, are roughly accurate, based on my solo-mining limitation of mining below difficulty of 10.0 coins only. (Also note, you can use this info to "correct" the horrible estimations that all these website value-estimators provide. Reality is more important than "perfect-world-estimates".) A LUT (Look-up-table) or formula would suffice for this adjustment. However, it is reverse-exponential in nature.

Sadly, unless you know the pseudo-random code... there is no way to predict the "lotto-winnings" of coins that use random "super-payouts", but examining the block-chain data can tell you when those are paid-out/hit, thus, immediately devaluing the coins value. (Telling you to stop mining it, or adjusting the reward potential.) Though, they all tend to use the same code, and others actually DO know which blocks will pay-out, and only mine when those coins reach those pay-out times. They leave others to mine the low-value crap. (That is your supportive community! Also, that is your coins support, for you, as a miner.)

Then there is instant-adjusting coins. (Those are ones that change diff per-block, which charts do not correctly display. They take averages or just spot-check diff, which may not be a real indication of actual value adjustment.)

Yes, that was the brief reply.

[BtcBling]

We need more info on the 6MHs and 12 MHs model.

power consumption, and price?

agreed!

there is no clear ETA, price of two new models. still under developing. when its ready,i will announce in forum.

is this the 2nd gen model or just 2x of the first?

Im also interested in these new models, any ideas about pricing would be great. Do you have updated web interface in english for new models?

[S4VV4S]

We need more info on the 6MHs and 12 MHs model.

power consumption, and price?

agreed!

there is no clear ETA, price of two new models. still under developing. when its ready,i will announce in forum.

is this the 2nd gen model or just 2x of the first?

Im also interested in these new models, any ideas about pricing would be great. Do you have updated web interface in english for new models?

+1

I would also like to have a rough estimate on the price.

[film2240]

I'm a Hero member who's interested in reviewing this ASIC as I'm very impressed with the hashrate vs power use.I have a large twitter following (over 160K) so if you send it to me,I'll make sure to give you good quality feedback .

Is this compatible for mac users or for PC only?Can you embed your drivers into Asteroid/Mac miner by sending them to the development teams of that software?As it'll allow me to test it then

[worldlybedouin]

We need more info on the 6MHs and 12 MHs model.

power consumption, and price?

agreed!

there is no clear ETA, price of two new models. still under developing. when its ready,i will announce in forum.

is this the 2nd gen model or just 2x of the first?

Im also interested in these new models, any ideas about pricing would be great. Do you have updated web interface in english for new models?

+1

I would also like to have a rough estimate on the price.

I got a resposne from asiabtc via PM on this subject.  He has informed me that the new products are still being worked on and he does not have specs or prices yet.

[worldlybedouin]

One thing missing from the original post...

Where to get these... (Prices in USD would be nice too. We can figure-out BTC-rate later, since BTC value is not constant. Or LTC, if accepted.)

asiabtc has quoted (and accepted from myself and at least one other in this thread) $2600 for the 3Mh/s set which consists of a controller, power supply, and qty 10 of the hashing units.  Each hashing unit contains 5 gridseed chips (total of 50 gridseed chips, or 10 hashing units).  Specs for the gridseed chips are here: https://github.com/gridseed/gc3355-doc.  He is also offering express shipping for $150 US.  He accepts BTC or USD at least.  I paid in BTC using preev conversion rate.

He told me he expected shipment by Feb 15.  I asked what the absolute latest ship date would be and he said Feb 20 FWIW.

I believe eightycylinders is referring to me as "at least one other" in this thread.  I did place an order with asiabtc for 3.44BTC (~$2750 at Bitstamp spot price at time of order).  I also have the same understanding on the ship dates.  I purchased a "package" called LA3M which includes the 10 hashing devices, PSU, USB hub, mico-controller (WiiBox), and all cables.  I plan on placing a full review here and on litecointalk of my purchase experience and how the units perform over an extended period of time.  I'll also provide what I can in terms of "lessons learned" if the bulk of the information/software is still not re-written into English and I end up having to decipher the Chinese versions.

[SheepOnDrugs]

Definitely interested in those new setups, keep us updated!

[eightcylinders]

For solo-mining, a simple database of "history", would suffice for "actual reward". To compare to any websites "projected reward". (EG, coinchoose data, coinwarz data, where-to-mine data.)

<<snipped>>

Okay I get what you are saying.  This is really like multipool.us except in cgminer.  This is *the* project I want to do, but it is much larger in scope than "merely" modifying cgminer to profit optimize the modes of the LA3M.  Breaking this into pieces, there are two somewhat interrelated projects: (1) getting cgminer to be able to switch modes on the LA3M device (amongst SHA256 only, scrypt only, and mixed mode), and (2) choosing the right coin to mine.  The two projects do intersect because in determining which mode to operate in, you have to pick an SHA256 coin and a scrypt coin for the program to evaluate on an expected profits versus power usage curve.

To simplify matters, and not to put too much load on cgminer (which often is running on low power devices like the rPi), I have separated these projects completely by requiring (for now) that the user must designate the two coins to be used for comparison.  cgminer itself would not select the coins or optimize solo mining by selecting optimized coins (yet).

I left open room for using multipools and merged mining by using coin feed that could be programmed to supply numbers for a fictitious coin so as to match the results of a merged mining pool or multipool; but that is a very different matter as compared to having cgminer optimize amongst a universe of alt coins (for one, my kludge is outside of cgminer; for a second, it doesn't require me to do any work other than handle the input paramater which may be a data feed).   

