[ANN] Noirbits Update required Improved algos !!!!

[OnlyC]

7000% wtf 

[1715618833]

1715618833

[1715618833]

1715618833

[1715618833]

1715618833

[barwizi]

7000% wtf 

huh?

[s7ereotype]

To many rejects 30%

Now 10%

[stbgefltc]

Yeah, I think I know where it comes from... there's in source a fix to avoid 51% attacks that could change difficulty at will... Thing is it looks back over the last 4 retargets to adjust hashrate. With yesterday's spike, and the 4x rewind to recalc diff, difficulty went way lower than it should have... It's now adjusting with the new retargets that went very quickly, a couple minutes max. between each.

Final word, we really really need a new diff algo, or this is gonna kill the coin...

[stbgefltc]

To many rejects 30%

Now 10%

Yeah, that's to be expected... you can decrease scantime and expiry on cgminer, but hashrate is so off the charts compared to difficulty that it's bound to happen.

[s7ereotype]

WTF with this pool http://nrb.miners-united.com ?
Power 800 khs instead of 3000 mhs shows how other pools.

[Oldminer]

WTF with this pool http://nrb.miners-united.com ?
Power 800 khs instead of 3000 mhs shows how other pools.

Yea found heaps of blocks but not showing and hash rate is 1/4 of what it should be..

[barwizi]

Yeah, I think I know where it comes from... there's in source a fix to avoid 51% attacks that could change difficulty at will... Thing is it looks back over the last 4 retargets to adjust hashrate. With yesterday's spike, and the 4x rewind to recalc diff, difficulty went way lower than it should have... It's now adjusting with the new retargets that went very quickly, a couple minutes max. between each.

Final word, we really really need a new diff algo, or this is gonna kill the coin...

lets get to it tonight

[stbgefltc]

WTF with this pool http://nrb.miners-united.com ?
Power 800 khs instead of 3000 mhs shows how other pools.

Yea found heaps of blocks but not showing and hash rate is 1/4 of what it should be..

Yeah I need to check my hashrate calc... but don't worry, you are getting paid for shares, not hashrate. I think the hashrate calc had trouble keeping up with the rate of shares being pushed.

I'm not home, so I can't check, but my hashrate is half of what it should be. Do your miners report any slowdowns, or is everything ok ?

Hashrate is returning to normal with the diff increase... but anyhow, you are getting paid for shares, not hashrate.

[TheRickTM]

Yeah, I think I know where it comes from... there's in source a fix to avoid 51% attacks that could change difficulty at will... Thing is it looks back over the last 4 retargets to adjust hashrate. With yesterday's spike, and the 4x rewind to recalc diff, difficulty went way lower than it should have... It's now adjusting with the new retargets that went very quickly, a couple minutes max. between each.

Final word, we really really need a new diff algo, or this is gonna kill the coin...

lets get to it tonight

Let me know when we got a new windows client we need to push and I'll get it up on the website as well.

[TheRickTM]

What would also be good is if we could detect if say 5 blocks were found too quickly in a row and then do a retarget then. Once we make the changes to the difficulty to allow it to lower faster, it just means that more people will jump back on until it sky rockets again and then comes back down for those who are leaving their miners on it for that period of time keeping the coin going. Just another thing for us to consider.

[oatmo]

What would also be good is if we could detect if say 5 blocks were found too quickly in a row and then do a retarget then. Once we make the changes to the difficulty to allow it to lower faster, it just means that more people will jump back on until it sky rockets again and then comes back down for those who are leaving their miners on it for that period of time keeping the coin going. Just another thing for us to consider.

Yes. We are seeing a real seesaw right now. We have to figure out the right ballast for the algorithm. They should design a test simulator into the open source along with a test suite which does stuff like this, so the algorithm can be refined before the coin is launched.

Having said that, I went to bed last night adding hashes to the network, because I thought that in another 20 blocks the hashrate would adjust down. It did, but went too low and we got 500 blocks over 6 hours, but I have almost no coin this morning, so I think that large miners jumped on it as soon as the difficulty went low, taking all the coin. We definitely need it to react much faster to adds and drops.

Instead of using the average over the last 4 retargets, how about an exponential weighted average, like 50%, 30%, 12%, 8%. Unlimited drops (how about only when we hit the 4 hour limit, but not for a regular 60 block retarget).

[cyberkiller]

Nice to see you on www.DigitalCurrencyTalk.com

[stbgefltc]

What would also be good is if we could detect if say 5 blocks were found too quickly in a row and then do a retarget then. Once we make the changes to the difficulty to allow it to lower faster, it just means that more people will jump back on until it sky rockets again and then comes back down for those who are leaving their miners on it for that period of time keeping the coin going. Just another thing for us to consider.

Yes. We are seeing a real seesaw right now. We have to figure out the right ballast for the algorithm. They should design a test simulator into the open source along with a test suite which does stuff like this, so the algorithm can be refined before the coin is launched.

Having said that, I went to bed last night adding hashes to the network, because I thought that in another 20 blocks the hashrate would adjust down. It did, but went too low and we got 500 blocks over 6 hours, but I have almost no coin this morning, so I think that large miners jumped on it as soon as the difficulty went low, taking all the coin. We definitely need it to react much faster to adds and drops.

