Author

Topic: Here's why no one was GPU-mining Litecoin from the start (Read 12680 times)

donator
Activity: 1654
Merit: 1351
Creator of Litecoin. Cryptocurrency enthusiast.
There was no proof that the Litecoin chain bloating was done by a Solidcoin supporter. I only suspected it becuase they were so anti-Litecoin. And I don't remember of any Solidcoin chain bloating attacks. But there were some attacks on Solidcoin that caused the clients to crash. This was before Litecoin launched IIRC.
legendary
Activity: 1484
Merit: 1005
As I recall

2a) BCX didn't really have anything to do with it, rather it was lolcust (who is possibly also artforz) who implemented scrypt for Tenebrix.  Tenebrix failed, but coblee's LiteCoin succeeded.
2b) Reaper was originally made to do the SC2 kernel and SC2 mining; mtrlt was pressed by the SC2 community to make a scrypt miner because of artforz claim that it didn't work.  He claimed to have it working about a month before it was released; I pressed him to release the open source code for which there was a bounty, and eventually he did so.  RealSolid/CoinHunter had nothing to do with this aside from community pressure AFAIK, though.
3) Not necessarily true; people lost confidence in LTC after the GPU miner was released.  There were other major events as well, namely:
- Chain bloating exploit first for LTC implemented by the SC2 guys, then SC2 chain bloating implemented by the LTC guys
- A SC2 exploit which let anyone sign the trust node blocks, which allowed someone to run off with a lot of SC2 that was intended originally as an indirect tax on the users
3b) Artforz' pride was strongly hit by the entire LTC experience and his own naivety and overconfidence in his code.  Before the GPU miner came out, artforz spent a lot of time and posts defending his implementation.  There are technical threads where he later admits he was wrong well before mtrlt's miner came out.  I only understood them in retrospect, but the algorithm mtrlt uses derives from the information in these threads.
legendary
Activity: 2128
Merit: 1073
OK, I wanted to sumarize what happened around the time that the Litecoin was launched. I want to do this while I still have some recollection of the events and discussions that occurred on the IRC.

1) There was a period of alt-chain wars where various groups launched a slight or extensive modification of Bitcoin and the groups that tried to take those down.

2) There were two prominent competing groups in the history of Litecoin:

2a) ArtForz and BitcoinExpress group. Their first main contribution was the fix for the off-by-one error in the Bitcoin difficulty calculation and exploits for the coins that could be attacked with the so called "time travel" attack that relied on this bug (among other faults). Their second main contribution was a selection of Colin Percival's scrypt() algorithm and its parameters, which got included in Litecoin through Tenebrix and Fairbrix. The scrypt() parameters were meant to be make GPU mining impossible, but they turned out to just within their outer limits of possibility. But that discovery was made by the other group and the this group GPU-mined the scrypt() coins only after the other group published their code. 

2b) RealSolid/CoinHunter group. Their main contribution was funding the developer "mtrlt" to develop "reaper": a GPU miner for both coins based on scrypt() (as an attack) and SolidCoin (as a defense). The GPU implementation of scrypt() from the "reaper" was later added nearly verbatim into the popular cgminer Bitcoin mining software.

3) Neither group succeeded at causing massive damage to the coins of the other group.

3a) SolidCoin adoption stalled due to erratic behavior of its lead developer.

3b) ArtForz got preoccupied by maintenance of his Bitcoin mining operation with FPGAs and metal-programmable structured ASICs. First, he stopped getting involved in the scrypt()-based coins he co-created and then he ceased to publicly participate in discussions anywhere.

Please feel free to quote me and make the corrections that you think are required to make this historical account reasonably fair and accurate.

Thanks.
-ck
legendary
Activity: 4088
Merit: 1631
Ruu \o/
The scrypt opencl kernel in cgminer was copied almost verbatim from reaper.
efx
sr. member
Activity: 378
Merit: 250
"ckolivas finally included scrypt() mining into his software?"


Are we talking about cgminer here...? I think you'll find reaper was the first, ckolivas used the kernel from it for cgminer scrypt support some time later.  Apparently, realsolid decided that publicly releasing a GPU miner and 'exposing the scam' was going to destroy litecoin, so he pressured the creator (mtrl or something like that) to release it.

