Consolidated Litecoin Mining Guide for 5xxx, 6xxx, and 7xxx GPUs

[pox67]

This Litecoin mining is less a Science at this point than it seems to be Sorcery.

I had to completely and radically change the cgminer settings to get the same performance from this card on a riser in the 3rd slot. 

Why on earth would two identical cards require wildly different settings in different slots on the same motherboard but in the end perform the same?  I think there is a LOT better performance that could be had from GPUs in scrypt mining - this inconsistency shows unoptimized and possibly buggy software imho. (at least on RADEON 79XX).

I agree, there are sweet spots that you can find though.
For instance at 950 core I get the same performance on a 7970 at mem speed 1225 and 1425. If scrypt is all about memory speed then that makes no damn sense. Everywhere between 1225 and 1425 mhz is very bad KHs.

I think the 7 series cards might have a lot of power hidden in the. But then if someone writes a better miner why would they release it?

[1715144320]

1715144320

[1715144320]

1715144320

[1715144320]

1715144320

[1715144320]

1715144320

[Melfice]

I'm running HD 7950 and 7770.
Here's the command line for cgminer that I pass it:
cgminer.exe --scrypt -o stratum+tcp://site -u username -p pass --shaders 1792,640 -w 256,256 -v 1 --intensity 18,14 -g 1 --thread-concurrency 21712,8000 --lookup-gap 2,2
and I get an error for the hd7950 device:
Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
Is there any way to fix this? It seems that no program accepts high values for thread concurency for me although I already set a global value for GPU_MAX_ALLOC_PERCENT 100
I also seem to get a lot of HW errors for 7950. I have catalyst 13.3 and app-sdk 2.7 installed.

For some reason, I don't get HW errors now, but still can't run high thread concurrencies, always error pops up. I've left only 7950 in the system to test it alone, but not much changed, HW errors returned

[Tejsei3]

already ordered the 7970=P
Still have 2x pci-e16 slots over, so might go for those in the future.
Btw, are ALL of those cards voltage locked? Or just the newer editions? And what temperature are your cards at when mining?
Even if i manage to burn it out after a while, 2 year waranty should help, no?

the newer revision of the gigabyte card is voltage locked.

i dont believe that the sapphire dual-x cards are locked

I got a Sapphire Dual-X 7950 (100352-2L) recently and it is voltage unlocked, 68C at 620 kH/s. Also got a 100352-4L, which is reference design, also unlocked, but runs much hotter 76-82C for the same 620 kH/s.

[anexsia]

Hi guys,

I'm a Litecoin mining newbie and I was wondering if someone could help me out with my Reaper config?

I'm mining LTC with an Radeon 6870 1GB using Reaper (because for some reason I can't get cgminer to work) and I'm getting about 280kh/s BUT when I'm running Reaper...my computer becomes pretty unresponsive and sometimes I get a lot of stales. I'm mining using the Pooledbits.com LTC pool and accumulating about 0.8 Litecoins daily if that matters any.

Here's my current Reaper config:

worksize 256
aggression 18
threads_per_gpu 1
sharethreads 18
lookup_gap 2
gpu_thread_concurrency 6144

Is there any optimal settings I should be adjusting for my 6870 and is there a way I can make it so my computer isn't so unresponsive while running Reaper and I can actually browse the web normally/do other tasks while running it?

Thanks!

[chrcoe01]

the aggression i believe is what's making it unresponsive, but lowering this will lower your hash rate.  it's up to you if you want to use the computer effectively or mine effectively

[markm]

whats the best GPU i can actually buy right now for BTC or LTC mining?

i found ATI Radion 5870s for sale online but they were for Mac.

i'm not looking to make money, i just want to mine BTC/LTC.

Are the 5870's for the Mac Pro really any different than normal ones? They seem to use PCI-e 16 like normal...

-MarkM-

[FullLife]

Looking for help to improve kH/s/W.

Here's my setup:
MSI Z68MA-ED55 (B3) motherboard
Intel G530 processor
4GB DDR3 1066 RAM
7950 Sapphire 100352-2L (purchased March 2013, voltage unlocked!)
16X-16X riser cable
Corsair TX750W power supply
AMD 13.1 drivers
AMD SDK 2.8
cgminer 2.11.3 (-I 20, --tc 22400 -w 256 -g 1 -l 1)
Sapphire Trixx

Best settings: 925/1250/.962V 560kH/s 3.29 kH/s/W.

