3x5830 1 runs about 100Kh/s slower than the other 2

[Nolo]

I just set this rig up last night.  Here's a screen shot:

cgminer version 2.8.5 - Started: [2012-10-27 08:05:39]
--------------------------------------------------------------------------------
 (5s):594.6K (avg):542.3Kh/s | Q:9  A:73  R:0  HW:0  E:811%  U:34.4/m
 TQ: 0  ST: 4  SS: 0  DW: 2  NB: 1  LW: 10  GF: 1  RF: 7  WU: 34.4
 Connected to http://coinotron.com:8322 with LP as user Nolo.1
 Block: c3acb8e2fc5380108c4fd3791364d80a...  Started: [08:05:39]
--------------------------------------------------------------------------------
 [P]ool management [G]PU management [ S]ettings [D]isplay options [Q]uit
 GPU 0:  68.0C 2761RPM | 275.0K/266.0Kh/s | A:32 R:0 HW:0 U:15.10/m I:18
 GPU 1:  77.0C 4108RPM | SICK / 0.000h/s | A: 0 R:0 HW:0 U: 0.00/m I:18
 GPU 2:  38.0C 2997RPM | 275.2K/276.3Kh/s | A:41 R:0 HW:0 U:19.35/m I:18
--------------------------------------------------------------------------------

 [2012-10-27 08:07:28] Accepted 0000084f Diff 30/16 GPU 0
 [2012-10-27 08:07:33] Accepted 00000e7c Diff 17/16 GPU 2
 [2012-10-27 08:07:33] Accepted 00000563 Diff 47/16 GPU 2
 [2012-10-27 08:07:34] Accepted 0000011b Diff 231/16 GPU 2
 [2012-10-27 08:07:37] Accepted 00000080 Diff 511/16 GPU 0
 [2012-10-27 08:07:39] Stale share detected, submitting as user requested
 [2012-10-27 08:07:39] Accepted 000001fe Diff 128/16 GPU 2
 [2012-10-27 08:07:39] Stale share detected, submitting as user requested
 [2012-10-27 08:07:39] Accepted 00000173 Diff 176/16 GPU 0
 [2012-10-27 08:07:42] Accepted 00000f3b Diff 16/16 GPU 0
 [2012-10-27 08:07:44] Accepted 000007b3 Diff 33/16 GPU 0

Here's the command I'm running to start it:

cgminer --scrypt -o http://xxxxxxx -u user -p pass --thread-concurrency 5600 -I 18 -g 1 -w 256

GPU 1 runs consistently 100Kh/s slower than GPU 0 & GPU 2.  That's assuming it even starts, which sometimes it doesn't.  Sometimes it sits at 0 while the other 2 mine away where they should be.  
I've tried running all 3 in crossfire, and not using the crossfire cables.  (By the way I can't tell that this makes a difference, does it matter?)

Also while in this screengrab it says that GPU 1 is SICK, it's temps shoot way up which tells me its atleast trying.  

Suggestions?  



And here it is after I reset it a couple of times and GPU 1 started:

cgminer version 2.8.5 - Started: [2012-10-27 08:20:16]
--------------------------------------------------------------------------------
 (5s):752.4K (avg):720.1Kh/s | Q:7  A:51  R:0  HW:0  E:729%  U:41.0/m
 TQ: 0  ST: 4  SS: 0  DW: 0  NB: 1  LW: 3  GF: 1  RF: 0  WU: 42.6
 Connected to http://coinotron.com:8322 with LP as user Nolo.1
 Block: c3acb8e2fc5380108c4fd3791364d80a...  Started: [08:20:16]
--------------------------------------------------------------------------------
 [P]ool management [G]PU management [ S]ettings [D]isplay options [Q]uit
 GPU 0:  61.5C 2685RPM | 273.5K/281.0Kh/s | A:23 R:0 HW:0 U:18.49/m I:18
 GPU 1:  69.0C 4019RPM | 178.9K/182.7Kh/s | A:11 R:0 HW:0 U: 8.84/m I:18
 GPU 2:  55.5C 2913RPM | 273.2K/277.5Kh/s | A:19 R:0 HW:0 U:15.28/m I:18
--------------------------------------------------------------------------------

