|
24 hours and no blocks. :{ Is there any way to check if this is working okay? I'm pulling 750w/hr at the wall so I don't want to waste power
|
|
|
|
Hopefully, I left it mining overnight and I didn't find one yet, but I'm hopeful today I'll find one. I remember soloing before with my CPU that sometimes you'd wait 24 hours for a block and then you'd find a streak of them in a few hour span. |
|
|
|
|
Mining BTC (520 mh/s on a 7950): 165 w/hr per card
Mining LTC (510 kh/s on a 7950): 227 w/hr per card
|
|
|
|
I been mining litecoin for a few days now, but I realized... What can I do with it? When I started mining, I assumed there would at the very least be some place where I could exchange it for BTC. But I can't even find that!
Is litecoin a waste of energy?
btc-e.com |
|
|
|
|
just started soloing with 1650kh/s (3x 7950s), hopefully i'll get some blocks soon. thanks for the guide!
|
|
|
|
|
It really says something about the bitcoin community when China has the most transparent hardware companies
|
|
|
|
DING we have a winner! 590 k/hash :-D
Now I'm going to try to get the 7870 up to par.
np. good luck. LhY6GBzeLCjhmNP4F9UWMNmWkyhGVrpQwv Payment sent. No luck yet for the 7870..  anything equivelent or close to what worked on the 7950 seems to fall on its face. Thanks! Try these thread concurrency values with the same settings otherwise as for the 7950, but with aggressions between 13 and 20 13568, 14336. 16000, 16384, 17408, 18432 The concurrencies are really strange, for instance I have no explanation of why 8000 is optimal for a 7770 |
|
|
|
Hello, my name is Sonny Vleisides.
...
Kind regards,
Sonny Vleisides
This is exactly the kind of thing somebody pretending to be Sonny would say. Prove it or you don't exist. And, when my super computer arrives next week I expect it to power itself. But...he posted a picture taken in China Missouri on Christmas 2010! What more proof do you need? Edit: unsigned numbers in EXIF parsers are bad! Hmmmmm. |
|
|
|
12.6+ AMD drivers do not let you change the memory speed of 7xxx (7950, 7970, etc) cards below a certain threshold. To underclock memory, use Afterburner and a dll file. 1) Install MSI afterburner. 2) Download this dll file and place it in your Afterburner directory. 3) Edit the afterburner config text according to this thread and set unofficial overclocking mode to 2. 4) Run MSI Afterburner. 5) Set the overclocking mode to kernel instead of software in settings. 6) Underclock memory as per usual. It took me a long time to figure this out from various threads, so hopefully now people will find it easy. This also allows you to overclock memory above the overdrive maximum. |
|
|
|
DING we have a winner! 590 k/hash :-D
Now I'm going to try to get the 7870 up to par.
np. good luck. LhY6GBzeLCjhmNP4F9UWMNmWkyhGVrpQwv |
|
|
|
I am very much considering cancelling my BFL pre-order and going with yours in light of all the bad press that BFL has gotten lately. I just wished you could get them sooner, as by the time they will ship, the payout would have dropped, and the difficulty will be insane, stretching the ROI on one of your units.
Or you could, instead of cancelling your order, sell your order number to somebody who wants a smaller order number than one would get if ordering now. I just can not imagine how do people believe that 20 employes including management with a nice looking old woman soldering singles working in a small place will assemble and even test 8000+ orders and catch up with promised delivery time. It is not making any sense for me having in mind that testing process will take a LOOT of time if you know what i mean... Let us say they will outsource hardware assembling. I am quite sure that they will do testing by themselves for obvious reasons.. The math is 8000/20=400 units per person. ASICS probably shall be flashed in a first place which costs time. Each test shall continue at least 24 hours (amusing they were not be able to fire all 400 units at once) which makes about ...i do not even want to calculate it Any way everybody takes own decision... It's China dude. You can hire people to assemble and solder boards 12 hours a day 7 days a week for $200 or less a month by the tens of thousands. |
|
|
|
I'll post it later For 550kh/s+ on my 7950 I use this config: host ???
port ???
