I agree that unclear is the operative word. I have feeling that Intel's James Reinders is heavily under the influence of the marketing department. There's some talk that the current board is a coprocessor, yet the ISA manual clearly shows the unit booting in the 16-bit segmented real mode.

Intel is either doing artificial market segmentation or something didn't work out in the memory controller/quickpath/chipset interface portion of the design.

I also wonder if the references to the "original Pentium" are similar to the branding exercise that happened with the announcement of the orignal Atoms. The Atoms had completely redesigned microarchitecture called Bonnell, but with various features disabled. Yet the marketing described them as "reissue of the classic Pentium" while pretty much the only thing that they had in common was lack of deep speculation and in-order execution.

Mining SoC is a technical nonsense. The mining controller has to be on a separate chip with separate cooling to reliably monitor the error rate of the mining portion of the hardware.

SoC only makes sense if you aren't planning on overclocking or otherwise pushing the envelope. This isn't going to happen in the mining business.

I wonder how this number is going to change once we include the information that the basic core resembles Pentium which was dual pipeline and that now the cores are 4 way hyperthreaded.

From the Pentium days I remember straightforward Fortran & C code easily retiring more than 1 instruction per clock: 1.3-1.6 with nothing more than "-Ofast".

I would double your number to 560 Mhash/s. This should be a safe assumption that the pipeline utilization could get close to 100%.

I'm sorry, I though it was obvious after the housing bubble and how widely publicized it was.

In the USA "own a home" most frequently means "have a mortgage to pay" or "have a home equity line of credit to pay".  The actual home-owners are surprisingly rare, especially in the more prestigious or upscale locations. The stress caused by high debt load in turn causes many family problems.

On the other hand "voluntary debt slavery" may be considered a new social glue for the USA. There's no easier man to scare that the one who has a mortgage and lifestyle to maintain.

Edit: OK, I think I just gotten myself trolled. Here's the quote from the redneck link I posted above. And with this quote I'm going to end participating in this thread.

Living around Rednecks has had a profound effect on my way of looking at life.  Once while driving my fancy new car, I visited a country cousin and committed what Redneck Theologists call the “sin of pride” when I mentioned that the car had cost me a fortune.  In good humor, he pointed to his giant farm tractor and said, “Well, I’m impressed. See that reaper over there?  She cost me over $200,000.00; I paid cash; and I reckon I only take her out a few weeks a year”.  Man, was I ever humbled in the presence of such profound Redneck wisdom.  I now think twice before bragging.

The phrase "my own home" means something else in the USA than throughout the rest of the world.

In fact, the American society is so diverse, that even in the USA this phrase lost any meaning without the cultural context.

http://dba-oracle.com/redneck.htm

Age isn't a number. The best phrase I heard about this is from my former landlord and a real-estate manager who insisted on face-to-face meetings to ascertain the credit rating: "fifty going on fifteen".

Please feel free to call yourself "young at heart" if "youngster" was offensive. I apologise.

I presume that with your car dealership story you've meant to convey that you are an experienced point-of-sale credit user, that you've obtained car loans from the car salespeople, morgage from the builder's salespeople, player credit at the casino cage, etc.

In that case the secret you are missing could probably be expressed in the old Jewish proverb: "Don't ask the tailor if the suit fits."

I apologise if I misunderstood you and the "nothing more than a signature" was in fact a signature on your check that your presented to the fleet sales rep at the car dealer's.

Ah, a land-less and family-less youngster whith a post-secondary education (or while obtaining a post-secondary degree) with "stellar" debt-slave score. I just hope that you aren't black, because my pot-shot was all equal-opportunity, non-discriminatory and race-blind.

I'll tell you the story. Years ago I went to the bookstore to pick up a special order. Due to mistakes the book I was handed was something like "Proceedings of a TRW conference on numerical modeling in consumer finance {or some such}". It was raining so I spent couple of hours flipping through it, but I now still regret not buying it.

There was a ton of super-interesting nuggets in it, but I particularly remember the "behavioral modeling" chapter. There was a matrix of data from MBNA who used to run credit cards affiliated with professional associations. In most of the upper bands of income the most profitable (for MBNA) card was ADA (American Dental Association) and the worst performing was AICPA (American Institute of Certified Public Accountants). Not surprisingly dentists had higher credit scores than accountans, normalized for all other factors.

The moral of this story? The "credit score" isn't meant to measure the likelihood of you repaying the debt. It is meant to measure how profitable of a customer you are going to become for the credit grantor. There exist people who wouldn't consider "high credit score" to be something to be proud of.

I'm not interested in providing credit counseling to the "how-much-per-month" crowd. There are at least two industries providing that advice in the USA: one for-profit and one non-profit. They have quite respectable success ratios. But as they say in a joke:

Q: How many psychotherapists does it take to change a lightbulb?
A: Normally just one, but the lightbulb has to want to change.

What can I say to that? I envision for you the future of using check cashing, payday loan, rent-to-own & pawn shop establishments.

It can happen to anyone; I'm not sure if this is the correct link for Collateral Lender of Beverly Hills. It is a lifestyle choice and also probably lasts a lifetime.

http://www.collaterallender.com/

Or maybe your future wife will be your credit manager? I've seen that too.