 [2012-10-27 08:21:18] Accepted 00000b0e Diff 23/16 GPU 0
 [2012-10-27 08:21:19] Accepted 0000037e Diff 73/16 GPU 2
 [2012-10-27 08:21:19] Accepted 000000aa Diff 385/16 GPU 0
 [2012-10-27 08:21:20] Accepted 00000378 Diff 73/16 GPU 0
 [2012-10-27 08:21:21] Accepted 00000307 Diff 84/16 GPU 1
 [2012-10-27 08:21:22] Accepted 00000bfd Diff 21/16 GPU 2
 [2012-10-27 08:21:24] Accepted 000009cb Diff 26/16 GPU 2
 [2012-10-27 08:21:29] Accepted 000003c7 Diff 67/16 GPU 2
 [2012-10-27 08:21:31] Accepted 00000f59 Diff 16/16 GPU 0
 [2012-10-27 08:21:31] Accepted 00000a95 Diff 24/16 GPU 2
 [2012-10-27 08:21:32] Accepted 000007ff Diff 32/16 GPU 1

Notice about 100Kh/s slower than 0 & 2

[Nolo]

Ok, I just figured out what's causing the drop.  It doesn't appear to have anything to do with the cards themselves.  

If I plug my monitor into GPU 0, GPU 0 loses about 100Kh/s, GPU's 1 & 2 run fine.
If I plug my monitor into GPU 1, GPU 1 loses about 100Kh/s, GPU's 0 & 2 run fine.
If I plug my monitor into GPU 2, GPU 2 loses about 100Kh/s, GPU's 0 & 1 run fine.

So the issue is obviously the display is causing the reduction.  The question is now, how do I fix this?  Running windows 8 x64.


Edit:  See post below.  I apparently was wrong. 

[tacotime]

Try installing logmein and accessing the computer remotely using the logmein display driver, if you're on windows anyway

[SaintDevil]

Try installing logmein and accessing the computer remotely using the logmein display driver, if you're on windows anyway

I love Logmein interface, but I found that is it loading slow on slow networks.
I use TightVNC, not sure about it's compatibility with win 8. 

[Nolo]

Try installing logmein and accessing the computer remotely using the logmein display driver, if you're on windows anyway

Well apparently I was completely wrong.  I installed logmein, disconnected all dispalys to the pc, rebooted...logged in from my laptop...fired up cgminer, and the same exact problem. GPU 1 100Kh/s slower.  Maybe when I was switching the card that the display was in, it was renumbering the GPU's?  I have no idea.  I just know I'm still getting the same problem with no display.

[SaintDevil]

Try installing logmein and accessing the computer remotely using the logmein display driver, if you're on windows anyway

Well apparently I was completely wrong.  I installed logmein, disconnected all dispalys to the pc, rebooted...logged in from my laptop...fired up cgminer, and the same exact problem. GPU 1 100Kh/s slower.  Maybe when I was switching the card that the display was in, it was renumbering the GPU's?  I have no idea.  I just know I'm still getting the same problem with no display.

I thought what if it is logmein, just installed it on my pc and doesn't make a difference.
What if you try running it in a separate instance of cgminer, or try reaper and see if it gives you the same results.
I have had one of them wouldn't start at all, but that's because I didn't have enough ram.

[Nolo]

Ok new information.  I unplugged two of the other cards.  I figured out it is the first card.  When I run it by itself this is what i get:

cgminer version 2.8.5 - Started: [2012-10-27 11:33:32]
--------------------------------------------------------------------------------
 (5s):178.9K (avg):183.4Kh/s | Q:1  A:15  R:0  HW:0  E:1500%  U:10.7/m
 TQ: 0  ST: 2  SS: 0  DW: 0  NB: 2  LW: 8  GF: 0  RF: 0  WU: 10.7
 Connected to http://coinotron.com:8322 with LP as user Nolo.1
 Block: b1a2ae8f4044d8b60c5d08f5e3a5d8d7...  Started: [11:34:19]
--------------------------------------------------------------------------------
 [P]ool management [G]PU management [ S]ettings [D]isplay options [Q]uit
 GPU 0:  65.0C 3137RPM | 177.4K/183.4Kh/s | A:15 R:0 HW:0 U:10.67/m I:18
--------------------------------------------------------------------------------

