ATTN Litecoin GPU Miners - Scrypt support for cgminer

[nitrox]

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.

system RAM : 1GB ( free >512MB sfter system load )

[1715491705]

1715491705

[-ck]

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.

system RAM : 1GB

This could well be a limiting factor because it needs both system and gpu ram... Maybe try concurrency of 1792 but increase lookup gap from the default of 2.

[nitrox]

--thread-concurrency 1792  - work = 4.0Kh/s
--thread-concurrency 3584  - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
--thread-concurrency 5376  - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)

[penek]

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

Try this
--scrypt --worksize 256 --lookup-gap 2 --thread-concurrency 8192 -g 1 --intensity 16
--scrypt --worksize 256 --lookup-gap 2 --thread-concurrency 16364 -g 1 --intensity 16 (if buffer permit)
Result I don't promise

[-ck]

--thread-concurrency 1792  - work = 4.0Kh/s
--thread-concurrency 3584  - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
--thread-concurrency 5376  - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)

Clearly your low system ram is screwing you over. Are you setting an intensity with those? --thread-concurrency 1792 will safely take -I 12

[penek]

--thread-concurrency 1792  - work = 4.0Kh/s
--thread-concurrency 3584  - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
--thread-concurrency 5376  - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)

I have this error (in linux)
Resolved by "export GPU_USE_SYNC_OBJECTS=1" and "export GPU_MAX_ALLOC_PERCENT=100"
how to do it in windows I don't know

[nitrox]

I have this error (in linux)
Resolved by "export GPU_USE_SYNC_OBJECTS"
how to do it in windows I don't know

windows 7 x64
gpu driver 12.6
sdk 2.7

[penek]

I have this error (in linux)
Resolved by "export GPU_USE_SYNC_OBJECTS"
how to do it in windows I don't know

windows 7 x64
gpu driver 12.6
sdk 2.7

same drivers and sdk in linux x64

[-ck]

--thread-concurrency 1792  - work = 4.0Kh/s
--thread-concurrency 3584  - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
--thread-concurrency 5376  - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)

I have this error (in linux)
Resolved by "export GPU_USE_SYNC_OBJECTS=1" and "export GPU_MAX_ALLOC_PERCENT=100"
how to do it in windows I don't know

The only way to do it in windows is to reboot into linux.

[nitrox]

This could well be a limiting factor because it needs both system and gpu ram... Maybe try concurrency of 1792 but increase lookup gap from the default of 2.

--scrypt --shaders 1792 -g 1 --thread-concurrency 1792

and :

--lookup-gap 2 - work = 4.0Kh/s
--lookup-gap 3 - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
--lookup-gap 4 - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
--lookup-gap 5 - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
--lookup-gap 6 - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
--lookup-gap 7 - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)

[Tittiez]

Looks like there are new commits, I'll build them when I wake up tomorrow. Cheers

[nitrox]

Try this
--scrypt --worksize 256 --lookup-gap 2 --thread-concurrency 8192 -g 1 --intensity 16
--scrypt --worksize 256 --lookup-gap 2 --thread-concurrency 16364 -g 1 --intensity 16 (if buffer permit)
Result I don't promise

--scrypt --worksize 256 --lookup-gap 2 --thread-concurrency 8192 -g 1 --intensity 16
- crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)

--scrypt --worksize 256 --lookup-gap 2 --thread-concurrency 16364 -g 1 --intensity 16 (if buffer permit)
- crash = Error -61: clCreateBuffer (padbuffer8), decrease CT or increase LG

[penek]

This could well be a limiting factor because it needs both system and gpu ram... Maybe try concurrency of 1792 but increase lookup gap from the default of 2.

--scrypt --shaders 1792 -g 1 --thread-concurrency 1792

and :

--lookup-gap 2 - work = 4.0Kh/s
--lookup-gap 3 - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
--lookup-gap 4 - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
--lookup-gap 5 - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
--lookup-gap 6 - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
--lookup-gap 7 - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)

only --lookup-gap 2
and -g 1

[nitrox]

only --lookup-gap 2
and -g 1

crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)

[-ck]

This could well be a limiting factor because it needs both system and gpu ram... Maybe try concurrency of 1792 but increase lookup gap from the default of 2.

--scrypt --shaders 1792 -g 1 --thread-concurrency 1792

and :

--lookup-gap 2 - work = 4.0Kh/s
--lookup-gap 3 - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
--lookup-gap 4 - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
--lookup-gap 5 - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
--lookup-gap 6 - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
--lookup-gap 7 - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)

--scrypt -g 1 --thread-concurrency 896
Then try lookup gaps of 1,2,3,4 on top

...but I think you're screwed. Mine bitcoin or get more system ram.

[nitrox]

--scrypt -g 1 --thread-concurrency 896
Then try lookup gaps of 1,2,3,4 on top

...but I think you're screwed. Mine bitcoin or get more system ram.

how much total ram do you suggest

[-ck]

--scrypt -g 1 --thread-concurrency 896
Then try lookup gaps of 1,2,3,4 on top

...but I think you're screwed. Mine bitcoin or get more system ram.

how much total ram do you suggest

Note I am guessing this is the problem, I am not totally sure, so I really don't even know how much to recommend. I can mine on 4x7970s with 4GB of system ram at shaders of 2048 (you have less shaders on a 7950).

[penek]

This could well be a limiting factor because it needs both system and gpu ram... Maybe try concurrency of 1792 but increase lookup gap from the default of 2.

--scrypt --shaders 1792 -g 1 --thread-concurrency 1792

and :

--lookup-gap 2 - work = 4.0Kh/s
--lookup-gap 3 - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
--lookup-gap 4 - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
--lookup-gap 5 - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
--lookup-gap 6 - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
--lookup-gap 7 - crash = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)

only --lookup-gap 2
and -g 1

I came to the result start from:
--scrypt --worksize 64 --lookup-gap 2 -g 1 --thread-concurrency {number of shaders by card} --intensity 10
and increase "worksize" and multiplying number of shaders in different variants

Edit: may be you have problem with system

Edit 2: try to reinstall (delete and install new) drivers (from amd.com)

[nitrox]

I came to the result start from:
--scrypt --worksize 64 --lookup-gap 2 -g 1 --thread-concurrency {number of shaders by card} --intensity 10
and increase "worksize" and multiplying number of shaders in different variants
Edit: may be you have problem with system
Edit 2: try to reinstall (delete and install new) drivers (from amd.com)

--scrypt --worksize 64 --lookup-gap 2 -g 1 --thread-concurrency 1792 --intensity 10  - work = ~60kH/s

with increasing --thread-concurrency = Error -5: Enqueueing kernel onto command queue. (clEnqueueNDRangeKernel)
only whit --thread-concurrency 1792 work but poor

with increasing over --intensity :
on intensity 11 - Accepted 100% at 135kH/s
on intensity 12 - Accepted / Rejected = 50/50 at ~270kH/s  (target-miss)
on intensity 19 - Rejected 100%  at ~490kH/s  (target-miss)

with increasing worksize to 256 with intensity 11, thread-concurrency 1792 lookup-gap 2
Accepted / Rejected = 75 / 25 at 135kH/s  (target-miss)

[-ck]

Scrypt mining has now been merged into the master branch of cgminer, so you can switch to that and from now on it will be in the main release as well.