Given high electricity cost my goal is to maximize kH/s/W.

My current best setup is 925/1250 at 0.962V for 560 kH/s at 170W or 3.29 kH/s/W, after many hours of systematically mapping gpu clock (850-1250), memclock (1250-1750), and gpu v (0.900-1.250). I've had the rig up to 740 kH/s stable (1225/1750/1.250V) but at 345W or 2.14 kH/s/W the efficiency just isn't there (would be great for a free power situation!).

Looking for advice from the experts on what to do next to improve my efficiency.

Thanks,

Tejsei

Wow, that's an absolutely fantastic efficiency!  I saw a tiny improvement (~ 5 Kh/s) changing the thread concurrency from 22400 to 16384.  See if that helps you at all.

[Nemesis]

just curious, why does the host computer need so much memory?

Doesnt the miner run on the GPU? and use GPU ram?

[xorxor]

club 3d royalking 7950 0,981v 1000/1250 589kh/s 165w , 3.57kh/J

win7 64
sdk 2.7
drv 13.1
cgminer 2.11.3

setx GPU_MAX_ALLOC_PERCENT 100
cgminer.exe --scrypt  --thread-concurrency 22400 -I 19 -g 1 -w 256

low power partialy due to 89% efficient enerprise dell PSU turned into gpu-booster[155w] + 10w from pci-e .
I checked the pci-e power only on iddle, maybe it also increases on mining so there may be an error.

[Cablez]

I finally got around to messing around with a new system to test my supposedly bunk 7 series cards.  New install of win7 on a new rig setup. My 7950 can now get up to 680kh/s at 1155/1875 with an I=18. It isn't the most stable but that can be remedied.  I am just happy I can increase intensity without taking a hit in hashrate. Will have to pop my 7970 on there and see what it can do.

For anyone with cards that hit low ceilings I would suggest a fresh install to clean up the OS and to keep it minimal. Seems to have worked for me. Using 12.8 driver by the way.

[mig91]

does anyone know why 5970s mostly crash in ltc mining?
http://forum.litecoin.net/index.php/topic,880.0.html

i have same problem with that guy. i have tryed 11.12, 12.8 and newest drivers with reaper and cg and it crash when i start mining

[Anvi]

So I had all 4 cards running in my case, but with them being all crammed together, it was way too hot.  So I pulled everything out and sat it on a rack.  I have 2 of the cards plugged into the board, a 3rd card plugged in via 16x riser cable, but couldn't get the 4th card to work with another 16x riser cable. So, I ordered a 1x to 16x riser cable and hopefully this will solve my problem.

In the meantime, I'm running 3 cards and getting pretty nice hashrates.  These are my settings and results using cgminer 2.11.3 and AMD Catalyst 13.1 drivers:

--intensity 13 -g 2  715/710 = ~1500Kh/s
--intensity 13 -g 2  925/1000 = ~1950Kh/s
--intensity 13 -g 2  1000/1206 = ~2100Kh/s
--intensity 13 -g 2  1022/1244 = ~2150Kh/s

Other settings:
--thread-concurrency 8192 -w 256

I've literally spent most of a entire week playing around with settings and trying to find the sweet spot and I feel like I'm pretty much there.  As you can see, I actually underclocked my cards to get these hashrates.  Stock clocks are 1100 core/1500 memory.  Unfortunately the default voltage is about 1.2 volts on my cards and its locked (@#&($$^(@&$!!), otherwise, I'd be able to save 100-150W easily by undervolting.

https://i.imgur.com/rzwfhgI.png

I'm seriously thinking about getting rid of these cards and getting 7950s or 7970s that don't have the voltage locked.  Having the voltage locked is such a huge drawback, I'm just wasting too much power right now.

I'm amazed that you can get over 700kHash with those clocks I managed to get 665kHash with 988/1244, but just can't manage 700+ without going 1700+ on memory clocks. Catalyst 13.1 just crashes on CGMiner 2.11.4, so I'm using 13.3 beta3.. perhaps I should try the older cgminer?

