ATTN Litecoin GPU Miners - Scrypt support for cgminer

[nitrox]

Your aggression seems like it's too high try this in cgminer or reaper
     worksize 64
     aggression 13
     threads_per_gpu 4 (or 2 or 1)
     sharethreads 18
     lookup_gap 2
     gpu_thread_concurrency 8192

105~130kH/s - going low

...... already thinking to drop mining LTC ......

[1714792526]

1714792526

[1714792526]

1714792526

[1714792526]

1714792526

[1714792526]

1714792526

[tacotime]

i try use reaper v13 beta 4 x64 with that settings :
worksize 64
aggression 20
threads_per_gpu 1
lookup_gap 5
gpu_thread_concurrency 8192

its work , but only 240~220 kH/s for HD7950OC i think is too little.
my old HD6950 reached ~360 kH/s

but i stil lito to use cgminer
i try same worksize , thread_concurrency and lookup_gap but again fail.

Use cgminer and these settings

--scrypt -I 19 --shaders 1792 --thread-concurrency 8960 -g 1

Should get 500+ kh/s

also try --thread-concurrency of 17920, may be a little faster

for the 7xxx series you must always use -g 1 to achieve high intensities

[nitrox]

Use cgminer and these settings
--scrypt -I 19 --shaders 1792 --thread-concurrency 8960 -g 1
Should get 500+ kh/s
also try --thread-concurrency of 17920, may be a little faster
for the 7xxx series you must always use -g 1 to achieve high intensities

----------------------------------------------------------------------------------------
 [2012-07-28 22:42:05] Started cgminer 2.5.0
 [2012-07-28 22:42:06] Probing for an alive pool
 [2012-07-28 22:42:06] Long-polling activated for http://ltc.xurious.com:9332/LP
 [2012-07-28 22:42:28] Maximum buffer memory device 0 supports says 536870912, your scrypt settings come to 0
 [2012-07-28 22:42:28] Error -61: clCreateBuffer (padbuffer8), decrease CT or increase LG
 [2012-07-28 22:42:28] Failed to init GPU thread 0, disabling device 0
 [2012-07-28 22:42:28] Restarting the GPU from the menu will not fix this.
 [2012-07-28 22:42:28] Try restarting cgminer.
Press enter to continue:
----------------------------------------------------------------------------------------
same for --thread-concurrency 8960 and 17920  

[tacotime]

Use cgminer and these settings
--scrypt -I 19 --shaders 1792 --thread-concurrency 8960 -g 1
Should get 500+ kh/s
also try --thread-concurrency of 17920, may be a little faster
for the 7xxx series you must always use -g 1 to achieve high intensities

----------------------------------------------------------------------------------------
[2012-07-28 22:42:05] Started cgminer 2.5.0
 [2012-07-28 22:42:06] Probing for an alive pool
 [2012-07-28 22:42:06] Long-polling activated for http://ltc.xurious.com:9332/LP
 [2012-07-28 22:42:28] Maximum buffer memory device 0 supports says 536870912, your scrypt settings come to 0
 [2012-07-28 22:42:28] Error -61: clCreateBuffer (padbuffer8), decrease CT or increase LG
 [2012-07-28 22:42:28] Failed to init GPU thread 0, disabling device 0
 [2012-07-28 22:42:28] Restarting the GPU from the menu will not fix this.
 [2012-07-28 22:42:28] Try restarting cgminer.
Press enter to continue:
----------------------------------------------------------------------------------------
same for --thread-concurrency 8960 and 17920  

Hum, that's strange, that's 5x the number of SPUs which should work.  You can try the 7970 setting of --thread-concurrency 10240, I don't know why that wouldn't work.

That is the normal error for incorrect buffer sizes though.  I wonder why it doesn't work?

Also --lookup-gap should always be 2, but that's the default

[nitrox]

same whit --thread-concurrency 10240 ....

[tacotime]

same whit --thread-concurrency 10240 ....

try --shaders 2048 and --thread-concurrency 10240

I'm starting to think something is wrong with your cgminer build though, these should not give 0 memory values

[nitrox]

try --shaders 2048 and --thread-concurrency 10240
I'm starting to think something is wrong with your cgminer build though, these should not give 0 memory values

--------------------------------------------------------------------------------
 [2012-07-28 22:57:09] Started cgminer 2.5.0
 [2012-07-28 22:57:09] Probing for an alive pool
 [2012-07-28 22:57:10] Long-polling activated for http://ltc.xurious.com:9332/LP
 [2012-07-28 22:57:23] LONGPOLL from pool 0 detected new block
 [2012-07-28 22:57:33] Pool 0 http://ltc.xurious.com:9332 alive
 [2012-07-28 22:57:33] Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
 [2012-07-28 22:57:33] GPU 0 failure, disabling!
 [2012-07-28 22:57:33] Thread 0 being disabled
--------------------------------------------------------------------------------

[dishwara]

Ok I found my issue, cgminer.conf was conflicting and it overrides the command line settings! So check your config file and make sure things like work size and intensity etc match what youre attempting on the command line. Currently cranking along at 203kh/s on my 5770 w/intensity 18, working on tuning the 6970 still it overheats way to easy and forces the machine to reboot.

