Topic: ASIC testing on main net is wrong, PERIOD. (Read 5701 times)

HOLY CRAP.  Whichever ASIC company is doing this.. is earning a shizzload of BTC right now.  This is so wrong on so many levels.. but can our opinion really change their minds?  I think not.

Many pools are beginning to offer a higher difficulty share. Instead of giving all of the miners difficulty 1 shares, the pool assigns anywhere from 4 - 32 or even HIGHER difficulty. The only difference is the speed the share is completed and the amount of bandwidth used, as both 100 difficulty 1 shares is the same as 1 difficulty 100 share.

The speed the share is completed in will be negligible in terms of variance on large ASIC rigs.

I joke about wire wrap, of course.  I am not sure about your first sentence, but I take feeling you are teasing my about my English.  Look down for explanation in signature.

I am serious but fun for mining chip.  I have many old FPGA developer kits and have been creating small footprint unrolled bitcoin miners as hobby.  Performance is bad alone but together, I can make all worth two dollars per hour spent on hobby.  I breaking even after 30 million bitcoins are found.  Mathematics makes proof !

FPGA conversion to ASIC (by yourself or by vendor) is not too expensive if you have research or student access.  Is cost more than GPU card and less yield than full custom run, but you can work with fab scheduling companies to put chips on margin of wafer already being processed.  Simpler is better, because charge by mm^2 and layers.

I have done for other purposes in grad school and am looking into bitcoin ASIC for fun.  I probably go for many simple hasher chips than many core, so is not commercially viable.  Simple chip to wire wrap, too.  My estimates are less $100 chip, packaged, for 50-100 chips.  Could be more, is always case.

I plan to test on main net.    You must fear my 50 times .5 gigahashes per second in lunchbox.

Why? I know of no mining ASIC manufacturer that has said they will mine on the main-net.

In my opinion, if there were no pre-orders, then fine.

I still wouldn't like the idea of a company competing with me, but hey, they didn't take my money upfront. They could decide not to sell the product at all if they choose.

Now once you take FULL payment upfront, I own that product unless you refund my money.

I don't get why people don't understand that pre-payment = contract.

I still applaud ALL vendors who have stated they will only use testnet in a box.

Any other vendor who took full payment pre-order should be held to the same standard.

If not, informed consumers should make the right choice.

People always bring up "free market", well free markets are an epic FAIL without informed consumers.

Replying to agree with op and to bump the thread.. this should be a huge shitstorm.

2 thinks:

1st, we're talking about a road test on PUBLIC roads. As in driving on the highway in a dealer car, right next to dozens of other cars that have been bought by customers. In our analogy, public roads are the main-net. Test tracks are the test-net. How are you saying that driving an untested car on public roads not like mining an untested miner on the main-net? This is assuming the design has been proven solid, but that specific unit has not.

2nd, car dealerships totally loan out dealer cars for employee use. Not every employee gets one, but 6 out of the 20 employees at the dealership were given dealership cars for their personal use. They would drive them for a few weeks to a month or two, get a new one, and then sell the one they were driving. The goal was to "become familiar with all the cars on our lot", but that's not the REAL reason they did it. Anyways, I don't know how this helps our analogy, but whatevs.

Your example is actually pretty close to a test track. It might be on public roads, but it was a pre-determined route used by the dealership to determine if there are issues that need correcting. This is like regular TestNet and a closed track is like TestNet-in-a-Box. Unfortunately the public TestNet track is populated mostly with by the State Steam Powered Auto club (developers), and are far more concerned with making sure that every little part of the track is known so they can test very precise details in their contraptions (test code build/hardware.) They can get all pissy if we bring in our dragsters and make a lot of noise and tear up their track, so they have built us our own Private Track called TestNet-in-a-Box.

So we have TestNet-in-a-Box where we can take the minirig-sc, give it the beans, and make sure the wheels stay on.

Testing on MainNet is more like the dealer letting all his employees USE the new cars for a week instead of their own, and every car he sold as "NEW!" was actually used for the carpool (or hauling home drunks) for a week.

Here's Avalon-ASIC's statement:

Source: http://www.reddit.com/r/Bitcoin/comments/10vmxa/avalon_asic_ama/c6h2pu2

IIRC, they never actually said they would mine on the main-net. They never actually said they would be using a test-net, either, at least before that 2012/10/2 post.