 [2012-10-27 11:33:45] Accepted 00000410 Diff 63/16 GPU 0
 [2012-10-27 11:33:45] Accepted 0000076b Diff 34/16 GPU 0
 [2012-10-27 11:34:02] Accepted 00000aaf Diff 23/16 GPU 0
 [2012-10-27 11:34:12] Accepted 00000ee3 Diff 17/16 GPU 0
 [2012-10-27 11:34:15] Accepted 000007b0 Diff 33/16 GPU 0
 [2012-10-27 11:34:16] Accepted 00000ee5 Diff 17/16 GPU 0
 [2012-10-27 11:34:19] LONGPOLL from pool 0 detected new block
 [2012-10-27 11:34:24] Accepted 00000e5d Diff 17/16 GPU 0
 [2012-10-27 11:34:30] Accepted 00000123 Diff 225/16 GPU 0
 [2012-10-27 11:34:30] Accepted 00000414 Diff 62/16 GPU 0
 [2012-10-27 11:34:35] Accepted 00000406 Diff 63/16 GPU 0
 [2012-10-27 11:34:43] Accepted 00000fbe Diff 16/16 GPU 0
 [2012-10-27 11:34:50] Accepted 000000cc Diff 321/16 GPU 0

Thoughts?

Try installing logmein and accessing the computer remotely using the logmein display driver, if you're on windows anyway

Well apparently I was completely wrong.  I installed logmein, disconnected all dispalys to the pc, rebooted...logged in from my laptop...fired up cgminer, and the same exact problem. GPU 1 100Kh/s slower.  Maybe when I was switching the card that the display was in, it was renumbering the GPU's?  I have no idea.  I just know I'm still getting the same problem with no display.

I thought what if it is logmein, just installed it on my pc and doesn't make a difference.
What if you try running it in a separate instance of cgminer, or try reaper and see if it gives you the same results.
I have had one of them wouldn't start at all, but that's because I didn't have enough ram.

I have 6 gigs of RAM on this machine.  

I'm only getting about 140k in Reaper on this card

[tacotime]

my guess is that the card is throttling the core and memory speed by the bios somehow.  is it running at the same speeds as the other ones in afterburner or gpu-z?

[Nolo]

my guess is that the card is throttling the core and memory speed by the bios somehow.  is it running at the same speeds as the other ones in afterburner or gpu-z?

All the cards are the same when running according to gpu-z except:
I just noticed that in Catalyst Control Center, the "Activity" for 2 of the cards are at 99%, while the Activity for 1 of them is at around 67-70%. 

What's causing that?

[tacotime]

that's weird.  does it do this for non-mining applications such as furmark?  it sounds like the card may be throttling inappropriately (either something is wrong with the BIOS or one of the thermal diodes is broken and is reporting incorrect values).
i would try flashing the bios with one of the ones from the other cards.  you can get the bios for one of the functional cards from gpu-z.

[SaintDevil]

my guess is that the card is throttling the core and memory speed by the bios somehow.  is it running at the same speeds as the other ones in afterburner or gpu-z?

All the cards are the same when running according to gpu-z except:
I just noticed that in Catalyst Control Center, the "Activity" for 2 of the cards are at 99%, while the Activity for 1 of them is at around 67-70%. 

What's causing that?

Something else is using the graphics card at that moment.
I get something like that when other programs trying to use it, and my hashrate goes down. I did notice that when I move my mouse fast the usage goes down too. It goes back up again if I disconnect.


 

[Bigal]

Have you tried disabling ULPS?

[Nolo]

Ok after all damn day of working on this, I think I have some more information.

I've been getting a Video TDR Bluescreen.  That seems to indicate a lack of power.

I have a Kingwin 800W Mach 1 PSU.  http://www.newegg.com/Product/Product.aspx?Item=N82E16817121026

4xHDD
1xSSD
1 DVDRW
i7 920
EVGA Classified e760
Several case fans.
6 Gigs RAM
3x 5830

Nothing is overclocked.

Is this PSU up to the task?  

