7950 cg-miner setting for litecoin?

notaniel101 · November 8, 2013

Since the weather is getting cooler I am not as concerned with running up a massive cooling related bill from mining/folding. I had been mining for a couple months, but got tired of the changing hash rates, anywhere from 650-ish to 450 per car (running two). As soon as it would get the the 650 or relative peak, it would start falling down to the bottom. If you want a graph I'm sure I can find one and put it up. If it helps I was mining on give-me-ltc, now known as give-me-coins. I can also post the cg-miner setting I was using if it helps.

@RuecanOnRails

RuecanOnRails · November 8, 2013

The hash rate can fluctuate like that for a few reasons. One it could be your settings, such as cgminer trying to use to much vram (scrypt is very memory intensive). It could also be from lag to your mining pool, or the share difficulty was to low. They way it works is you request an easy to solve share based on the current block, you solve it, upload it then download the next one. If there is lag between that step you can see your hash rate dip as it starts a new share. By increasing the difficulty per share you can try to find a balance. Small fluctuations are normal, but going from 650 down to 450 is a big jump. You might be able to make a couple changes and aim for a stable 630 K/hash instead of 650, it might be enough to stabilize everything.

Are you running your cards in crossfire? If so try removing crossfire as it changes the memory usage of each card from independent to mirrored.

The first thing I'd look at is the settings you use for cgminer and the clock speeds.

https://litecoin.info/Mining_Hardware_Comparison

You can use that link to find hash rates others have achieved. You can try their various settings to see what sort of results you get.

Another thought would be your temps, If your card gets too hot it will auto throttle down and cause drops in hash. There a throttle built into the gpu's bios, but there's also a software one in Cgminer. I think it's defaulted to prevent the cards from going above 90, if you save the cgminer config you can open it with notepad and take a look then try to increase the limit.

If overheating is a factor, you can purchase some pcie risers and separate the cards for better airflow.

notaniel101 · November 10, 2013

The hash rate can fluctuate like that for a few reasons. One it could be your settings, such as cgminer trying to use to much vram (scrypt is very memory intensive). It could also be from lag to your mining pool, or the share difficulty was to low. They way it works is you request an easy to solve share based on the current block, you solve it, upload it then download the next one. If there is lag between that step you can see your hash rate dip as it starts a new share. By increasing the difficulty per share you can try to find a balance. Small fluctuations are normal, but going from 650 down to 450 is a big jump. You might be able to make a couple changes and aim for a stable 630 K/hash instead of 650, it might be enough to stabilize everything.

Are you running your cards in crossfire? If so try removing crossfire as it changes the memory usage of each card from independent to mirrored.

The first thing I'd look at is the settings you use for cgminer and the clock speeds.

https://litecoin.info/Mining_Hardware_Comparison

You can use that link to find hash rates others have achieved. You can try their various settings to see what sort of results you get.

Another thought would be your temps, If your card gets too hot it will auto throttle down and cause drops in hash. There a throttle built into the gpu's bios, but there's also a software one in Cgminer. I think it's defaulted to prevent the cards from going above 90, if you save the cgminer config you can open it with notepad and take a look then try to increase the limit.

If overheating is a factor, you can purchase some pcie risers and separate the cards for better airflow.

Thanks for that, I had the int set at 13 and the tc around 8000 because any higher and it wouldn't have enough buffer memory, but I finally got the set max alloc % 100 command to work and that helped, along with higher tc and int. Now it still dips, but not nearly as badly. The avg rate is above 1000 now, although with the settings it should be higher, but that might just come down to drivers.