This is great to hear and kudos to BFL.

They take a lot of S on here, but they deserve credit on a very ethical decision in my opinion.

Sounds like the hardware folks are in unison on this which is great for the community.

They contradicted themselves?

...and it looks like BFL has committed to not test on mainnet:

https://forums.butterflylabs.com/showthread.php/52-ASIC-Pre-Shipment-Testing-Policy


Further down, there's something about testnet-in-a-box somehow not being fully up to the task of testing a TH/s-class device (bitcoind can't handle the throughput), but that they'd figure something out.  Given that the 20 or so people who've ordered Mini Rigs could potentially have such a device banging away at their bitcoind instances when they arrive, what exactly does this mean for them?

IIRC, he's hosting (aka controlling) a bunch of people's FPGA Single and MRs, while they're waiting for the trade in program. He pays them their BTC, minus a cut to host and power them, but he can control where they mine at.

If Butterfly labs were to be using the rigs to mine, would we assume that they would mine on EclipseMC? Thinking against that, their total hashrate hasn't really changed since ozcoin sent a lot of their miners to Eclipse (it went from 1.3TH/s to 1.8TH/s)

Or then again, if BSL_josh/Inuba was smart, he's figure a way around it like...

*using another pool
*Mine on his own
*change the sites programming to not reflect test rigs.

I did happen the see Inuba was mining at 550GH/s a few weeks ago, does he have a mining farm that big?

hmmm

Conspiracy theorists, DISCUSS!

0.o

Unless they turn over the mined BTC to the customer by using an online wallet, a customer supplied BTC address of choice or a printed wallet in the shipment. Heck, even a branded USB stick with the wallet.dat on it could do...

Frankly I consider machines running for a burn-in of 24h in a test net a pure waste of electricity and an environmental cost we can live without.

this would more be an issues for pools and mining sofware devs to address - not the manufacturers of ASIC - in fact I am expecting some of the less well maintained pools to fall over when someone tries to mine on them with ASIC.
stratum and GBT were developed specifically because of the expected extra loads from currently FPGA minirigs and upcoming ASIC hardware on pools.

one ASIC manufacturer is sending software devs devices to code on so there is actually working software when his product gets delivered.
I'm sure pool operators would be happy to be sent devices to test against the pools before release too All we have to work with is "there are lots of high hashrate machines coming - no real specs yet - but get ready".
I am also sure mining software devs will test against prepared pools - as ckolivas is with slush's pool and btcguild as he develops stratum support for cgminer. This makes sense. not so much manufacturers who can indeed do burn on testing on testnet.

I worked at a car dealership several years ago - and while I have no doubt that the manufacturer tested a few sample models on a test track, not every single car manufactured was run through a test track run. However, almost every single car WAS tested (by the dealership) on public roads, before it was sold to the customer. We would slap some special dealer plates on the car, and go for a carefully chosen 8 mile loop, one that included hills, flats, stop lights, highway stretches, etc. At the end, we'd fill up the gas tank, and return. Yes, these were cars directly off the trucks, and had not gone through a State Inspection yet.

i certainly hope there will be some testing on mainnet. i don't want to get a product from any company, if it hasn't been proven on real live pools.

how will pools respond to some much additional hashing power? will a sustained 10-20x increasing in hashing power have an adverse effect on pools? i don't want to see pools going down after ASICs come out.

will the product work with stratum? how about getblocktemplate?

i think it's unreasonable to think/ask that that no testing occurs on main net. bfl, btcfga, and avalon should all have engineers/developers performing functional and stress testing on REAL world mining scenarios before they send their products out to customers. i sure as hell don't want to get an ASIC miner only to find out that there is a critical bug that cripples my ability to mine...

i don't know how big each company's development team is , but I'd say it'd be reasonable to have maybe 3-5 of their test machines running on main net during development. of course, once they've finalized their development and testing, there is little need to keep those machines attached to main net. then before shipping, customer machines can be "burned-in" on a test net .

Dear BFL,

If you test on main net, please forward the earned BTC to my BTC address.

