Topic: The YACoin Superfun Premine Thread - page 4. (Read 8999 times)

Quote from: bitdwarf on May 09, 2013, 06:23:43 PM

Yet the Windows build gave me 60 KH/s while my Ubuntu build gave me 400 KH/s with my 2700K -- totally different story.

Anyway, who wants to buy some YAC?

400 KH/s?
I'm getting merely 200-400kH/s on my Intel hexacore. That said I cannot use the AVX optimisation as it runs inside Virtualbox. However, I don't see how this is a totally different story -- we have the same with ASICs for Bitcoin, GPUs that arrived for Litecoin etc.

YAC might not be more "pure" than other coins. But at least it is a new concept and not a simple clone such as FTC and CNC. I have my doubts whether it is the right concept. But only time will tell.

I also find it hypocritical to ask for a "fair" coin. What is fair? Fair, so that everybody can profit? Fair, so that "I" can profit enough and in particular more than the other guy that I'm envious of? Fair, so I can dump it onto buyers and cash out?

Is anybody even remotely thinking about the original motivation of a coin? To make digital payments easy and secure. Mining surely was more of a necessary evil (consuming Megawatts and 100k$/day) than a wanted aspect of the network. I could think of a strategy that should limit the required mining power.

Unfair is when I make lots more than everyone else for no greater effort -- which is what I've done repeatedly. Like I said, I got a bunch of YACs. But I figure people not into the crypto p'n'd' scene will fail to see the value of something that has ponzi-scheme like distributions in the first week that it's out.

tacotime

legendary

Activity: 1484

Merit: 1005

This is one of the reasons that the YACoin chain makes no sense itself, because it eventually intends to use values of N in the gigabyte range.

Won't people be mainly mining PoS by the time Scrypt PoW uses more memory than a cheap home computer can hold? At worst it's just a kill switch for PoW to try to make it economically unfeasible to use specialized hardware.

No, because suddenly you have blocks that are worth 100 coins because the difficulty is so low and no one can mine quickly. The difficulty algorithm in the code increases block reward with decreasing difficulty.

jomay

full member

Activity: 167

Merit: 100

Quote from: Come-from-Beyond on May 09, 2013, 06:22:41 PM

Yet the Windows build gave me 60 KH/s while my Ubuntu build gave me 400 KH/s with my 2700K -- totally different story.

Anyway, who wants to buy some YAC?

400 KH/s?
I'm getting merely 200-400kH/s on my Intel hexacore. That said I cannot use the AVX optimisation as it runs inside Virtualbox. However, I don't see how this is a totally different story -- we have the same with ASICs for Bitcoin, GPUs that arrived for Litecoin etc.

YAC might not be more "pure" than other coins. But at least it is a new concept and not a simple clone such as FTC and CNC. I have my doubts whether it is the right concept. But only time will tell.

I also find it hypocritical to ask for a "fair" coin. What is fair? Fair, so that everybody can profit? Fair, so that "I" can profit enough and in particular more than the other guy that I'm envious of? Fair, so I can dump it onto buyers and cash out?

Is anybody even remotely thinking about the original motivation of a coin? To make digital payments easy and secure. Mining surely was more of a necessary evil (consuming Megawatts and 100k$/day) than a wanted aspect of the network. I could think of a strategy that should limit the required mining power.

tacotime

legendary

Activity: 1484

Merit: 1005

I meant scrypt(1024, 1, 1) should be set to something like scrypt(1024, 128, 1)*. Setting the second parameter to "1" means that IT'S NOT GPU RESISTANT. Artforz fucked a lot of guys with his set of parameters.

* scrypt(CPU cost, memory cost, parallelization cost)

N isn't CPU cost, it's CPU and memory cost. And even at that, there's little different between using N or r to increase memory usage past a certain point because you become very quickly memory bandwidth limited. You're welcome to test it out with scrypt-jane if you want to, though. You'll see that when memory consumption is equivalent with either N or R being used to increase it the time to get a hash is about the same past 128 KB or so of memory usage.

TheSwede75

full member

Activity: 224

Merit: 100

Quote from: Come-from-Beyond on May 09, 2013, 06:22:41 PM

Quote from: Come-from-Beyond on May 09, 2013, 06:09:39 PM

You might be thinking of the GPU mining code -- no one knows if Artforz was able to get a GPU miner for Litecoin running after he realized that the TMTO issue with scrypt allowed for faster mining by GPU. It's still a possibility; Artforz mined privately with his GPU miner when he implemented it for the Bitcoin chain, then only gave out the source code later.

I recall an issue related to the chosen parameters (1024, 1, 1). This set of parameters leads to degeneracy of the algo (in mathematical sense), making it's easier to create GPU mining soft.

That wasn't the problem (although the author of this coin is trying to use that to sell you on it).

