Sapphire 7990 second gpu runs much hotter

[esmadja]

I have a Sapphire 7990 running cgminer 3.7.2 with the following config
cgminer --scrypt -o http://us.wemineltc.com:80 -u user -p password --gpu-vddc 1.15 --auto-fan --auto-gpu -g 2 -I 12,11 --gpu-engine 600-1100,300-900 --gpu-memclock 1500,1000 --thread-concurrency 8192,8192 -w 256,256 --temp-target 75,75 --temp-overheat 80,80 --temp-cutoff 85,85 --temp-hysteresis 3

Couple of things I would appreciate help with
--gpu-vddc 1.15 doesn't seem to change the voltage notice gpu 0 is at 1.20 and gpu 1 is at 1.17
I am trying to reduce the voltage in order to reduce temp on gpu 1

More important is the fact that gpu 1 consistently seems to under perform i.e. it never mines at more than 300Kh/s compared to gpu 0 which has sometimes mined at around 450Kh/s and more at around 75C

The real issue is that gpu 1 will frequent heat up to 90C and while cgminer says it is reducing or even disabling gpu 1 it seems to stay at the nasty temp levels.

Is it possible that there is something wrong with the card or gpu 1?  I've heard people claim to hit over 1000Kh/s with stock fans.

I am running a Cooler Master haf case with Fan on the side pushing air in, fan on the front pushing air in and fan on the top pushing air out.

This is true even running at night with the window open the inside room temp is well under 60F
If I change -I 12,11 to -I 13,11 it increases gpu 0 Kh/s but will always send gpu 1 to over 90C???

Thanks for any help.
 
 cgminer version 3.7.2 - Started: [2013-12-14 09:50:09]
--------------------------------------------------------------------------------
 (5s):722.8K (avg):669.4Kh/s | A:50800  R:0  HW:0  WU:585.2/m
 ST: 2  SS: 0  NB: 38  LW: 1033  GF: 0  RF: 0
 Connected to us.wemineltc.com diff 208 with stratum as user myuser
 Block: bbf31f63...  Diff:2.93K  Started: [11:09:43]  Best share: 42.4K
--------------------------------------------------------------------------------
 [P]ool management [G]PU management (S)ettings [D]isplay options [Q]uit
 GPU 0:  74.0C 4037RPM | 427.5K/377.6Kh/s | A:29120 R:0 HW:0 WU: 333.9/m I:12
 GPU 1:  78.0C 3773RPM | 292.0K/292.0Kh/s | A:21680 R:0 HW:0 WU: 251.2/m I:11
--------------------------------------------------------------------------------

GPU 0: 303.3 / 378.8 Kh/s | A:129  R:0  HW:0  U:1.55/m  I:12
74.0 C  F: 85% (4035 RPM)  E: 1100 MHz  M: 1500 Mhz  V: 1.200V  A: 99%  P: 0%
Last initialised: [2013-12-14 09:50:09]
Intensity: 12
Thread 0: 151.3 Kh/s Enabled ALIVE
Thread 1: 151.9 Kh/s Enabled ALIVE

GPU 1: 287.0 / 292.2 Kh/s | A:101  R:0  HW:0  U:1.21/m  I:11
75.0 C  F: 85% (3767 RPM)  E: 750 MHz  M: 1000 Mhz  V: 1.170V  A: 99%  P: 0%
Last initialised: [2013-12-14 09:50:09]
Intensity: 11
Thread 2: 143.7 Kh/s Enabled ALIVE
Thread 3: 143.7 Kh/s Enabled ALIVE

[E]nable [D]isable ntensity [R]estart GPU [C]hange settings
Or press any other key to continue

[1714472981]

1714472981

[1714472981]

1714472981

[1714472981]

1714472981

[1714472981]

1714472981

[goxed]

If you are not aiming for phenomenal hash-rates

Try the following
a) download cgminer-2.11.2
b) engine clock=700, Memory=1230
c) Vdd=0.95V

should give 495KH/s per core and it will leave ur gpu much cooler.

