Dual 7970's hashrate issues

[maursader]

Hey guys, I've just gotten these two new gigabyte 7970's and they're posting hash rates of around 600kh/s each. (See screenshot below)



I have configured other 7970's to be around the 720-750 mark in the past, but this is kind of underperforming for what I was expecting from the gigabytes.

In GPU-Z they show up as Radeon 7900 Series, and my GPU usage is 99% on both cards, not to mention the heat is around 86*C for GPU 0 and 78*C for GPU 1.

I have cleared the bin files as per this readme

cgminer says that my thread concurrency should be 22400 (as per what was selected and put in the bin filename)

Here's my .bat file

setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
cgminer --scrypt -o stratum+tcp://stratum.wemineftc.com:4444 -u maursader.1 -p x --shaders 2048 --intensity 13 --worksize 256 --thread-concurrency 22400 -g 1 --gpu-engine 1100 --gpu-memclock 1500 --auto-fan

Please let me know your thoughts, as this has been driving me up the wall! I removed the look up gap because I read that it sacrifices hashing power on the first GPU.

Cheers

Ps. For hardware, I've got a gigabyte am3 motherboard with the phenom x2 945 black edition, 12 GB DDR3, 750W antec power supply.. etc.

[FiiNALiZE]

Put your -g value up to 2.

This should net you at least 650+ on each card.

Also, put a fan by your rig. 86c is way to hot to run 24/7.

[maursader]

I'll give that a shot when I get home thanks!

Anyone else have any other suggestions? I will be home in an hour.

[maursader]

Put your -g value up to 2.

This should net you at least 650+ on each card.

Also, put a fan by your rig. 86c is way to hot to run 24/7.

Doing this reduced GPU 0 to 7-8kh/s and GPU 1 was unaffected.

Thank you for the suggestion though.

[FiiNALiZE]

Put your -g value up to 2.

This should net you at least 650+ on each card.

Also, put a fan by your rig. 86c is way to hot to run 24/7.

Doing this reduced GPU 0 to 7-8kh/s and GPU 1 was unaffected.

Thank you for the suggestion though.

Try lowering your engine and memclock by 100 each, keep -g 2

Post a screenshot of what happens afterwards.

[kaneda2004]

His thread concurrency may be too high to run -g 2, the 3gb cards run out of memory doing multiple threads with high thread concurrency.
My setup below:
setx GPU_MAX_ALLOC_PERCENT 100
color 02
cgminer --scrypt --api-listen --api-allow W:127.0.0.1 -o SERVER -u USER -p PASS --worksize 256 -g 2 -I 13 --lookup-gap 2 --shaders 2048 --thread-concurrency 19520 --gpu-fan 0-100 --auto-fan

and use MSI afterburner to setup 1085 core / 1500 mem clock.



                         

[maursader]

Put your -g value up to 2.

This should net you at least 650+ on each card.

Also, put a fan by your rig. 86c is way to hot to run 24/7.

Doing this reduced GPU 0 to 7-8kh/s and GPU 1 was unaffected.

Thank you for the suggestion though.

Try lowering your engine and memclock by 100 each, keep -g 2

Post a screenshot of what happens afterwards.

Current bat file settings:

color F0
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
cgminer --scrypt -o stratum+tcp://stratum.wemineftc.com:4444 -u maursader.1 -p x --intensity

13 --worksize 256 --thread-concurrency 22400 -g 2 --gpu-engine 1000 --gpu-memclock 1400 --auto-

fan

I had made the adjustments as per above.



I'm about to try kaneda's suggestions

[maursader]

His thread concurrency may be too high to run -g 2, the 3gb cards run out of memory doing multiple threads with high thread concurrency.
My setup below:
setx GPU_MAX_ALLOC_PERCENT 100
color 02
cgminer --scrypt --api-listen --api-allow W:127.0.0.1 -o SERVER -u USER -p PASS --worksize 256 -g 2 -I 13 --lookup-gap 2 --shaders 2048 --thread-concurrency 19520 --gpu-fan 0-100 --auto-fan

and use MSI afterburner to setup 1085 core / 1500 mem clock.



                         

Unfortunately, it did not side with me on this, the hash rates are still sub 600.