Regardless of the value of N, the GPU will likely be faster until you hit some value in the hundreds of megabytes that requires all of the 8-16 GB of RAM available on a modern computer. The reason for this is because you can adjust the amount of memory usage in exchange for number of ALU cycles (this is exactly what the "lookup_gap" variable does in reaper/cgminer). But at a value of N large enough to prohibit faster GPU mining, a single hash is measured in seconds to minutes. This makes it impossible to use for a blockchain hash and was detailed in the memcoin thread I created a long time ago. YACoin is actually more or less exactly what I specified in the original memcoin thread. This is one of the reasons that the YACoin chain makes no sense itself, because it eventually intends to use values of N in the gigabyte range.

I meant scrypt(1024, 1, 1) should be set to something like scrypt(1024, 128, 1)*. Setting the second parameter to "1" means that IT'S NOT GPU RESISTANT. Artforz fucked a lot of guys with his set of parameters.

* scrypt(CPU cost, memory cost, parallelization cost)

Sad part being that he knew this VERY well as it was pointed out early in development. Then again, why fix something that you can profit off?

bitdwarf

sr. member

Activity: 406

Merit: 250

The cryptocoin watcher

This is one of the reasons that the YACoin chain makes no sense itself, because it eventually intends to use values of N in the gigabyte range.

Won't people be mainly mining PoS by the time Scrypt PoW uses more memory than a cheap home computer can hold? At worst it's just a kill switch for PoW to try to make it economically unfeasible to use specialized hardware.

Come-from-Beyond

legendary

Activity: 2142

Merit: 1010

Newbie

Quote from: Come-from-Beyond on May 09, 2013, 06:09:39 PM

Quote from: WindMaster on May 09, 2013, 06:18:12 PM

You might be thinking of the GPU mining code -- no one knows if Artforz was able to get a GPU miner for Litecoin running after he realized that the TMTO issue with scrypt allowed for faster mining by GPU. It's still a possibility; Artforz mined privately with his GPU miner when he implemented it for the Bitcoin chain, then only gave out the source code later.

I recall an issue related to the chosen parameters (1024, 1, 1). This set of parameters leads to degeneracy of the algo (in mathematical sense), making it's easier to create GPU mining soft.

That wasn't the problem (although the author of this coin is trying to use that to sell you on it).

Regardless of the value of N, the GPU will likely be faster until you hit some value in the hundreds of megabytes that requires all of the 8-16 GB of RAM available on a modern computer. The reason for this is because you can adjust the amount of memory usage in exchange for number of ALU cycles (this is exactly what the "lookup_gap" variable does in reaper/cgminer). But at a value of N large enough to prohibit faster GPU mining, a single hash is measured in seconds to minutes. This makes it impossible to use for a blockchain hash and was detailed in the memcoin thread I created a long time ago. YACoin is actually more or less exactly what I specified in the original memcoin thread. This is one of the reasons that the YACoin chain makes no sense itself, because it eventually intends to use values of N in the gigabyte range.

I meant scrypt(1024, 1, 1) should be set to something like scrypt(1024, 128, 1)*. Setting the second parameter to "1" means that IT'S NOT GPU RESISTANT. Artforz fucked a lot of guys with his set of parameters.

* scrypt(CPU cost, memory cost, parallelization cost)

tacotime

legendary

Activity: 1484

Merit: 1005

Quote from: Come-from-Beyond on May 09, 2013, 06:09:39 PM

Quote from: Come-from-Beyond on May 09, 2013, 06:09:39 PM

You might be thinking of the GPU mining code -- no one knows if Artforz was able to get a GPU miner for Litecoin running after he realized that the TMTO issue with scrypt allowed for faster mining by GPU. It's still a possibility; Artforz mined privately with his GPU miner when he implemented it for the Bitcoin chain, then only gave out the source code later.

I recall an issue related to the chosen parameters (1024, 1, 1). This set of parameters leads to degeneracy of the algo (in mathematical sense), making it's easier to create GPU mining soft.

Indeed, at the time someone immediately called out ArtForz about the choice of scrypt(1024,1,1) in pretty much the very next forum post. I believe the wording was something along the lines of "are you joking?" and "don't you know anything about OpenCL?" when ArtForz suggested it would be GPU resistant.

Artforz logic was that the L2 CPU cache would be fast enough that it could compete with a GPU implementation. I spoke with him myself around the time of the release of Tenebrix; the numbers he gave me for his naive GPU implementation were in line with those when running reaper with lookup_gap = 1 (which offers at most a 200% advantage).

WindMaster

sr. member

Activity: 347

Merit: 250

Quote from: Come-from-Beyond on May 09, 2013, 06:09:39 PM

You might be thinking of the GPU mining code -- no one knows if Artforz was able to get a GPU miner for Litecoin running after he realized that the TMTO issue with scrypt allowed for faster mining by GPU. It's still a possibility; Artforz mined privately with his GPU miner when he implemented it for the Bitcoin chain, then only gave out the source code later.

