SCRYPT Parameters Explained..

[spartacusrex]

I am trying to understand the Scrypt parameters used in Litecoin..

As I understand it,

Computer power : N  = 1024
Memory Usage   :  r   = 1
Parallellization   : p  = 1

Can someone explain how these all fit together and why they were chosen ?

I found this - '.. With the scrypt parameters used in litecoin’s implementation (N = 1024; p = 1; r = 1), each thread needs about 64-128 KB..'

How much memory do you actually need to mine them, then ?

Can someone explain how to make the scrypt function require CONSIDERABLY more memory than is currently required.. enough so that GPU mining really is not possible ? Is this Possible ? say 1 or 2 GB min.. ?

Thanks..

   

[Mike270]

I think you need to think bigger... my GPU already has 3GB, whereas my PC has 24GB.

[UKMark]

I'm mining successfully with 3 x 2GB DDR5 GPU's and 4GB of system RAM - It's pretty much useless for any background tasks though But it's a 24/7/365 miner so not an issue for me.

[spartacusrex]

So - Is the whole GPU resistant SCRYPT over now ?

Seems like the GPUs have just as much RAM as the CPUs anyway..

ASICs and FPGA boards are still not used for SCRYPT mining though.. right ? Will that stay like that for some time.. ?

I like SCRYPT because it means you can mine bitcoin and 'some' litecoin variant at the same time. Is this still true ?

I am working on something a little different.. something new.. 

[Impaler]

I guess at the point when the typical video-card has more memory then the typical motherboard yea, no value would be able to exclude GPUs while still being minable on a CPU.  But I don't think we are to that point quite yet.

I think I also read somewhere that the opposite strategy might work, reducing memory to the point that it all fits in L1-cache which is faster then GPU memory might give the CPU the advantage again, but I'm not sure, this needs to be answered by a real expert.

Ultimately PoW is just never going to be remotely well distributed (if we define that as most mining done by regular peoples regular PC's), it will always flow towards centralization and is thus not a viable long term basis for crypto currency.

[Etlase2]

Ultimately PoW is just never going to be remotely well distributed (if we define that as most mining done by regular peoples regular PC's),

It can be if you can come up with something better than just a brain-dead algorithm.

[spartacusrex]

Well, maybe we need to understand/realise/'get-used-to-the-fact'  that someone/anyone with LARGE computers needs to MAINTAIN the NETWORK ?

Is that so BAD ?

If people want to be able to log on to the network, in the leaf-peer non-miner mode, then someone is going to have to deal with their requests. And there in lies the problem. As they need to be rewarded for doing this, as is only right.

PoS just doesn't make sense to me, as someone who is prepared to get 51% of the HASH power, would surely also be prepared to buy a large enough PoS to mess around with. no ?

Insert Quote
Quote from: Impaler on Today at 11:06:31 AM
Ultimately PoW is just never going to be remotely well distributed (if we define that as most mining done by regular peoples regular PC's),

It can be if you can come up with something better than just a brain-dead algorithm.

The PoW system is the very basis of the timestamp server at the heart of bitcoin. I think ,for now, it's all we have and will have.. Although I do like the whole p2pool mining.. that looks interesting.. maybe it can be embedded deep into a new coin.

AND so - in conclusion, SCRYPT is only valuable in that it can be run at the same time as a normal bitcoin miner, so that a miner can do both. ?

Well, that's something.

[spartacusrex]

@Etlase2 - just read your DECRITS post you link to. Very interesting. At least it's original.