Although it sounds like a rather illogical thought processes, I've come to expect nothing less from the MicroScammer.  
legendary
Activity: 2128
Merit: 1073
So in the interest of history: who was working on OpenCL/CUDA mining for scrypt() and/or solidcoin v2?

My recollection is that there were two broad camps centered around very vocal persons: BitcoinExpress and RealSolid. Both were threatening to "rape" the coin of the other camp. Both offered some sort of bounty for development of the GPU mining software. The bounties were both defensive (mine "our" coin) and offensive (mine "their" coin).

How many people have gotten close to collecting those bounties? How many closed-source implementations existed before ckolivas finally included scrypt() mining into his software?


efx
sr. member
Activity: 378
Merit: 250
Oh no, this thread shall not be allowed to drop down until you have all had time to read it. I am so tired of hearing the same story recycled over and over. Here and on the btc-e chat, the delusional conspiracy is still rampant.

 If you guys actually spent a few minutes considering the implications, you would have already noticed that the entire 'artforz gpu premine' fabrication falls apart.

Did you children really listen to coinhunter and proceed to chase away one of the few seemingly decent (just my personal opinion) members of this community?


If we look at the issue from a basic level and assume artforz had some coin hoarding/patient ltc 'dumping' affliction ....Why would he ever admit that he had any sort of working scrypt kernel hashing on a GPU if he's just sitting back scraping up coins? I know most of you haven't even bothered looking at the scrypt kernel...The sequence of events as they have been told from the beginning by the OP (and others) is highly probable.


I guess it's also "inconceivable" that the data is correct (and verifiable, imagine!) and drawing conclusions from said data is the only logical course of action. I suppose most of you would rather rely on the emotionally driven panic you normally see around this place.

No offense to SaltySpitoon, one guy cannot stop a free-falling a380.


+1 tacotime
donator
Activity: 1654
Merit: 1351
Creator of Litecoin. Cryptocurrency enthusiast.
Here's the math:

On 1/13/2012
Litecoin difficulty: 0.65
Litecoin price: $0.02/ltc (0.003 btc/ltc)
Bitcoin difficulty: 1,250,757.74
Bitcoin block reward: 50 btc
Bitcoin price: $6.5/btc

GPU mines about 1000x the speed on Bitcoin compared to Litecoin

A 5970 would mine bitcoins around 750 mhash/s and get about 0.6 btc a day at that time, which is about $4.23
If mining litecoins at 750 khash/s, it would make 1160 ltc a day, which is about $23.20

Ok, so you'd make 5.5 times more. There'd be no risk. Just sell all the litecoins you'd make for bitcoins and if that pushes the price down to a point where it's no longer profitable to mine litecoins, then switch back to bitcoin mining. And that price would be 0.0005 btc/ltc. (I don't remember litecoin price ever being that low)
legendary
Activity: 1484
Merit: 1005
My guess is artforz couldn't figure out how to code a GPU miner effectively.  The key to reaper's code is to effectively halve the size of the lookup table (lookup_gap = 2) and read sequentially in order to achieve the fastest hash rate.  If you use the entire lookup table in RAM (lookup_gap = 1) you lose about 40% performance, which was much closer to what artforz was originally quoting for GPUs when the chain opened.

There are threads later where he admits he was wrong after Litecoin's release.  So, there's no need for conspiracy theories or even coblee to sit around and try and prove there was no premine.
donator
Activity: 1654
Merit: 1351
Creator of Litecoin. Cryptocurrency enthusiast.
Thanks for all the numbers. Clearly artforz would never do anything like only using a portion of his miners on LTC and BTC. Inconceivable.

If you had miners that can make 10x the amount mining litecoins instead of bitcoins, would you not put all your miners on litecoins? It's not like anyone would suspect anything. They'd just think some botnets were mining litecoins with a lot of CPUs.

I will do that math to figure out exact how much more profitable it was.
hero member
Activity: 632
Merit: 500
Booo! Get out of here with your "facts"! You're preventing me of masturbating my paranoia with conspiracy theories.
sr. member
Activity: 966
Merit: 311
Thanks for all the numbers. Clearly artforz would never do anything like only using a portion of his miners on LTC and BTC. Inconceivable.
donator
Activity: 1654
Merit: 1351
Creator of Litecoin. Cryptocurrency enthusiast.
When GPU mining Litecoin became a reality, people kept spreading FUD that ArtForz and/or I have been GPU-mining Litecoin from the start.
Let me put this issue about GPU-mining from the start to rest once and for all.
I will use cumulative difficulty to figure out how much hashpower has been working on a chain since the start.