I don't see cgminer.conf in the unziped files on windows.
Can point me where i can see cgminer.conf.
Thanks.

[tacotime]

try --shaders 2048 and --thread-concurrency 10240
I'm starting to think something is wrong with your cgminer build though, these should not give 0 memory values

--------------------------------------------------------------------------------
 [2012-07-28 22:57:09] Started cgminer 2.5.0
 [2012-07-28 22:57:09] Probing for an alive pool
 [2012-07-28 22:57:10] Long-polling activated for http://ltc.xurious.com:9332/LP
 [2012-07-28 22:57:23] LONGPOLL from pool 0 detected new block
 [2012-07-28 22:57:33] Pool 0 http://ltc.xurious.com:9332 alive
 [2012-07-28 22:57:33] Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
 [2012-07-28 22:57:33] GPU 0 failure, disabling!
 [2012-07-28 22:57:33] Thread 0 being disabled
--------------------------------------------------------------------------------

Yeah, that's what I expected...  with reaper and --thread-concurrency of 7168, 8960, or 16000 and number of threads 1 (intensity 19) or 4/5 (intensity 13) should give you at least 400kh/s with a lookup gap of 2 and a worksize of 128 or 256

[nitrox]

I'm starting to think something is wrong with your cgminer build though, these should not give 0 memory values

i use windows build "ckolivas-cgminer-43752ee" i think is newest at the moment

[nitrox]

Yeah, that's what I expected...  with reaper and --thread-concurrency of 7168, 8960, or 16000 and number of threads 1 (intensity 19) or 4/5 (intensity 13) should give you at least 400kh/s with a lookup gap of 2 and a worksize of 128 or 256

reaper chrash, chrash, chrash, chrash, chrash, chrash ......

whit :
---------------------------------
worksize 256
aggression 19
threads_per_gpu 1
lookup_gap 5
gpu_thread_concurrency 7168
---------------------------------
run normal but < 250kH/s

[tacotime]

Yeah, that's what I expected...  with reaper and --thread-concurrency of 7168, 8960, or 16000 and number of threads 1 (intensity 19) or 4/5 (intensity 13) should give you at least 400kh/s with a lookup gap of 2 and a worksize of 128 or 256

reaper chrash, chrash, chrash, chrash, chrash, chrash ......

whit :
---------------------------------
worksize 256
aggression 19
threads_per_gpu 1
lookup_gap 5
gpu_thread_concurrency 7168
---------------------------------
run normal but < 250kH/s

something is very wrong, lookup_gap 2 should work fine.  might want to try uninstalling 12.6 drivers and reinstalling them, and also uninstalling app sdk 2.6 and installing 2.7.  other than that i do not know, something is broken.

[nitrox]

something is very wrong, lookup_gap 2 should work fine.  might want to try uninstalling 12.6 drivers and reinstalling them, and also uninstalling app sdk 2.6 and installing 2.7.  other than that i do not know, something is broken.

exactly the same result crash, crash, crash, crash, crash, crash, crash ...

but BTC digging well ~ 530MH/s

[rahlquist]

Ok I found my issue, cgminer.conf was conflicting and it overrides the command line settings! So check your config file and make sure things like work size and intensity etc match what youre attempting on the command line. Currently cranking along at 203kh/s on my 5770 w/intensity 18, working on tuning the 6970 still it overheats way to easy and forces the machine to reboot.

I don't see cgminer.conf in the unziped files on windows.
Can point me where i can see cgminer.conf.
Thanks.

If you have one it will be in the same directory as the cgminer.exe most likely. If you dont have one then this isnt whats holding you back.

[rahlquist]

something is very wrong, lookup_gap 2 should work fine.  might want to try uninstalling 12.6 drivers and reinstalling them, and also uninstalling app sdk 2.6 and installing 2.7.  other than that i do not know, something is broken.

exactly the same result crash, crash, crash, crash, crash, crash, crash ...

but BTC digging well ~ 530MH/s

Ok its a stupid question but I havent seen it asked. You are shutting down BTC when testing for LTC right?

[nitrox]

Ok its a stupid question but I havent seen it asked. You are shutting down BTC when testing for LTC right?

of course, also restart PC when i change digging currency.

[nitrox]

Come on guys, if not you no one can't help, give ideas   

[-ck]

Come on guys, if not you no one can't help, give ideas   

Start it with just your login details and: --shaders 1792 -T -D --verbose -g 1
and see what it reports.

[nitrox]

Come on guys, if not you no one can't help, give ideas    

Start it with just your login details and: --shaders 1792 -T -D --verbose -g 1
and see what it reports.

here is log file:
http://www.uploadmb.com/dw.php?id=1343553631

[-ck]

Come on guys, if not you no one can't help, give ideas    

Start it with just your login details and: --shaders 1792 -T -D --verbose -g 1
and see what it reports.

here is log file:
http://www.uploadmb.com/dw.php?id=1343553631

Everything looks in order there, but it's still failing. How much system ram do you have? Try forcing smaller multiples of 1792 for thread concurrency. Eg --thread-concurrency 1792 , then 3584, then 5376.