I recall an issue related to the chosen parameters (1024, 1, 1). This set of parameters leads to degeneracy of the algo (in mathematical sense), making it's easier to create GPU mining soft.

Indeed, at the time someone immediately called out ArtForz about the choice of scrypt(1024,1,1) in pretty much the very next forum post. I believe the wording was something along the lines of "are you joking?" and "don't you know anything about OpenCL?" when ArtForz suggested it would be GPU resistant.

tacotime

legendary

Activity: 1484

Merit: 1005

Quote from: Digicoiner on May 09, 2013, 05:57:33 PM

You might be thinking of the GPU mining code -- no one knows if Artforz was able to get a GPU miner for Litecoin running after he realized that the TMTO issue with scrypt allowed for faster mining by GPU. It's still a possibility; Artforz mined privately with his GPU miner when he implemented it for the Bitcoin chain, then only gave out the source code later.

I recall an issue related to the chosen parameters (1024, 1, 1). This set of parameters leads to degeneracy of the algo (in mathematical sense), making it's easier to create GPU mining soft.

That wasn't the problem (although the author of this coin is trying to use that to sell you on it).

Regardless of the value of N, the GPU will likely be faster until you hit some value in the hundreds of megabytes that requires all of the 8-16 GB of RAM available on a modern computer. The reason for this is because you can adjust the amount of memory usage in exchange for number of ALU cycles (this is exactly what the "lookup_gap" variable does in reaper/cgminer). But at a value of N large enough to prohibit faster GPU mining, a single hash is measured in seconds to minutes. This makes it impossible to use for a blockchain hash and was detailed in the memcoin thread I created a long time ago. Most of the ideas for YACoin came from the specifications for that coin, which were later revealed (to my chagrin) to make no sense. This is one of the reasons that the YACoin chain makes no sense itself, because it eventually intends to use values of N in the gigabyte range.

WindMaster

sr. member

Activity: 347

Merit: 250

and the fact there were no clear instructions on how to CPU mine. The instructions were made more clear later on. I didn't realize you just had to run the wallet with a few switches to start mining. Also all those pictures posted a couple of days ago of people eating popcorn was a signal some people were raking in the coins.

Possibly this is more an issue of people forgetting the past, back when CPU mining BTC was common. I'd tend to say it wasn't a flaw of the YAC launch, just that a lot of people had forgotten how to CPU mine with the client itself, or entered the cryptocurrency scene and went straight to GPU mining without having ever attempted CPU mining. Otherwise, the "gen" parameter to enable CPU mining in the client seems to be a long-term legacy dating to the original Bitcoin client and wasn't any sort of revolutionary or undocumented feature specific to Yacoin.

Come-from-Beyond

legendary

Activity: 2142

Merit: 1010

Newbie

Quote from: Come-from-Beyond on May 09, 2013, 06:00:00 PM

Quote from: Come-from-Beyond on May 09, 2013, 06:00:00 PM

Are you kidding me? I was one of the first Tenebrix and Litecoin miners, and I mined with an AMD cluster. Artforz just had trouble optimizing for Intel, so for a while AMD and Intel were neck and neck until pooler did some tricks with cache access and assembly to finally get the performance out of intel GPUs that everyone was expecting from them. End of the day, a few months after Tenebrix was released you could finally mine with 60 KH/s on your Intel CPU as compared to 40 KH/s with Artforz' original code. Whoop-de-doo, a 50% improvement.

Well. I may be wrong and that was unintentional.

You might be thinking of the GPU mining code -- no one knows if Artforz was able to get a GPU miner for Litecoin running after he realized that the TMTO issue with scrypt allowed for faster mining by GPU. It's still a possibility; Artforz mined privately with his GPU miner when he implemented it for the Bitcoin chain, then only gave out the source code later.

I recall an issue related to the chosen parameters (1024, 1, 1). This set of parameters leads to degeneracy of the algo (in mathematical sense), making it's easier to create GPU mining soft.

tacotime

legendary

Activity: 1484

Merit: 1005

Are you kidding me? I was one of the first Tenebrix and Litecoin miners, and I mined with an AMD cluster. Artforz just had trouble optimizing for Intel, so for a while AMD and Intel were neck and neck until pooler did some tricks with cache access and assembly to finally get the performance out of intel GPUs that everyone was expecting from them. End of the day, a few months after Tenebrix was released you could finally mine with 60 KH/s on your Intel CPU as compared to 40 KH/s with Artforz' original code. Whoop-de-doo, a 50% improvement.

Well. I may be wrong and that was unintentional.

You might be thinking of the GPU mining code -- no one knows if Artforz was able to get a GPU miner for Litecoin running after he realized that the TMTO issue with scrypt allowed for faster mining by GPU. It's still a possibility; Artforz mined privately with his GPU miner when he implemented it for the Bitcoin chain, then only gave out the source code later.

Come-from-Beyond

legendary

Activity: 2142

Merit: 1010

Newbie