How to reduce Hardware Errors

[vamsibannu]

Hello Miners,

I recently built a rig with 3 sapphire vapor-x HD 7950 graphics cards on windows 8 pro OS and i am using air cooling with 8 cooling fans and i am able to recieve about 1.4 MH/s all run together at I = 15

My CGMiner configuration file is as below :

{
"pools" : [
   {
      "name" : "Goldcoin Hashfaster (gld.hashfaster.com)",
      "url" : "stratum+tcp://gld.hashfaster.com:3336",
      "user" : "*****.****",
      "pass" : "***"
   }
],
"api-allow" : "W:127.0.0.1",
"api-listen" : true,
"expiry" : "1",
"failover-only" : true,
"queue" : "0",
"round-robin" : true,
"scrypt" : true,
"intensity" : "15",
"worksize" : "256",
"no-submit-stale": true,
"scan-time": "1"
}

My .BAT batch file code is as below :

color 07
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
cgminer --scrypt  --auto-fan -o stratum+tcp://gld.hashfaster.com:3336 -u *******.**** -p **** --thread-concurrency 21712 --lookup-gap 2 --intensity 15 --worksize 256 -g 2

I have attached the CGMiner screenshot.
My problem is though i am mining at about 1.4MH/s, in my pool i could see only about 175 KH/s being submitted.
I studied that this may be due to too may Hardware errors (HW = high) being generated.

I would be happy if anyone can help me resolve this issue.

Regards,
Vamsi.

[mrtsufo]

Are you matching your card to these suggested settings?

https://litecoin.info/Mining_Hardware_Comparison

Also, with regards to your other hardware: I found on my side that I got greater stability upon removing a stick of RAM that was different from the other suggested stick.  It wasn't a matched pair. Little things sometimes.  Good luck!

[dustint]

try to drop your intensity down to 12 or 13

[vamsibannu]

I did try changing intensity from 12 to 18. I was able to achieve about 400 to 450 KH/s per card without overclocking, but the problem with high number of Hardware Errors still remains the same and hence i could see that the hashrate in my pool is only about 15% of what my miner shows.

[manto]

Try chaning the thread concurrency. Look on github for recommended value for a 7950.

[AZIZ1977]

try worksize 256 intensity 13 thread-concurrency 8192

[pawl]

Ive found that -i 13 is the highest you can go to when using -g 2. You can increase intensity beyond 13 but u will have to do it using only one GPU thread, ie -g 1

you can start cgminer with -thread concurrency set to 0, that will force cgminer to find the best TC that your card can handle. you can tell what value it finds by looking at the .bin file in your cgminer directory should look a bit like this "scrypt130511Tahitiglg2tc10240w256l4.bin" with "...tc10240.." being the max value cgminer could run my Asus7970 at.

I would try aziz settings are they are pretty much default for 7 series cards. If cgminer runs with -w 256 -i 13 -g 2 -tc 8192 and ure still getting HW errors then make sure you havent clocked your cards too high. I find my Asus 7970 starts to [beep] the bed at around 1875mhz on the memory.

Once u have zero HW errors, then u can start to tweak your core/memory to find that sweet spot. by sweet spot i mean the point where u get a fast hash rate for least power. My 7970 seems to like 1080/1500 with it running on minimal volts for 740kh/s, if i go higher then i have to start upping voltate and tbh its only going from 740 to 760ish, not really worth the massive wattage jump.