There's going to be a lot of math here. First read up on this:
https://en.bitcoin.it/wiki/Difficulty#What_network_hash_rate_results_in_a_given_difficulty.3F
https://github.com/litecoin-project/litecoin/wiki/Mining-hardware-comparison

Here are the current state of things:
Code:
Current difficulty: 20.794
Number of hashes to solve a block: DIFFICULTY*pow(2,32) = 89,309,034,556.95
Seconds per block: 2.5 * 60 = 150s
Theoretical network hashrate (in mhash/s): DIFFICULTY*pow(2,32)/pow(10,6)/SECONDS_PER_BLOCK = 595 mhash/s (~2000 average GPUs)

Litecoin was launched on 10/13/2011 03:00:00 at block #3:
http://explorer.litecoin.net/block/dec173dda2735ff11376b68bdfda804cede230c1fa6f1a11765cddfd8edf4398

We can calculate how much hashpower has been put on the chain since the start using cumulative difficulty.
Let's check a recent block 294537 found on 2/12/2012 03:00:00
http://explorer.litecoin.net/block/a065026ba50a71e1d4979e078265dc9ccf15d0b393969cd35ec4c954bf2c22fb
You can see the cumulative difficulty on the block explorer page.

Code:
Cumulative difficulty: 2,421,540.599
Number of hashes: DIFFICULTY*pow(2,32) = 10,400,437,678,641,250
Time since start (in seconds): 2013-02-12 - 2011-10-13 = 488 days * 24*60*60 = 42,163,200 s
Theoretical network hashrate (in mhash/s): DIFFICULTY*pow(2,32)/pow(10,6)/TIME_SINCE_START = 246.67 mhash/s (~1000 average GPUs)

So we average about 1000 GPUs working on the chain. In other words, if you had 246.67 mhash/s pointed at the chain since launch, you'd have found just as many hashes.

Now, here's what you all wanted to know. How much hashing power was pointed at the chain during the first week.
Here's block 14807 found at 10/20/2011 03:00:00:
http://explorer.litecoin.net/block/6fcf032b2edfd3e06ee6cace9ed9b6c219d8dca06fa1f43a47cb1c5b7f87084f

Let's do the same math:

Code:
Cumulative difficulty: 438.193
Number of hashes: DIFFICULTY*pow(2,32) = 1,882,024,604,336
Time since start (in seconds): 7 days * 24*60*60 = 604,800 s
Theoretical network hashrate (in mhash/s): DIFFICULTY*pow(2,32)/pow(10,6)/TIME_SINCE_START = 3.11 mhash/s (~100 average CPUs OR 10 average GPUs)

A month later. Block 31011:
http://explorer.litecoin.net/block/7b08a3bfb5f2a865fc0061f6e3f5b97fa1690c8d357ccd814fd9f55641f83187

Code:
Cumulative difficulty: 5,949.565
Number of hashes: DIFFICULTY*pow(2,32) = 25,553,187,100,426
Time since start (in seconds): 31 days * 24*60*60 = 2,678,400 s
Theoretical network hashrate (in mhash/s): DIFFICULTY*pow(2,32)/pow(10,6)/TIME_SINCE_START = 9.54 mhash/s (~300 average CPUs OR 30 average GPUs)

After 3 months (http://explorer.litecoin.net/block/55d1323fa4d7175953fab43ef97c0ef18577d8f000e494740ccc867d42fe67f5), average hashrate is 18 mhash/s. You can do that math yourself.


Seems like the normal growth of a CPU-only (at the time) coin to me.

ArtForz had 24 5970s. 5970s can do 750 khash/s. If he put those 5970s on mining Litecoin, he would have 18 mhash/s, which is twice the work done on the chain in the first month. Litecoin was put on the exchange pretty quickly and mining litecoins was pretty profitable even with a CPU. If ArtForz had GPU scrypt mining from the start, would he not put those machines on mining Litecoin and make a killing?
Jump to: