Topic: China-made CPU coins, the truth is... (Read 3796 times)

botnets are largely CPU based.  Primecoin is an example of a true CPU coin.  There's a GPU miner but it's very inefficient because GPUs just aren't good at prime chains, it's complex math that pretty much requires a complex core, not a bunch of little simple math GPU cores.

PS:  Still trolling I see, eh number37?

well the term "cpu only" means that it is not still efficient to mine with gpu...

No such thing as a cpu only coin some one else will always find a way to mine gpu. if not they still got botnets out there

I am chinese, and I always keep myself away from chinese coins because I know them, I know this country very much! 

but do keep in mind it's the Chinese who drove up BTC price  xD

This uses Quark algo, and Quark is not Chinese.

The algo is used by many coins now.

I would stick this topic if I could.

 How about MOL?

Well why posting under a fake name? registered a new account eh? You mentioned Sam because we were both supporters of the Elephantcoin? You are a hater? Well loser look at what ELP is today, I saw 3 new mining pools created for ELP lately.

I used my account to post all the time, not like you. Why not you reveal you real user ID to everyone, so we know your history. Retarded why posting under a cover!

scam coins

This is the best guy who know the truth

next time, when criticize others, look at yourself first...

Union Coin is based on Quark algorithm. We know that there is a GPU miner for Quark but it's fast like CPU miner. Do you mean that Chinese have more powerfull GPU Quark miner (at least 10 times faster)?

How surprising that a number37 thread has flyingcat and sam53 replying noncontent

CPU's don't have the parallel execution resources that a CPU has, but CPU's use their transistors very differently.  Here's what I would add to an algorithm to guarantee that it's CPU-centric:

A 1MB table of pointers/references, which can be held in a L2/L3 cache.
Long, sequential series of reads.  Each pointer read in from the L2 cache points to a different location in the table.
Double-precision floating point operations.  Most CPU's have multiple FPU's, but GPU's almost always have single-precision only.  Double must be done in software.  IEEE 754/854 requires that results must be within 1/2 ulp and that, if a calculation produces a slightly larger result, that the FPU must produce results that are >= the previous results.
Many sequential compares, followed by sometimes-taken branches.  Typically, CPU branch prediction schemes produce greater accuracy than the static predictors generated by the compiler. 
Branches that depend on the results of two very close double-precision operations must be tested for equality, CPU's handle this very well.  GPU's and ASIC's will not.
64-bit integer operations with long strings of dependencies.  Modern, 64-bit CPU's tend to handle mask, shift, and logical operations well.  Use 16-bit signed-extended masks and shifts by variable amounts.
CPU's can also store these 64-bit values to the L1 cache, then load a variable part of the last load plus the remainder in a subsequent load.
Software division of very large (128-bit+) integers.  These must be done meticulously, and are notoriously difficult for highly parallel systems.  In general, only one or two operations can be issued at one time before the results of the previous operation have returned.
Parameter passing through shared L1/L2 caches, which GPU's and even supercomputers are not always ideal for.  If NSA uses its supercomputers and data centers to attack the Block Chain, their hardware will not be the ideal for these algorithms, even if they are using massively parallel systems with thousands of CPU's, because only one CPU can be working on a specific attack at a time.

In effect, the whole Block Chain must be put together over the course of several years as long sequences of many billions of PFLOP's and sequential loads from the L2/L3 cache.  These long chains of data-, load-, and control-dependencies inhibit parallelism in even the best code.

GPU's would not handle such well, but neither would supercomputers or ASIC's.  Only the CPU in your laptop or desktop is optimized for handling long sequences of data-dependent and control-dependent operations.  As important, the caching systems on modern CPU's are better oriented for database-style sequential accesses. 

Unless the algorithm is written this way, it is unlikely that CPU-mining will remain viable for long, even if that was not the intent from the beginning.

Suter coin did this again, I feel very sorry to these guys, they don't care honour and long time business.
The Chinese people who were borned before 1980 are hopeless.

You from China too, right? Check these coins, they all have Chinese websites, and check details there, you know who made them

Are u sure all of this coin is from China?
Why?

freedom of taking advantage, lol

Why are you stupid?)))
Do you jealous  to Chinese.
I think they, unlike you, could do that.
But they do not do it because they understand  the freedom and community better then you.