1BTC bounty -- increase 7970 hashrate on bamt/cgminer

[hashalfahalve]

Hi All,

Looking at the hardware wiki I saw people getting great hashrates with the 7970 with the lowest being 550 running on a windoze box and the average being ~650+.

I just got a shiny 7970 in the mail, and plugged it into bamt, installed the latest drivers and have it using cgminer instead of phoenix (phoenix wouldn't start at all).

It's mining with consistency at ~480mh.

1BTC to someone who can send me a pretty config file with optimal settings or point out something obvious I'm missing to get the hashrate up.

Yes, I'm a n00b.

Thanks!

[1714478751]

1714478751

[1714478751]

1714478751

[1714478751]

1714478751

[1714478751]

1714478751

[hashalfahalve]

And the winner of the million dollar, er, 1btc prize is af_newbie! Thank you sir! I currently have 0.86btc in my account, with about .2 pending, as soon as it hits 1 I'll send it all to you. 


With increasing gpu-engine, is my only concern maintaining adequate cooling? Or are there other dangers I should be looking out for with changing that number?

[hashalfahalve]

 cgminer -c cgm.conf

Does this config work for multiple cards? When I try it with two 7970 GPUs the computer shuts off completely. I think it might be an underrated PSU, but want to make sure I'm not missing anything obvious besides that.

[DeathAndTaxes]

 cgminer -c cgm.conf

Does this config work for multiple cards? When I try it with two 7970 GPUs the computer shuts off completely. I think it might be an underrated PSU, but want to make sure I'm not missing anything obvious besides that.

Yes you can specify per GPU values like this

Code:

"gpu-engine" : "1125, 1050, 1125, 1070",

With one value it applies that value to all cards.  

Either way works.

Rig powering off sounds like a power issue.

[hashalfahalve]

For some reason it looks like GPU0 is using the 1125 setting from the conf file, but GPU1 is set at 925.

[hashalfahalve]

Shoot, I rebooted and now they're both stuck at 925..

I'm on BAMT, is there a bamt/debian utility to manually change the clock?

[hashalfahalve]

Thanks, yes it seems to work though it's not very consistent. I set the max allowed clock and memory to 1200,1050

then in cgm.conf I set them at 1180,975 and it's running at 1180,1375

When it does run at 1200,1050 it seems to freeze up.

I'll have to keep playing around with it.

BTW I just sent your bounty many thanks for your continued help!

[DeathAndTaxes]

then in cgm.conf I set them at 1180,975 and it's running at 1180,1375

answer was already given

1180 - 150 = 1030.  975 is too low so it resets to stock.

Try to change them from within cgminer but make sure the diff between engine and memory is kept at 150 MHz  (memory lower by 150).  Bigger spread will not work as cgminer uses ADL APIs and those will not work with larger difference in engine to memory clocks.

cgminer also has a mem-diff parameter.  You can set it to 150, set no memclock and cgminer will always set memclock to core clock - memdiff.
I'll have to keep playing around with it.

[hashalfahalve]

I'm having an issue with the core clock winding down back to 925, do you have any thoughts on that? It will start running at the speeds I set, the memory clock doesn't change, but after a few minutes the core clock drops down to 925. I thought it might be the auto-gpu but the card isn't overheating, it's right at 66 degrees, and even when I take out auto-gpu it still drops down after a bit.

[hashalfahalve]

It's weird.

I have two cards, and one is running happily at 1200/1050 without issue.

The other one the max I can get it at is 1165/1015. If I push it to 1170 even, it works for a few minutes then the hash rate dials down and down then cgminer says it's "sick" and dies.

Same card same brand/company.


So you're saying even though they're identical cards, they have different OC properties?

[JWU42]

I need to try again - never did get it (bamt and 7970s) to work about 4 weeks ago so stuck with lubuntu.

[1l1l11ll1l]

Running BAMT, how can I see what memory clock my cards are at? Reading this thread I thinking they're probably not running at the rate I tell them to...

[JWU42]

7970 can run as low as engine less 150.

1050 Eng = 900 Mem
1125 Eng = 975 Mem

etc...

Hitting "G" in cgminer (which will be running in screen -- 'screen -r' will attach you) will show you all the details.

[-ck]

7970 can run as low as engine less 150.

1050 Eng = 900 Mem
1125 Eng = 975 Mem

etc...

Hitting "G" in cgminer (which will be running in screen -- 'screen -r' will attach you) will show you all the details.

Which you can do at startup with the memdiff option:
--gpu-memdiff -150

[JWU42]

I should do that 

[GernMiester]

Answers from CKolivas about cgminer... just listen, i bet he is right.....

[akipfer]

what about my xfx HD7950
let's say coreCLock is 900
what mem? why? ad i do i make the calculations

[Spendulus]

<Separate problem from priors>

I had 2 7970 running fine, added #3.  Now I see one card at right about 1/2 production, like 298 Mhash, the other 2 cards at 550-600 Mhash.

So what would cause a card to half run?

PS is 1200 Thermaltake, and voltages look fine.