Have you tried disabling ULPS?

I've tried disabling ULPS by setting all EnableULPS in the registry to 0.  A program I have says its disabled, but it still underclocks all the cards when I view them in Catalyst.

that's weird.  does it do this for non-mining applications such as furmark?  it sounds like the card may be throttling inappropriately (either something is wrong with the BIOS or one of the thermal diodes is broken and is reporting incorrect values).
i would try flashing the bios with one of the ones from the other cards.  you can get the bios for one of the functional cards from gpu-z.

No problem in Furmark all three run great there...all my cards report the same bios. 

[SaintDevil]

Ok after all damn day of working on this, I think I have some more information.

I've been getting a Video TDR Bluescreen.  That seems to indicate a lack of power.

I have a Kingwin 800W Mach 1 PSU.  http://www.newegg.com/Product/Product.aspx?Item=N82E16817121026

4xHDD
1xSSD
1 DVDRW
i7 920
EVGA Classified e760
Several case fans.
6 Gigs RAM
3x 5830

Nothing is overclocked.

Is this PSU up to the task?  

1 5830 consumes around 300 watt. system overload O_o? 

[Nolo]

Ok after all damn day of working on this, I think I have some more information.

I've been getting a Video TDR Bluescreen.  That seems to indicate a lack of power.

I have a Kingwin 800W Mach 1 PSU.  http://www.newegg.com/Product/Product.aspx?Item=N82E16817121026

4xHDD
1xSSD
1 DVDRW
i7 920
EVGA Classified e760
Several case fans.
6 Gigs RAM
3x 5830

Nothing is overclocked.

Is this PSU up to the task?  

1 5830 consumes around 300 watt. system overload O_o?  

Well isn't that just jim damn dandy.  Guess I'll start looking for one    1000W?

[SaintDevil]

Well isn't that just jim damn dandy.  Guess I'll start looking for one    1000W?

But just to make sure before you buy a new one, test 2 cards at a time, then switch one for the 3rd.
And then if mhs stays same definitely not enough power.

But with all your other components running along, I would look into 1200 

[Bigal]

Ok after all damn day of working on this, I think I have some more information.

I've been getting a Video TDR Bluescreen.  That seems to indicate a lack of power.

I have a Kingwin 800W Mach 1 PSU.  http://www.newegg.com/Product/Product.aspx?Item=N82E16817121026

4xHDD
1xSSD
1 DVDRW
i7 920
EVGA Classified e760
Several case fans.
6 Gigs RAM
3x 5830

Nothing is overclocked.

Is this PSU up to the task?  

1 5830 consumes around 300 watt. system overload O_o?  

Well isn't that just jim damn dandy.  Guess I'll start looking for one    1000W?

I would think the power supply would be sufficient at stock clocks and volts, I think it's more likely a driver issue (what version are you using?) with monitors not being connected to the cards, might need a couple dummy plugs.

[vitruvio]

my guess is that the card is throttling the core and memory speed by the bios somehow.  is it running at the same speeds as the other ones in afterburner or gpu-z?

All the cards are the same when running according to gpu-z except:
I just noticed that in Catalyst Control Center, the "Activity" for 2 of the cards are at 99%, while the Activity for 1 of them is at around 67-70%. 

What's causing that?

Something else is using the graphics card at that moment.
I get something like that when other programs trying to use it, and my hashrate goes down. I did notice that when I move my mouse fast the usage goes down too. It goes back up again if I disconnect.


 

It's normal ft this card to run memmory at 600 Mhz?

[SaintDevil]

It's normal ft this card to run memmory at 600 Mhz?

Not originally, I underclocked it.

I would think the power supply would be sufficient at stock clocks and volts, I think it's more likely a driver issue (what version are you using?) with monitors not being connected to the cards, might need a couple dummy plugs.

Yes but don't forget that mining pretty much uses all of the graphics cards power, also the power supply may not provide continuous power. If it says 800w it is probably just on the pick.

[tacotime]

Well, if it still does that with only one card in though, there's something fundamentally wrong with the card I would think.  He should test it on another system before he buys a new PSU.  I run 3x 7950s on an 850w PSU and I've never seen a hiccup.