LTC mining with Reaper on a 7970

[philips]

Reaper reports higher hashrate compared with cgminer but keep this in mind:

Now finally, and you can believe me or not on this, but raper sends work to the GPU WITHOUT CHECKING if it was accepted, and gets the return buffer WITHOUT CHECKING if it actually did any work, and then adds the number of hashes it would have expected the GPU to do with that work sent to it. This means that when you start with lots of threads, some of them may not even be doing anything. Or if you've set some borderline invalid values, it will appear to be working fine, report back a big hashrate, but generate less valid shares. So I implore you to check the share rate generation and pretty much ignore the reported hashrate when comparing notes. Remember that cgminer AND raper use virtually identical kernels so should hash at virtually identical rates.

https://bitcointalk.org/index.php?topic=92887.msg1048114;topicseen#msg1048114

Best bet would be to squeeze (properly) as much performance as you can from both apps, mine with both for few hours and compare share rates as they are reported by the pool.

[1715204751]

1715204751

[live627]

I find reaper to be the best doin 760Kh/s on my 5970 

protocol litecoin

worksize 256
aggression 16
threads_per_gpu 1
sharethreads 10
lookup_gap 2
gpu_thread_concurrency 5632

zclock speeds?

[Liquid]

800Mhz

[Greedi]

cgminer is the best, and can give more then reaper, its have been proven

Just need to configure right

[Liquid]

I use it for BTC and PPC but i cannot get LTC to work it keeps saying rejected

[Greedi]

I use it for BTC and PPC but i cannot get LTC to work it keeps saying rejected

what is you're cgminer conf?

[Liquid]

well i dont have to configure for BTC & LTC and i get the best hashrate and best results without lag on my system cos i use my computer

but when i try to use a configured bat file for LTC with all the correct settings i get lower hashrate and lag 

[tacotime]

cgminer is the best, and can give more then reaper, its have been proven

Just need to configure right

using -shaders i get 550kh/s and 99% hardware errors

[Greedi]

cgminer is the best, and can give more then reaper, its have been proven

Just need to configure right

using -shaders i get 550kh/s and 99% hardware errors

what card?, and what --shader are you using?, and what else?

[tacotime]

cgminer is the best, and can give more then reaper, its have been proven

Just need to configure right

using -shaders i get 550kh/s and 99% hardware errors

what card?, and what --shader are you using?, and what else?

7950, --shader 1792 -w 256 -v 1 -g 1

[Greedi]

cgminer is the best, and can give more then reaper, its have been proven

Just need to configure right

using -shaders i get 550kh/s and 99% hardware errors

what card?, and what --shader are you using?, and what else?

7950, --shader 1792 -w 256 -v 1 -g 1

in what O/C ? if you'r doing that

also, you don't need -w and -v

[tacotime]

in what O/C ? if you'r doing that

also, you don't need -w and -v

1GHz core / 1450mhz RAM

if i don't set flags i lose performance

cgminer is setting the thread concurrency with --shader to ~8192, which is too small.  cgminer has a bug where it can't use thread concurrencies over 8192 (gives a memory size error).  in order to achieve max hashing rates for a 7xxx series card you need to set the concurrency to 64 * (memory bus size in bits).

in the meantime i'm getting 550kh/s in reaper and finding blocks

[live627]

800Mhz

May I also know the memory clock?

[Liquid]

1000Mhz

[Tomatocage]

1GHz core / 1450mhz RAM

if i don't set flags i lose performance

cgminer is setting the thread concurrency with --shader to ~8192, which is too small.  cgminer has a bug where it can't use thread concurrencies over 8192 (gives a memory size error).  in order to achieve max hashing rates for a 7xxx series card you need to set the concurrency to 64 * (memory bus size in bits).

in the meantime i'm getting 550kh/s in reaper and finding blocks

Interesting.  I'm getting the same error in Reaper as well when I go above 8192.

[tacotime]

Interesting.  I'm getting the same error in Reaper as well when I go above 8192.

Your thread concurrency needs to be a multiple of 64, but otherwise any thread_concurrency within the memory size should work.

[Tomatocage]

Interesting.  I'm getting the same error in Reaper as well when I go above 8192.

Your thread concurrency needs to be a multiple of 64, but otherwise any thread_concurrency within the memory size should work.

Thanks... but anything I try that yields a buffer size over 512MB doesn't work.  My card is supposed to have 3GB too :-/

[meebs]

Interesting.  I'm getting the same error in Reaper as well when I go above 8192.

Your thread concurrency needs to be a multiple of 64, but otherwise any thread_concurrency within the memory size should work.

Thanks... but anything I try that yields a buffer size over 512MB doesn't work.  My card is supposed to have 3GB too :-/

What is an example of something that didnt work for you?

The setup I run on my 7950 uses 1.5gb on each of my cards and 3gb of system ram.

Make sure you have lots of system ram ready available.

[Tomatocage]

Interesting.  I'm getting the same error in Reaper as well when I go above 8192.

Your thread concurrency needs to be a multiple of 64, but otherwise any thread_concurrency within the memory size should work.

Thanks... but anything I try that yields a buffer size over 512MB doesn't work.  My card is supposed to have 3GB too :-/

What is an example of something that didnt work for you?

The setup I run on my 7950 uses 1.5gb on each of my cards and 3gb of system ram.

Make sure you have lots of system ram ready available.

worksize 256
vectors 1
aggression 13
threads_per_gpu 4
sharethreads 32
lookup_gap 2
gpu_thread_concurrency 8192

I'm seeing only 2Gb out of 6Gb total system ram being used right now.  Anything beyond 8192 for gpu_thread_concurrency yields an error:

LTC buffer size: 1024MB.
Buffer too big: allocation failed. Either raise 'lookup_gap' or lower 'gpu_thread_concurrency'.

[tacotime]

change thread_per_gpu to 1