Instead of using the average over the last 4 retargets, how about an exponential weighted average, like 50%, 30%, 12%, 8%. Unlimited drops (how about only when we hit the 4 hour limit, but not for a regular 60 block retarget).

The issue with implementing a test suite is... well the source code. It needs heavy refactoring to achieve simulation of hashrate peaks without the hardware.

Other than that, I second the unlimited drop only if 4 hour interval is reached. I don't know where the 4 times rewind comes from (its not in Litecoin source), but we've seen its disastrous effects : it's a shitload of crap. We need to get rid of it ASAP, but we can't push it just like that. We can conditionally remove it starting at block N, where N needs to be decided so it gives enough time for users to update the client. That way we can push the change immediately, and have it effective in about a week.

Edit Also, I think we need to raise min target, it's currently way too low compared to the available hashing power.

[barwizi]

as i said, this is a community thing, if the active people are in agreement then let's do it.

[oatmo]

What would also be good is if we could detect if say 5 blocks were found too quickly in a row and then do a retarget then. Once we make the changes to the difficulty to allow it to lower faster, it just means that more people will jump back on until it sky rockets again and then comes back down for those who are leaving their miners on it for that period of time keeping the coin going. Just another thing for us to consider.

Yes. We are seeing a real seesaw right now. We have to figure out the right ballast for the algorithm. They should design a test simulator into the open source along with a test suite which does stuff like this, so the algorithm can be refined before the coin is launched.

Having said that, I went to bed last night adding hashes to the network, because I thought that in another 20 blocks the hashrate would adjust down. It did, but went too low and we got 500 blocks over 6 hours, but I have almost no coin this morning, so I think that large miners jumped on it as soon as the difficulty went low, taking all the coin. We definitely need it to react much faster to adds and drops.

Instead of using the average over the last 4 retargets, how about an exponential weighted average, like 50%, 30%, 12%, 8%. Unlimited drops (how about only when we hit the 4 hour limit, but not for a regular 60 block retarget).

The issue with implementing a test suite is... well the source code. It needs heavy refactoring to achieve simulation of hashrate peaks without the hardware.

Other than that, I second the unlimited drop only if 4 hour interval is reached. I don't know where the 4 times rewind comes from (its not in Litecoin source), but we've seen its disastrous effects : it's a shitload of crap. We need to get rid of it ASAP, but we can't push it just like that. We can conditionally remove it starting at block N, where N needs to be decided so it gives enough time for users to update the client. That way we can push the change immediately, and have it effective in about a week.

Edit Also, I think we need to raise min target, it's currently way too low compared to the available hashing power.

I think the spikes up will be taken care of with 30 block retargets instead of 60, with a better weighting over the last 4 retargets. The real problem is when the hashing power drops off. This morning my client says that network has 180MH/s up from 5MH/s last night. Is that all from coinchoose reporting a rediculously low difficulty and everyone jumping on? If so, then I change my vote to remove us from coinchoose for a week to get the client updated before we go back on. I'm now thinking that even a max of a 3 hour retarget would be better than 4 to handle fluctuations like this (and I still think an unlimited drop in the case of a time forced retarget only).

Now it's reporting 130MH/s, so maybe we will be spared the huge drop which causes a 4 day wait for the retarget, followed by the difficulty dropping way too low.

[JDDev]

Yes, the problems you're describing are resulting from the listing on coinchoose

EDIT:
Essentially what happens is as soon as the difficulty adjusts down, the profitability of the coin on coinchoose jumps by a large amount. This makes it 500% profitable to mine, and a large amount of miners jump on the coin. Then once the difficulty adjusts to the new hashrate, then those people stop mining noirbits making the hashrate low and the difficulty extremely high, making it take hours to find a block. This is bad for miners and for users of the coin as transfers also take hours.

When its listed on coinchoose there is no incentive for miners to mine through high difficulty periods as they know once the high difficulty period is over, other miners will jump on the coin taking away some of the profit, and making it high difficulty again. Also, this causes crashes in price (The highest offer is down to 0.00006 right now) because miners who only are mining the coin at low difficulty, and not because they believe that the coin is good will just dump when they get their NRB.

[sal002]

As I mentioned to JDDev - if you all come to a consensus to remove from CoinChoose, I will remove it.  I think there is a different issue if the profitability is jumping that high, though, as CoinWarz would also pick it up.

[oatmo]

Yes, the problems you're describing are resulting from the listing on coinchoose

EDIT:
Essentially what happens is as soon as the difficulty adjusts down, the profitability of the coin on coinchoose jumps by a large amount. This makes it 500% profitable to mine, and a large amount of miners jump on the coin. Then once the difficulty adjusts to the new hashrate, then those people stop mining noirbits making the hashrate low and the difficulty extremely high, making it take hours to find a block. This is bad for miners and for users of the coin as transfers also take hours.

When its listed on coinchoose there is no incentive for miners to mine through high difficulty periods as they know once the high difficulty period is over, other miners will jump on the coin taking away some of the profit, and making it high difficulty again. Also, this causes crashes in price (The highest offer is down to 0.00006 right now) because miners who only are mining the coin at low difficulty, and not because they believe that the coin is good will just dump when they get their NRB.

Then lets ask them to delist us for a week while we get this stuff fixed unless there is violent opposition.

[JDDev]

If we could get a few more posts in this thread asking to delist for a while, then it would happen