The network hash rate is REALLY picking up this weekend, we're at 10 blocks per hour almost.  If this keeps up we'll see 6m by next difficulty change.

You can regenerate the lookup table on the fly instead of storing the whole thing (just work in the space you have; for every lookup, you can generate the whole table by building the beginning of it in the RAM you have and then overwriting it to get to the next series of values).  This is used to speed up the GPU implementation a bit (lookup gap = 2, use approximately half size memory tables as I understand it).

Note that by having to regenerate the table in sequential fragments, things get very expensive very quickly in terms of having to overwrite the RAM buffer a lot and the ALU cycles required to regenerate the table.  This is why it is considered memory hard, and why it's proving a giant pain in the ass to make FPGAs for so far.

The implementation of scrypt for litecoin uses N=1024, r=1, p=1, hence the 128 KB of usage.  You can use more RAM by increasing N or p.  See: www.tarsnap.com/scrypt/scrypt.pdf

Using more than 128 KB with N=1024, r=1, p=1 will yield no faster results

Overclocking is with another program eg afterburner

Please let me know what your output is in the console when you are trying to connect to a pool with cgminer (it should show some error)

This I have never seen. Which miner are you using and over what length of time ddoes this occur?

Thanks for that! Hopefully I'll add it to my guiminer tomorrow after I test it.

It was compiled under Win7 and doesn't seem to work on x86 or x64 windows XP because of an incompatible build of wx.  I will try to compile a version later for Windows XP.

Yeah, this too eventually.

Well, the source code is published on github, but no technical explanation of many of the properties or of the new update has ever been given.

It depends how much you trust a single, secretive developer with your money.  Unlike bitcoin development, improvements to the chain are completed by a single person without public testing or public awareness.

If the whole team of bitcoin developers have already introduced a couple catastrophic errors into their chain (overflow transaction generation bug, recent blocksize incompatibility bug) that required a hard-coded fork of the chain to correct, would you really feel confident that PPC can, in the long term, provide you with security?

I would love to recommend PPC and its technical paper, but I can't recommend the former because the latter simply doesn't exist in any adequate fashion.

Sorry, made some edits for clarity.

Mass introduction of ASICs will be within 1-3 months, 3 months maximum.  There is simply too much money to be made there right now, people are becoming keen to the fact that bitcoin mining consumers will throw millions of dollars at them if they can provide ASICs for them.

My guess is that we'll see all three likely scenarios play out concurrently;
1) Drop in BTC price.
2) Hike in BTC difficulty.
3) Hike in LTC price.

No, I gave the cost in power to mine 1 USD in LTC and BTC.

So, Litecoin is twenty times less profitable than mining BTC right now with the inefficient Avalon ASICs.  Once BFL ASICs come out, or some other better competitor, we could see the cost to mine 1 USD in BTC decrease another 5-10 fold in the next few months.

Given this, we might guess that
1) BTC price will drop 20-100 fold or difficulty will increase 20-100 fold
or
2) LTC price will increase 20-100 fold (or LTC difficulty will drop 20-100 fold; unlikely)
 upon the mass introduction of SHA256 ASICs so long as no LTC scrypt ASICs or super efficient FPGAs come out (I doubt they will, personally).

Crossposting here to show that yes, Litecoin is still cheap

Could you update this too with the cost of power per USD earned at current difficulty?

The calculation is like this:
For LTC right now it takes 131.7 KH/s (diff = 84.2242236212) to mine 1 USD equiv in 24 hours.
A 7950 pulling 190 w/h will get about 575 KH/s.

190 w/h * (131.7 KH/s / 575 KH/s) = 43.52 w/h * 24 h = 1044.44 w = 1.04444 kw at 10 cents a watt = 10.44 cents.

However, for BTC we should use the wattage of most efficient miner, Avalon, which is 60 GH/s at 500 W (~400 w DC)
At diff 4,847,647, we need 0.2059 GH/s to mine 1 USD equiv. in 24 h.

500 w/h * (0.2059 GH/s / 60.0 GH/s) = 1.716 w/h * 24 h = 41.18 w = 0.04118 kw at 10 cents a watt = 0.4118 cents.
You could also do the calculation for GPUs.

In other words, it currently costs 20x as much power to mine 1.00 USD in LTC as compared to BTC.  Such a calculation would be immensely helpful to people trying to figure out what the valuation should be of scrypt versus SHA256 alternative currencies.

You do realize someone mining with 3 ghs of hardware makes thousands of dollars per day mining litecoin, right?

None of mine are hardware locked. I'll try and post a pic of my pcb after, wondering if they did a silent revision.

What was it? I'd like to fix the reaper kernel too if it was a problem with that.

Offering 1.2 BTC shipped to Canada

I'm frankly surprised no one has pursued legal action against BFL at this point.

Updated cgminer settings for 7970

http://www.litecoinpool.org/pools
Total   3,744,400 kH/s