user xxx
pass xxx
protocol litecoin
worksize 256
vectors 1
aggression 20
threads_per_gpu 1
sharethreads 32
lookup_gap 2
gpu_thread_concurrency 24576
This person managed 700kh/s on water: http://anonymouse.org/cgi-bin/anon-www.cgi/http://img38.imageshack.us/img38/2250/reaperg.jpgedit: After playing with this for a while and plotting thread_concurries and kh/s quantities, I can say the following: - For any given card there is a thread_concurrency value range which is optimum with lookup_gap == 2. For the 7770, this span is 6080-11456. Below this range the number of stales increases, with it becoming mostly stale in the 4000-ish range. Above this range, the speed in kh/s drops dramatically. Within this range, speed in kh/s seems to plateau and is nearly constant. - Using the lowest thread_concurrency values seems to use the least power (I guess because less memory is being used). - You can run 1550MHz+ RAM on these cards using MSI Afterburner with this file in the Afterburner folder and overlocking mode set to 2 in the Afterburner configuration. |
|
|
|
You should be able to get more kh/s mining ltc than you get mh/s mining btc. As I said, a 6870 should pull 350kh/s+ easy. I get 550 kh/s on a 7950. You do have to wait a little while for reaper to reach the correct reported hash rate too, it took me about 6 minutes before it stated 550 kh/s.
What driver and SDK version are you using? I'm using the latest on win 7.. and no matter what settings I try I can't seem to get above 300 on either of my cards on cgminer or reaper. Bounty of 25 LTC to whoever can get me some proper hashrates on a 7870 or 7950  try host  port user xxx pass xxx protocol litecoin worksize 256 vectors 1 aggression 16 threads_per_gpu 1 sharethreads 16 lookup_gap 2 gpu_thread_concurrency 8000 in reaper. 7770 gives me ~190kh/s overclocked with same settings and gpu_thread_concurrency 8000 (64 * 125) with ccc 12.8 and app sdk 2.6 a 7870 has 2x the shaders as a 7770 so you should get ~380kh/s and my 7950 gets ~550kh/s The trick must be finding the right thread concurrency for your card, which at this point almost seems random. I would just multiply 64 by progressive numbers above 100 up to 256 and you'll probably find something that works the fastest. |
|
|
|
|
LTC has been around for almost a year. It's not dead and it's probably not going to die. Like BTC, the algorithm is self adjusting so that it remains profitable to mine. The naysayers are welcome to sit around in a circlejerk waiting to profit from nonexisting ASICs they bought six months ago while I continually make money with a GPU farm.
|
|
|
|
At ~600W per unit, what heat management solution will be used?
I would guess it will be computationally split between 6-10 dies... This would enhance yield. I doubt they would be making 4000mm^2 dies as that's pretty much unheard of. And then those smaller dies would be cooled with $10 HSFs. |
|
|
|
Looks like a smokers case..... Jesus bro
Leave him alone, he lives inside a volcano |
|
|
|
I'm talking about the minimum memory clock that 7970 cards enforce, which is 150MHz less than current core clock. If you really found a way to get around that limitation with just cgminer, I'm sure many would be interested in how you did it. But it's more likely that you just think you are running at 300MHz memclock, while in reality the card picked something much higher when you requested it.
AFAIK, getting around this limitation is only possible with some Windows software that use the cards in undocumented ways. For example MSI Afterburner and Sapphire Trixx (both happen to be made by card manufacturers).
You need a dll file from the 12.1 version drivers in your afterburner directory. I posted it in a thread a week ago. It works fine on my 7950s which mine with mem at 200mhz and core at 1000mhz/1.03v. I get 514 mh/s at 140w per card, or 3.67 mh/s/w. |
|
|
|
........... but likely a collapse will follow of an even greater magnitude than that following the 2011 tanking to $2.
Are you saying you believe the price will collapse to $2? That would be incredible. Very few people would be investing in FPGA or ASIC technology if they believed the price will see $2 again. Instead of buying expensive paperweights, they'd wait for $2 bitcoins and buy them! It will be relative. 2011 saw BTC go from $30 to $2.50. 2013 may see bitcoin rise to $125 and drop to $5. I hold 50:50 btc:ltc, so as long as one of them goes up in valuation by an order of magnitude I win. |
|
|
|
|
Show Posts