Sincerely,
Someone who is not very concerned about thus issue.

p.s.  Please test my order first and for as long as you can.

well testnet is for um testing Bitcoin stuff
all aspects
with testnet in a box you can even choose difficulty and stuff - surely an advantage to high hashrate devices

um I like bad analogies - someone raised cars before
so if my car manufacturer is testing my new car I would prefer he tested it on a test track before it went onto public roads (I know no-one will get killed if an asics brakes fail - but hopefully you get my drift )

the difference between testnet and mainnet in the Bitcoin software is this flag
-testnet           Use the test network
cheers
Graet

So BFL will be using a test-net. Cablepair will be using a test-net. Are any manufacturers using the main-net?

I thought that was more for software/protocol changes?

Disregard any previous debate going on in this thread: does the testing of any new mining hardware (GPU, FPGA, or ASIC) require the use of a test-net?

I'm going to assume the answer is yes. When support for a hardware is written into a miner program, be it an openCL kernel, or a FPGA bitstream, I'm assuming it gets tested on a test-net first.

I'm going to also assume that BFL or anyone else adding support for BFL hardware will also do this, while configuring the firmware/mining software. But once that's done, why does a test-net need to be used for every single piece of hardware?

I'm asking this seriously, because I greatly respect your opinion, Graet. Half of me agrees with you, but I just like a good argument.

ASIC Pre-Shipment Testing Policy

Update:


https://forums.butterflylabs.com/showthread.php/52-ASIC-Pre-Shipment-Testing-Policy

but yours made better sense, my grammer sucks.

testnet and testnet in a box were designed for testing before bringing to main net - that's what should be used, no questions asked.No discussion entered into.

If it must be mainnet (for whatever imagined reason)
well... if only I could supply login details for a worker - they could test my asic against my account on a pool I choose- this issue would dissolve and could even be seen as a gesture of good will towards people that have pre-ordered and waited.
Otherwise it will be profitable mining at lower difficulty with asic than anyone else is able to, will push difficulty up before asics start getting delivered (its not going down after they start getting delivered) and affect every miner not just the ones ordering asic.

As a bonus miners would know when their ASIC is getting close - they could see hashes on that worker

radical idea!!! no way it could get adopted - too logical for Bitcoin

any asic maunfs planning to test on mainnet want to balls up and do this?

goodwill is priceless
my thought
Graet

I typed too many words Kaliecious beat me

I do think the asic suppliers should do a burn in on the main net. but what should be done is contact the purchaser of the unit that would be burned in and get a worker login for thier said pool, and burn in with the purchasers login information. inorder to keep burnins being done hush hush if a purchaser was to announce that burn ins are being done the go to last of shipping list.

I'm not "buying into" anything. I honestly just think you're making mountains out of molehills. I told you here:


They could burn in 15TH/s of ASICs (with our numbers of a 100TH/s network, that's 15%) for 24 hours, and the increase would only be 1%, and even that's only for 1 difficulty period! After 2 weeks, when that 15% increase is hashing 24/7 in the hands of customers, it wouldn't even matter!

And this is at a 5X network size. I've seen estimates anywhere from 5X to 100X. The larger the network, the less their burn ins affect it.

So with the worst case scenario, the smallest projected network increase and a high amount of ASICs being sold per week, we're talking about a 1% difficulty change for one difficulty period (~14 days). That's going to cost miners a fraction of a BTC, over the entire grand scheme of things. Again, why is this such a big deal?

I would say it is an issue of disclosure. If say you buy an ASIC being CLEARLY told before purchase that it would be used for 72 hours to main-net mine and the price you pay takes that into account that is 100% acceptable, as those are the conditions of the sale.

Other then that.. 100% wrong unless 95%+ of all BTC earned during testing was given to the buyer, in which case they are almost doing you a favor.

You mean, if they ran all the ordered rigs for about 5 minutes total? 

@crazyates

bro it's not just your units or the ones shipped with your batch.

They will keep testing and testing can't you see that?

You'll be competing against the next batch of orders before they are even shipped.

It's unbelievable that people buy into this stuff.

And yes, those 1000 coins matter, ANY coins matter in a finite supply.

Well I suppose the apologists won't ever quit on this one.

Thank "bob" cablepair (bASIC) said it, and I think ngzhang (Avalon) said it too, though I will have to go back and look to be sure.


I hope to see the opposite happen, and will "vote with my monies" to help make it so.