[vm1990]

or use guiminer scrypt alpha and set each gpu core to its own setting. tune gpu 1 down a bit....
its outdated but still does pretty damn well

[esmadja]

Thanks for the feedback I'll try downgrading cgminer, can you explain why that will help... I am curious.
Also these changes seem to have helped alot though I honestly don't see why.

cgminer --scrypt -o http://us.wemineltc.com:80 -u emont69.1 -p 69system00 --auto-fan --auto-gpu -g 2 -I 14,14 --gpu-powertune 15,15 --gpu-engine 500-1100,500-1100 --gpu-memclock 1500,1000 --thread-concurrency 8192,8192 -w 256,256 --temp-target 77,77 --temp-overheat 83,83 --temp-cutoff 89,89 --temp-hysteresis 3

 cgminer version 3.7.2 - Started: [2013-12-14 15:15:29]
-------------------------------------------------------------------------------
 (5s):919.4K (avg):838.3Kh/s | A:12544  R:0  HW:408  WU:353.0/m
 ST: 2  SS: 0  NB: 16  LW: 455  GF: 0  RF: 0
 Connected to us.wemineltc.com diff 256 with stratum as user emont69.1
 Block: 5d511665...  Diff:2.93K  Started: [15:54:23]  Best share: 66.4K
-------------------------------------------------------------------------------
 [P]ool management [G]PU management (S)ettings [D]isplay options [Q]uit
 GPU 0:  79.0C 4028RPM | 383.3K/408.3Kh/s | A:8192 R:0 HW:181 WU: 197.8/m I:14
 GPU 1:  79.0C 3759RPM | 516.3K/431.9Kh/s | A:4352 R:0 HW:227 WU: 155.2/m I:14
-------------------------------------------------------------------------------

GPU 0: 399.0 / 409.3 Kh/s | A:30  R:0  HW:178  U:0.87/m  I:14
74.0 C  F: 84% (4032 RPM)  E: 715 MHz  M: 1500 Mhz  V: 1.170V  A: 100%  P: 15%
Last initialised: [2013-12-14 15:15:29]
Intensity: 14
Thread 0: 199.2 Kh/s Enabled ALIVE
Thread 1: 200.4 Kh/s Enabled ALIVE

GPU 1: 468.6 / 419.4 Kh/s | A:15  R:0  HW:224  U:0.43/m  I:14
79.0 C  F: 85% (3753 RPM)  E: 600 MHz  M: 1000 Mhz  V: 1.170V  A: 98%  P: 15%
Last initialised: [2013-12-14 15:15:29]
Intensity: 14
Thread 2: 236.0 Kh/s Enabled ALIVE
Thread 3: 238.1 Kh/s Enabled ALIVE

[E]nable [D]isable ntensity [R]estart GPU [C]hange settings
Or press any other key to continue

[jomay]

Thanks for the feedback I'll try downgrading cgminer, can you explain why that will help... I am curious.
Also these changes seem to have helped alot though I honestly don't see why.

cgminer --scrypt -o http://us.wemineltc.com:80 -u emont69.1 -p 69system00 --auto-fan --auto-gpu -g 2 -I 14,14 --gpu-powertune 15,15 --gpu-engine 500-1100,500-1100 --gpu-memclock 1500,1000 --thread-concurrency 8192,8192 -w 256,256 --temp-target 77,77 --temp-overheat 83,83 --temp-cutoff 89,89 --temp-hysteresis 3

 cgminer version 3.7.2 - Started: [2013-12-14 15:15:29]
-------------------------------------------------------------------------------
 (5s):919.4K (avg):838.3Kh/s | A:12544  R:0  HW:408  WU:353.0/m
 ST: 2  SS: 0  NB: 16  LW: 455  GF: 0  RF: 0
 Connected to us.wemineltc.com diff 256 with stratum as user emont69.1
 Block: 5d511665...  Diff:2.93K  Started: [15:54:23]  Best share: 66.4K
-------------------------------------------------------------------------------
 [P]ool management [G]PU management (S)ettings [D]isplay options [Q]uit
 GPU 0:  79.0C 4028RPM | 383.3K/408.3Kh/s | A:8192 R:0 HW:181 WU: 197.8/m I:14
 GPU 1:  79.0C 3759RPM | 516.3K/431.9Kh/s | A:4352 R:0 HW:227 WU: 155.2/m I:14
-------------------------------------------------------------------------------