This has been a real head scratcher. haha

[tacotime]

Run this

Code:

cgminer --scrypt -o stratum+tcp://stratum.wemineftc.com:4444 -u maursader.1 -p x --intensity 13 --worksize 256 --thread-concurrency 8192 -g 2 --gpu-engine 1051 --gpu-memclock 1722 --auto-fan

[FiiNALiZE]

Run this

Code:

cgminer --scrypt -o stratum+tcp://stratum.wemineftc.com:4444 -u maursader.1 -p x --intensity 13 --worksize 256 --thread-concurrency 8192 -g 2 --gpu-engine 1051 --gpu-memclock 1722 --auto-fan

That memclock is crazy. I'm at 1100 core and 1320 mem and getting around 710kh/s.

Put your powertune to 5% at least in cgminer.

Also take out GPU_USE_SYNC (never seen that before) and --gpu-fan (it's redundant since you already have --auto-fan).

Put your thread-concurrency down to 8192 and see how it goes.

[superfastkyle]

--gpu-fan is not redundant auto fan only goes up to 85% without it

[jdebunt]

Here's my batfile :

setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1

cgminer -o stratum+tcp://stratum.wemineftc.com:4444 -u maursader.1 -p x --scrypt --gpu-engine 1075 --gpu-memclock 1485 --thread-concurrency 8192 -I 13 -g 2 -w 256 --lookup-gap 2 --no-submit-stale --auto-fan

742.2khash/sec , could push it to over 750 if i wanted to, but it's too hot here atm

Current temp : 74°

[FiiNALiZE]

--gpu-fan is not redundant auto fan only goes up to 85% without it

There's been research done and any fan speeds above 80% has no significant effect on temperature.

It only wears out the fan faster

[maursader]

Here's my batfile :

setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1

cgminer -o stratum+tcp://stratum.wemineftc.com:4444 -u maursader.1 -p x --scrypt --gpu-engine 1075 --gpu-memclock 1485 --thread-concurrency 8192 -I 13 -g 2 -w 256 --lookup-gap 2 --no-submit-stale --auto-fan

742.2khash/sec , could push it to over 750 if i wanted to, but it's too hot here atm

Current temp : 74°

Thank you for the suggestion, here are the results:

Run this

Code:

cgminer --scrypt -o stratum+tcp://stratum.wemineftc.com:4444 -u maursader.1 -p x --intensity 13 --worksize 256 --thread-concurrency 8192 -g 2 --gpu-engine 1051 --gpu-memclock 1722 --auto-fan

This caused the computer to lock up, lol.

I wonder if it's got anything to do with the thread concurrencies, and I only say this because when I run cgminer by itself, it tends to choose 22400 for the value.

Any thoughts on that?

[slarty]

Here's my batfile :

setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1

cgminer -o stratum+tcp://stratum.wemineftc.com:4444 -u maursader.1 -p x --scrypt --gpu-engine 1075 --gpu-memclock 1485 --thread-concurrency 8192 -I 13 -g 2 -w 256 --lookup-gap 2 --no-submit-stale --auto-fan

742.2khash/sec , could push it to over 750 if i wanted to, but it's too hot here atm

Current temp : 74°

Thank you for the suggestion, here are the results:

http://puu.sh/3CWB4.png

Run this

Code:

cgminer --scrypt -o stratum+tcp://stratum.wemineftc.com:4444 -u maursader.1 -p x --intensity 13 --worksize 256 --thread-concurrency 8192 -g 2 --gpu-engine 1051 --gpu-memclock 1722 --auto-fan

This caused the computer to lock up, lol.

I wonder if it's got anything to do with the thread concurrencies, and I only say this because when I run cgminer by itself, it tends to choose 22400 for the value.

Any thoughts on that?

I would suggest to set all the values for each card separated. even if they are the same cards with the same gpu it might be due to some manufacturing tolerances that they react different on the settings. Than you can try to find out which settings suits each card.
I had such an issue in the past with 2 x HD5770. they where in need of different settings in terms of CPU and memclock

RGDS

[maursader]

Just a thought, I am running a 750W antec power supply, do you suppose it is drawing too much power?