No, that's my point. There are almost ZERO "lasting effects" after ~2 weeks, and any that actually are there are minuscule. I wait 1 Difficulty change after I get my ASIC, and whatever burn-in they did no longer affects my rewards. I'm getting the same amount of Bitcoins as if they had used a test-net.

As far as the whole "there are only 21 million coins, and their stealing the coins we should be mining" argument: Lets say they mine 1,000 BTC in their burn-in. All this does is push forward the reward half by a matter of about 3 hours. Any "lasting effect" of this reward half coming 3 hours sooner is, again, minuscule.

The Bitcoin network was designed from it's very conception(with increasing difficulty, reward halves, and TX fees) to eventually level out.

Did they ever say they wouldn't test on main net? If they didn't, then it's not fair for us to add new requirements after money has changed hands.

Furthermore, making the coming ASIC-fueled difficulty increase smoother is a good thing for the bitcoin economy.

Chalk it up as a lesson learned. When the next big thing is announced, demand either testnet-only testing or lower prices up front. I suspect that when otherwise identical manufacturers compete, the ones who do main net testing will out-compete those who don't.

I'm not saying anything   Maybe Inaba has a quarter of a million dollars worth of mini-rigs, who knows?
I'm just stating the facts, which are highly suspicious.

ELLIPSIS ...

...aren't there still a lot of people waiting on their BFL FPGAs to be shipped to them?  Are you saying there is 500Gh/s of them undergoing "burn in" while customers still sit empty handed?

We can type all day in this thread, but really we can't stop these companies from cheating us by mining BTC for themselves with OUR units, already paid for with OUR funds, if they choose to be so dishonorable as to do so.

We can only hope that the honest maker(s) ship(s) first, and vote with our monies when placing additional orders.  And/or try to cancel orders with any companies who refuse to pledge not to mine main net with customer units.

However, they are making it so anyone that invested huge amounts of money ensuring their ASICs ship first will make less money than originally planned. I find that funny, screwing the people that made their profits possible.

I LOLed


+1

This is a very good analogy. I was trying to come up with one myself, but blanked.

Here here.

Imagine if someone sold you all tickets to pick berries at the best berry patch in the state, but then when berry season came around, the ticket sellers went in the first week and picked all the best easiest berries for themselves, only letting all of you ticket holders in beginning the second week to fight over the few remaining small hard-to-reach berries in the middle of briars.

I agree there HAS to be SOME testing on main net. you cannot sell a product that has not been tested in a real world environment. even if test net mirrors main net...it's still not exactly the same. how will the hardware react with live pools? how will it react under different load situations? etc...it'd be negligent to do zero testing on main net. Not saying they need to run each unit for hours at a time on main net, but zero testing is absolutely unacceptable.

First off, you're delusional that this money will flow to the consumer.

If these companies publish the amount of coins mined and reflect it in their pricing then fine.

Also, testnet in a box is as much "real world" as the main net. It's NO different for testing purposes, thus why it was developed.

@crazyates

I get you, but here's the point, it's not just your ASIC and since difficulty only adjust every 2016 blocks, well the effect could be lasting.

Nevermind that this testing will be ongoing, so we're talking a fairly permanent effect on the difficulty.

I see no one answered that the coins will also be gone. As a miner, you'll never be able to mine those coins.

We are all mining based on this same concept, there are 21 million coins available and nearly ten million are gone.

You don't need to have that kind of money - just write some impressive numbers on a napkin and people here will throw millions of dollars at you.

No, it is not "as simple as that".

99% of Bitcoin users do not have the kind of money to design an ASIC, so that is simply not an option.

BFL. People got pissed when their "burnin" account kept popping up on eclipsemc, now inaba has control of 500Gh/s of FPGAs  

Frankly, I'll be stunned if the ASICs don't get a bit of live testing. If they tested every unit for 7 days they'd make a profit on some of the early ones without even factoring in the sale price  

There are some offers from several manufacturers... Take them and deal with consequences, or design your own chip. It's simple as that.

Test net in a box seems like the best way to do ASIC testing... who is testing on main net?

ASIC chips are expensive i saw hard copys going for 7k...
its only going to get worse before it gets better at least asic gives the simple man a chance

I agree with OP 110%

I don't care what anyone says.

It is wrong for ASIC manufacturers to FREE ROLL with OUR MONEY with a product they would not have been able to make if it wasn't for our 100% paid in full pre-order, when test net in a box is sufficient enough for testing.