This limitation was there since forever. The root of it is in the costs of insuring and maintaining the account. I'm kinda thinking that "sudden" probably means "first time in my life that I read the account disclosure booklet. I used to simply throw it away."

One thing that US banking system has figured out exceedingly well is the default and fraud risk on personal accounts. There are so many flags that signal imminent personal bankruptcy, one of them is lack of normal monthly or biweekly budget planning. It manifests itself by too frequent too small withdrawals. The limits are on the quantity of individual withdrawals, not on the total amount withdrawn.

Railing against those limits is akin to railing against higher car insurance rates for unmarried men less than 25 years old. It leads nowhere because it has no wider social support.

Same difference. There is a 2016-way decimator (or subsampler) at the output of the I or PI controller.

The decimator is there to assure that the blockchain converges after an emergence of orphans right after the difficulty retarget. Without the decimator the blockchain would either completely split or took much longer time to converge to a majority.

The I vs PI argument is somewhat hair splitting because if it is PI then the proportions are about 1/2016 P and 2015/2016 I. Additionally there is at least one off-by-one bug in the computation of the feedback value so the equations aren't optimal but about 1/2016 away from the optimal.

Incidentally (or by design) integrator is also an approximation of an average. And average happens to be maximum likelihood (or maybe minimum variance or some such) estimator for the lambda parameter in the exponential distribution governing the block-winning lottery.

I have a feeling that Satoshi had a very peculiar sense of humor, intentionally setting a traps for people who kinda but not really understand what's going on. Very similar to the choice of stochastic knapsack solver for the wallet coin selector. I now always imagine him cackling at the discussions here from a safe distance.

I don't think Bitcoin requires that fix. The feedback system in Bitcoin is either asymptotically stable in the Lyapunov sense or extremely close to being asymptotically stable. By extrememly close I mean that the maximum amplitude of oscillations that could be introduced in it is on the order of LSB in the very narrow floating-point format that is used to represent the difficulty.

On the other hand the designers of alt-coins cranked up the gain in the feedback loop and applied asymmetric clipping in the feedback signal. The resultant system is probably still stable in the Lyapunov sense but no longer asymptotically stable.

Satoshi either understood the stability problem in the theory of closed-loop control systems or at least consulted somebody who understood the problems.

On the other hand the designers of alt-coins are like kids who stood up with microphone in front of the speakers and cranked up the gain on the amplifier.

http://en.wikipedia.org/wiki/Lyapunov_stability

The amount of work required to maintain the alt-coins can be easily reduced by more than an order of magnitude.

Just cease to maintain them as forks in the github sense.

They need to be rewritten to Perl or other text-processing language.

The input to said Perl program should be the Bitcoin source tree and the list of the rather trivial changes that need to be made to various constants and identifiers in the Bitcoin source.

The output will be altcoin source tree buildable in the exact same way as the Bitcoin source tree.

This is a proven methodology that was in use when the source programs were in Fortran and the text processing language was Snobol.

First of all Spartan-6 by itself isn't something that would compete with modern CPU. This is a 45nm device targeted to be cheap and power-efficient.

The absolute maximum for clock is 1080MHz, but only a fraction of the device can run at this speed because the chip doesn't have the heath spreader.

Internal block RAM is a dual-ported SRAM rated at 320MHz zero-wait-state access. When people synthesize soft-CPUs on Xilinx FPGAs they use it as L1 cache or simply there's no cache and the whole RAM is 536kB that is on chip.


Spartan-6 LX150 has 4 memory controller blocks that can control single-chip external DRAM (DDR, DDR2, DDR3, and LPDDR) up to 800Mtps.

But currently only ztex (and maybe Enterpoint) make boards that have external DRAM chips, and those aren't the boards in the hands of most miners.

Besides, emulating CPU or GPU in FPGA isn't the architectural road to take the advantage of the famed power efficiency of the FPGAs. Those devices achieve their optimum efficiency at relatively low clock speeds by doing maximum usefull work per clock and minimizing the on-chip logical resources that aren't used.

And rename the project to BFLcoin. 

You can plead with the developement team to implement a flag like GCC has to make the stochastic algorithms reproducible:

-frandom-seed=string

Edit: And I have to immediately correct myself. The results will be reproducible within the interblock time window. Once the new block arrives the fee computation will change again. Maybe also another option will be required to temporarily pause the local updates of the number of confirmations in the wallet:

-fpause-blockchain=chain-height

Anyone has any other suggestions how to make Satoshi client reproducibly testable? Maybe the pausing should be done also through the RPC interface to allow unpausing?

Immature time-wasting crank with learning disability detected. I'm now sorry that I engaged in discussion with you.

It isn't a bug, it is a feature. Satoshi Bitcoin client uses a stochastic knapsack solver to select the coins and compute the associated fee.

I don't buy this argument.

Quote from scrypt.c:

const int scrypt_scratchpad_size = 131583;
char scratchpad[scrypt_scratchpad_size];

Quote from XC6LX150 data sheet:

Total Block RAM (kb) = 4824

converting bits to bytes I'm getting 617472, which would make litecoin hasher easily fit into the popular Spartan-6 chip already in hands of many miners.

Same with me, I'm fine reading the recent BFL/ASIC chatter in whatever section it was originally posted.