However, this is admittedly just a first step to getting cgminer to auto-optimize amongst multiple coins.  Once you have it optimizing the power  of the LA3M, it will be easier (still not **easy**) to get it to look at the alt-coin universe and select coin(s) to mine.  I still think that you want the bulk of such a program outside of cgminer, and instead feed to cgminer the optimal coin(s) to mine data.  If you did it that way, my modifications to cgminer would provide about 90% of what you need (once you create a daemon that can feed cgminer the optimal coin mining data)

Sadly, unless you know the pseudo-random code... there is no way to predict the "lotto-winnings" of coins that use random "super-payouts", but examining the block-chain data can tell you when those are paid-out/hit, thus, immediately devaluing the coins value. (Telling you to stop mining it, or adjusting the reward potential.) Though, they all tend to use the same code, and others actually DO know which blocks will pay-out, and only mine when those coins reach those pay-out times. They leave others to mine the low-value crap. (That is your supportive community! Also, that is your coins support, for you, as a miner.)

I don't think that is right, or if it is, the code they are using is piss-poor.  High quality random number generators are very easy to find - and using the same code does not mean getting the same or even repeatable results.  And as a matter of probability (totally counter to intuition) the fact that a random reward block was just found does **not** decrease the probability that another random reward block will be found in a given time.  A miner may *choose* to stop mining at that point, but a program that computed expected values based on probability should not take into account past random events as they cannot be used to predict future random events if truly random.

[FilipZ]

I have checked your pages. Shame its out of shock. I wanna test it in serial setup.

But I'm focused on rentability. 6k USD for 6MG/s is equal to 6x290x rig. There is no difference of initial cost. And if you thing about sell it later, so its not possible in comparison with PC components.

[bobby35ny]

Well, I am in.  Sent 2,600 US + 150 for overnight shipping to asiabtc for the LA3M (3Mh/s total scrypt power, which is a controller, power, and set of 10 of the 300-350kh/s pucks).  He is very responsive on PM fyi for those of you wanting to order.  He lists price in US, but I sent him BTC using preev rate.  Due Feb 15, and he has promised it will ship no later than Feb 20 (I asked for absolute outside latest date come hell or high water, and that is what he gave me).

If one of the testers could post some exact dimensions and ideas for mounting these, that would be great.  I am going to fire up my 3D printer and see if I can make a tower something like a wine rack to put each unit in fan first; planning to make it modular so new ones can be snapped in... or any better idea for those who have them?

What do folks think about trying to optimize these for mining BTC vs. scrypt?  Power requirements are so much greater for BTC, it would seem that it might often be the case that you would mine only scrypt with these, but maybe sometimes it would be better to only mine BTC, or both??  That there are 3 modes seems to present a great option to be able to optimize cgminer real time...

Starting with the github cgminer code for these devices, I would add a "smart" mining option for cgminer that will switch between the scrypt and 256SHA mode depending on maximum profit.  I am thinking that the option would look something like:

--optimize-profit 256SHA-coin, scrypt-coin, averaging-function, time-window, recompute-time

Where:

   256SHA-coin is the desired 256-SHA coin for profit comparison purposes (does not need to be coin actually mined).  This would probably have to be a lookup table or service/deamon that provides information on current difficulty, average block solve time, block reward, and price for a given actual or fictitious coin... fictitious coins could be used for merged coin mining and/or multipool mining operations to provide average difficulty, solve time, etc. but the cgminer program would not be responsible for computing all that. 

   scrypt-coin is a reference to a lookup table or service or deamon that provides relevant info on the the desired scrypt based actual or fictitious coin

   averaging-function is the desired method of computing profits from the relevant coin data (e.g., moving averages, weighted moving averages, stochastics, etc. I should be able to use open source libraries used in stock data analysis very easily here)

    time-window is the time window to use when performing the averaging-function (12 hours, 4 hours, 5 days, etc.)

    recompute-time is the time between recomputation of the averaging-function (e.g, every 240 seconds)

Basically, the modification to cgminer would compute the expected profits from mining in each of the three modes (scrypt only, sha256 only, and both) during the recompute period, subtract out power costs, and then select the operating mode that maximizes expected profits.

Thoughts?  Better ideas?


   
 
    I too would love to know the dimensions!!
-Bobby






 

[asiabtc]

hi, buddy,
if LA3M can reach 4.5MHS for LTC only, how much will you pay for it?
thanks.

[firejuan]

Don't go raising prices again.  US buyers already pay an extra $600.  Yes, I know gridcip raised the price, but the Chinese price is ~12,000 yuan = ~$2,000 for the LA3M.

hi, buddy,
if LA3M can reach 4.5MHS for LTC only, how much will you pay for it?
thanks.

[FilipZ]

Don't go raising prices again.  US buyers already pay an extra $600.  Yes, I know gridcip raised the price, but the Chinese price is ~12,000 yuan = ~$2,000 for the LA3M.

hi, buddy,
if LA3M can reach 4.5MHS for LTC only, how much will you pay for it?
thanks.

yeah, 2000 USD sounds real

[asiabtc]

great works. LIGHTNINGASIC Members.
Lightningasic - gridseed-based rig hashing litecoin on a raspberry pi.
http://vimeo.com/85853611

[rammy2k2]

made a pre-order for 2 x 3 MH kits