Even if it's for only 24 hours, it is unacceptable.

I will not be ordering from any ASIC manufacturer that tests on the main net, and I hope most of the community adopts this stance.

I under stand the ethical argument about WHY you don't want companies to mine on the main net, but has anyone actually done the math on how much this is actually costing customers?


Hypothetical situation:
The date is 2013/1/1, and the network is 100TH/s.

I order a 60GH/s (or a 54GH/s, lets not play favorites ) unit on 2012/12/1, and it's expected to be delivered by 2013/1/15 (~6 weeks).

This ASIC is going to start mining eventually, whether it's by me, or the manufacturer.

60GH/s is ~0.06% of the 100TH/s network.

Lets say they mine for 24 hours, on the main net, and by the time they ship it to me, we have a new difficulty.

They're mining with 0.06% of the network for 1/14 of the difficulty period, or 7%. (I know 2016 blocks isn't going to take exactly 14 days, but shh!)

Their testing on the main net increased the network rate by 0.0042%. I hardly consider that worth whining and complaining about!

Also, by the time the next difficulty comes after I get my unit, their testing 4 weeks before has absolutely ZERO affect on the future difficulty. My mining has changed the difficulty, not theirs.

Now they do this with 100 orders (or the equivalent of 6TH/s). They mine with 6% of the network for 1/14 of the period, for an increase of 0.42% increase in difficulty. This might actually be noticeable, but still, hardly worth freaking out about.

It's just not FAIR!

There is testnet in a box plain and simple.

Also, if it's built into the price, then everyone should have been made aware prior to pre-order.

I repeat this is NOT a normal situation.

This is direct competition for a finite resource which is nearly half gone already.

This is direct competition for a finite resource which is distributed in relation to network hashrate.

There is almost no example of a similar situation where this would be common business practice.

The manufacturer's development cost is not our problem. It's already insane that people plunked money down nearly 4 months ago as it is.

That's fine, they took that risk.

They DON'T deserve to be competed against by the same entity or entities.

It's not just BFL, I urge Avalon ASIC, Reclaimer and others to pledge the same.

bASIC has already pledge to test on testnet in a box.

ASICMINER is a mining bond and they were straight up from the beginning about their intentions and have NOT taken any orders for hardware.

Let all be clear.  Estimating current BFL units sold the profit from 30 days of sustained QC/burn-in mining would equate to 80-150 USD discount per unit.


The other ASIC vendors have claimed and will survive without profiting from main-net.  While they may perform limited testing with main-net, they will not be performing 24 hour burn-in sessions with each unit produced.


There's no excuse.  There's no good reason (other than easy money) that they would tell us they would not do this, but then change their mind and consider it an issue of little importance to the mining community.


The BFL unit you have paid for is effectively 80-150 USD more expensive than you realize.  And it's not even the best $/GH.

I won't mind if each manufacturer only puts 1 TH only at a time for testing thier units.

I would prefer they use a test box,since testnet dosen't want to be overrun to skew any testing thier doing,but who am I  

What if they built it into the price? What if the only way they could actually afford to develop a custom ASIC chip was with a certain period of pre-mining?

It's not like anyone is complaining that it isn't possible for them to actually build a custom ASIC chip with what they were charging...

I can't believe people are actually advocating on behalf of companies selling us hardware and directly competing with us.

This is not like nearly any other ecosystem where it makes sense to allow a company to compete with you.

THERE ARE ONLY 21 MILLION BITCOINS!!!

Whatever they mine WE never get a chance at, EVER.

It will aartificially inflate the difficulty and it WILL NOT STOP due to ongoing testing.

This community really is insane sometimes, falling for obvious scams, pre-funding vaporware, etc.

Now you think companies "deserve" to profit with your pre-ordered hardware that you funded at 0% interest ?

And artificially inflate the difficulty for the foreseeable future as they "test" these devices ?

One week of ASIC mining in the early days can be HUGE for your profit margin at lowered difficulty.

Come on people, stop and think about this for a moment. You put out your money months in advance, with limited recourse at 0% interest.

YOU DESERVE THE RIGHT TO PROFIT, NOT THE HARDWARE COMPANY!

These should be tested on testnet or testnet in a box only.

Please demand that all ASIC vendors do this.