GPU 0: 399.0 / 409.3 Kh/s | A:30  R:0  HW:178  U:0.87/m  I:14
74.0 C  F: 84% (4032 RPM)  E: 715 MHz  M: 1500 Mhz  V: 1.170V  A: 100%  P: 15%
Last initialised: [2013-12-14 15:15:29]
Intensity: 14
Thread 0: 199.2 Kh/s Enabled ALIVE
Thread 1: 200.4 Kh/s Enabled ALIVE

GPU 1: 468.6 / 419.4 Kh/s | A:15  R:0  HW:224  U:0.43/m  I:14
79.0 C  F: 85% (3753 RPM)  E: 600 MHz  M: 1000 Mhz  V: 1.170V  A: 98%  P: 15%
Last initialised: [2013-12-14 15:15:29]
Intensity: 14
Thread 2: 236.0 Kh/s Enabled ALIVE
Thread 3: 238.1 Kh/s Enabled ALIVE

[E]nable [D]isable ntensity [R]estart GPU [C]hange settings
Or press any other key to continue

gpu-powertune=15 will allow the cards to exceed the temperature/power limits by 15%. There are cards that will throttle themselves if they get heavy workload and powertune=0. One GPU might have throttled itself before and that's why you might see lower hashrates. 79 degrees also seems pretty hot to me. Can you leave your case open? Or maybe turn around the side fan so it sucks hot air out of the case? For example, I got 2x7950 running in a closed case(!) at <65 degree with <50% fan simply by optimizing the fan layout of the case.

The card itself is easily able to get to 2x550 up to 2x730kH/s, but temperature will be a problem in that case. I've got 2x7990 doing 4x630kH/s at <=72 degree celsius, but obviously not in a closed case and undervolting is a must. The settings to max out kH for mine are: cgminer --scrypt -I 13 -g 2 -w 256 --thread-concurrency 8192

Hope this helps.

Edit: I just saw that you are getting "HW" (hardware error) shares. This is not normal at all and either means that your settings are wrong or that your hardware is flaky. I believe intensity=14 might simply be too high. But maybe the high temperature causes hardware errors, I don't know. I've never had any.

[esmadja]

Thanks, I've been trying to under volt but --gpu-vddc doesn't seem to work on my sapphire 7990
I've read you need to use msi afterburner beta 3.0 build 17, is that what you are using.

I am now running at an average of 950Kh/s at about 79C on each core, which is progress.

[digicoin]

>> I am now running at an average of 950Kh/s at about 79C on each core, which is progress.
wonderful speed. I love it

[jomay]

Thanks, I've been trying to under volt but --gpu-vddc doesn't seem to work on my sapphire 7990
I've read you need to use msi afterburner beta 3.0 build 17, is that what you are using.

I am now running at an average of 950Kh/s at about 79C on each core, which is progress.

Which settings did you use now? Did you use the -I 13 etc line? BTW, pushing mem clock to 1500MHz doesn't have a big advantage for me, so you might want to live with 1000MHz on both GPUs.

My experience is that cgminer cannot tweak the voltage on most cards. Get MSI afterburner and unlock it to support voltage control. There are some guides on the net that will help you with that process, e.g. http://forums.overclockersclub.com/index.php?showtopic=182403.

[esmadja]

Newest settings are below doesn't seem quite as good as -I 12,12
Also I am using Afterburner to change voltage to 1.1, 1.1 however it doesn't always seem to stick sometimes the second gpu will go back to 1.17 volts???
If I keep pushing gpu 1 back to 1.1v the Kh/s for both gpu's goes up.

cgminer --scrypt -o http://us.wemineltc.com:80 -u user -p password --gpu-vddc 1.05,1.05 --auto-fan --auto-gpu -g 2 -I 13,13 --gpu-powertune 10,10 --gpu-engine 500-1100,500-1000 --gpu-memclock 1500,1100 --thread-concurrency 8192,8192 -w 256,256 --temp-target 75,75 --temp-overheat 83,83 --temp-cutoff 89,89 --temp-hysteresis 3