[ricciolone]

hi i'm tryng to set a mining farm

2 ati 9750
 
but i can run one of it, not in the same time... how is it possible? the same result if i use two different instance of reaper or if i set
device 0   
device 1   

in che config file and run one istance of reaper...


if i use cgminer it seems ok  but in the pool i see 1/10 hash power...

win 7 x64

ati 13.1 all package


skeda zero

kernel reaper.cl
save_binaries yes
enable_graceful_shutdown no
long_polling yes
platform 0
device 0

mine litecoin




host localhost
port 8332
user ricciolone.farm0
pass 12345

protocol litecoin

worksize 256
aggression 20
threads_per_gpu 1
sharethreads 21
lookup_gap 2
gpu_thread_concurrency 28000













skeda uno

kernel reaper.cl
save_binaries yes
enable_graceful_shutdown no
long_polling yes
platform 0
device 1


mine litecoin





host localhost
port 8332
user ricciolone.farm1
pass 12345


protocol litecoin


worksize 256
aggression 20
threads_per_gpu 1
sharethreads 21
lookup_gap 2
gpu_thread_concurrency 28000





stratum


mining_proxy -pa scrypt -o stratum.give-me-ltc.com -p 3333

[Anvi]

Your TC is too high.. See first post of this thread. Plus it takes time (10-30 minutes) for pool to report your hash rate correctly.. It's just reporting the AVG over time, so nothing wrong with cgminer.

[ricciolone]

gpu_thread_concurrency 28000 = tc

 i have set it and have 620 hash in a single card... so it's fine!!

no cgminer runs for two hours and hash on the pool was 1/10 ...
I have the problem only on 7950... it works fine on 7750

[paperkat]

Stuck at around 300 khash/s on my 7970 atm

running factory clock settings, using reaper.

Code:

worksize 256
aggression 13
threads_per_gpu 1
sharethreads 18
lookup_gap 2
gpu_thread_concurrency 8192

have tried alot of diffrent clock settings/thread conc, the best i managed around 500khash/s.

Any suggestions?

[meebs]

Stuck at around 300 khash/s on my 7970 atm

running factory clock settings, using reaper.

Code:

worksize 256
aggression 13
threads_per_gpu 1
sharethreads 18
lookup_gap 2
gpu_thread_concurrency 8192

have tried alot of diffrent clock settings/thread conc, the best i managed around 500khash/s.

Any suggestions?

woah there cowboy...

use these settings..


worksize 256
vectors 1
aggression 20
threads_per_gpu 1
sharethreads 28
lookup_gap 2
gpu_thread_concurrency 21712


note: you'll need about 2gb of system ram PER CARD using reaper

i'd personally recommend cgminer.

[jhd]

Cgminer requires less memory

[aa]

.

[paperkat]

Stuck at around 300 khash/s on my 7970 atm

running factory clock settings, using reaper.

Code:

worksize 256
aggression 13
threads_per_gpu 1
sharethreads 18
lookup_gap 2
gpu_thread_concurrency 8192

have tried alot of diffrent clock settings/thread conc, the best i managed around 500khash/s.

Any suggestions?

woah there cowboy...

use these settings..


worksize 256
vectors 1
aggression 20
threads_per_gpu 1
sharethreads 28
lookup_gap 2
gpu_thread_concurrency 21712


note: you'll need about 2gb of system ram PER CARD using reaper

i'd personally recommend cgminer.

620khashes now, yay:D

I couldnt get the cgminer to work before for some reason, but will give it another try!
thanks!

Stuck at around 300 khash/s on my 7970 atm

running factory clock settings, using reaper.

Code:

worksize 256
aggression 13
threads_per_gpu 1
sharethreads 18
lookup_gap 2
gpu_thread_concurrency 8192

have tried alot of diffrent clock settings/thread conc, the best i managed around 500khash/s.

Any suggestions?

1) Why are you using one year old code, that was broken back then, to mine with? Use cgminer.

2) Of course you're getting 300kh/s at aggression 13. Increase it to 18, or 20.

had no clue it was broken:O