 cgminer version 3.7.2 - Started: [2013-12-16 19:02:08]
--------------------------------------------------------------------------------
 (5s):834.5K (avg):970.3Kh/s | A:5888  R:0  HW:0  WU:869.0/m
 ST: 2  SS: 0  NB: 1  LW: 63  GF: 0  RF: 0
 Connected to us.wemineltc.com diff 256 with stratum as user emont69.1
 Block: 4c2fb0e5...  Diff:3.09K  Started: [19:02:08]  Best share: 10.1K
--------------------------------------------------------------------------------
 [P]ool management [G]PU management (S)ettings [D]isplay options [Q]uit
 GPU 0:  76.0C 4118RPM | 326.1K/385.7Kh/s | A:1792 R:0 HW:0 WU: 347.5/m I:13
 GPU 1:  77.0C 3910RPM | 520.7K/595.3Kh/s | A:4096 R:0 HW:0 WU: 521.5/m I:13
--------------------------------------------------------------------------------

[jomay]

Newest settings are below doesn't seem quite as good as -I 12,12
Also I am using Afterburner to change voltage to 1.1, 1.1 however it doesn't always seem to stick sometimes the second gpu will go back to 1.17 volts???
If I keep pushing gpu 1 back to 1.1v the Kh/s for both gpu's goes up.

cgminer --scrypt -o http://us.wemineltc.com:80 -u user -p password --gpu-vddc 1.05,1.05 --auto-fan --auto-gpu -g 2 -I 13,13 --gpu-powertune 10,10 --gpu-engine 500-1100,500-1000 --gpu-memclock 1500,1100 --thread-concurrency 8192,8192 -w 256,256 --temp-target 75,75 --temp-overheat 83,83 --temp-cutoff 89,89 --temp-hysteresis 3

 cgminer version 3.7.2 - Started: [2013-12-16 19:02:08]
--------------------------------------------------------------------------------
 (5s):834.5K (avg):970.3Kh/s | A:5888  R:0  HW:0  WU:869.0/m
 ST: 2  SS: 0  NB: 1  LW: 63  GF: 0  RF: 0
 Connected to us.wemineltc.com diff 256 with stratum as user emont69.1
 Block: 4c2fb0e5...  Diff:3.09K  Started: [19:02:08]  Best share: 10.1K
--------------------------------------------------------------------------------
 [P]ool management [G]PU management (S)ettings [D]isplay options [Q]uit
 GPU 0:  76.0C 4118RPM | 326.1K/385.7Kh/s | A:1792 R:0 HW:0 WU: 347.5/m I:13
 GPU 1:  77.0C 3910RPM | 520.7K/595.3Kh/s | A:4096 R:0 HW:0 WU: 521.5/m I:13
--------------------------------------------------------------------------------

If the voltage jumps up in between it means in my experience that you don't undervolt properly. In other words, the automatic voltage regulation of the card kicks in and sets the voltage depending on your clock speed - if the clock goes up, voltage is automatically increased. Likewise you can easily lower the voltage if you simply change the engine clock to a low MHz number.

The trick is to get all the tools properly setup s.t. the voltage stays constantly low despite high clock.

Have you set up MSI Afterburner correctly? google "voltage unlock MSI Afterburner" for details. Also: did you install the drivers of your card manufacturer? That can also have an influence.

[esmadja]

I am using a Sapphire 7990

I followed the instructions for configuring msi afterburner;
1) edit msi config file
2) edit registry entries
3) reboot machine

I can change the voltage on the first gpu and it won't change, but not the other gpu.  I can't change the voltage on the second gpu 2 and have it remain constant.  In fact sometimes it seems I can't even change the voltage on the second gpu.

[jomay]

I am using a Sapphire 7990

I followed the instructions for configuring msi afterburner;
1) edit msi config file
2) edit registry entries
3) reboot machine

I can change the voltage on the first gpu and it won't change, but not the other gpu.  I can't change the voltage on the second gpu 2 and have it remain constant.  In fact sometimes it seems I can't even change the voltage on the second gpu.

Hmm, not sure I can help you much further, as I don't own a Sapphire and cards from different manufacturers tend to behave differently.

Did you check the box "force constant voltage" in the settings box , as well as "enabled voltage control" etc?

Regarding reading out voltage, I know that cgminer is completely unreliable and will not report the correct voltage. MSI Afterburner should be reliable, though. If you want definite accuracy and some more information on voltage, temps (also VRM temps) I'd use the the GPU-Z utility.