Nice,  they look like they would be much more durable than the powered 1x pci-e risers that I have. I had to re-solder a couple of pins on mine straight out out of the box.

this is the newer configuration with risers and second fan (also, PSU is lit up)
[/quote]

Adding a second fan in that configuration is actually reducing your total volume of airflow through the housing.  You would need to make a larger housing and position the fans next to each-other at one end of the housing if you want a larger volume of air to pass over the card.  You are probably seeing an increase in air velocity due to the restriction that you have employed,  but you could achieve that same effect by using one fan and reducing diameter of the housing at the exhaust end. 

[/quote]


What you are missing is that the fan blades of the second fan are actually causing a huge restriction in that configuration.

Your example of a push/pull configuration in a box/tower is creating more internal turbulence than a single fan, and you are probably getting more complete air movement *within* the box,  but if the box is otherwise sealed, you certainly aren't moving any more air in or out of the box due to having a second fan placed in the exhaust port. In fact you are moving much less out of it due to the fan being there and causing a restriction.

If you want to increase CFM with multiple fans,  then you would have to provide an individual matched intake and exhaust port for *each* fan.  Always have matched ports for each fan or expect to pay significant penalties in the form of CFM.

Hey guys,  out of the blue I cannot get work from the x11 pool,  cgminer crashes,  and I see this message in the logs.

"Failed to convert header to header_bin in parse_notify"

I don' t know what is going on,  everything was working fine until a few hours ago.

I lowered diff to one setting less than recommended based on the faq chart, that did seem to help so far.

Maybe every ten minutes or so.  I set difficulty based on my hashrate,   but I'll try lowering it to see if it makes a difference and report back.

WP,  I'm seeing network disconnects rather frequently on the x13 us east/west servers.  Not on my end, as I have another instance on the same rig mining x11 that keeps on chugging along at the same time.
Any insight?

Link to the thread:

https://bitcointalk.org/index.php?topic=623409.0

I've been seeing my highest hashrates using I=18.  If I go to 19 or higher, or 17 or lower, it will fall off.  

Try I=18 and then see if you can goose the memclocks up 10 at a time and see what happens.

(If I go above 1500 memclock, or  1100 engine clock, my gpus will start checking out.)

What are you using for gpu threads? 4 seems to work ok for me,  I will lose hash if I go to 1. I forget what happens if I go to 2 or 3.  

I have sapphire hd7950 though,  don't know much about the MSI cards other than they might drink a little more power.

Edit: Also wanted to add that I'm using the lasybear x11/x13 miner,  although I don't know if the modded x13 kernel loaded into a different sgminer basically would give the same performance. Maybe trying  that miner would help.  It is based  off of the "2% fee" miner but does not send hash out to anyone.

I'm seeing just under 2.7 Mh/s per 7950, my config & setup is posted over there. You probably have a little headroom somewhere. 

Not to butt in, but isn't that more or less normal?  I was expecting to see about 30% less hashrate on x13 vs. x11.   

Nice man,  thanks!  Bumped me right up to ~2.67x to 2.68x  Mh/s per card,  or ~15.9x to 16.0x Mh/s total.  

Rig is now pulling 8.5 amps.

Current config:
{
"pools" : [
   {
      "url" : "stratum+tcp://useast.wafflepool.com:3330",
      "user" : "VR6'S LOVE THAT BOOST!",
      "pass" : "d=0.16"
   },
   {
      "url" : "stratum+tcp://uswest.wafflepool.com:3330",
      "user" : "FOLLOW ME TO ISLAND DRAGWAY!",
      "pass" : "d=0.16"
   },
   {       "url" : "stratum+tcp://eu.wafflepool.com:3330",
      "user" : "COWS COME FROM PLANET ZIRCON!",
      "pass" : "d=0.16"
   }
]
,
"gpu-platform" : "0",
"thread-concurrency" : "8192",
"shaders" : "1792",
"worksize" : "256",
"lookup-gap" : "2",
"kernel" : "x13mod",
"kernel-path" : "/usr/local/bin",
"gpu-threads" : "4",
"intensity" : "18",
"gpu-engine" : "1100",
"gpu-memclock" : "1500",
"gpu-powertune" : "20",
"gpu-vddc" : ".900",
"temp-cutoff" : "80",
"temp-hysteresis" : "1",
"failover-only" : true,
"failover-switch-delay" : "60",
"log" : "4",
"no-pool-disable" : true,
"no-client-reconnect" : true,
"tcp-keepalive" : "30",
"queue" : "0",
"scan-time" : "1",
"expiry" : "1",
"shares" : "0"
}

I have not played around with the config yet,  but the above was optimized for x11, other than the difficulty setting which I have since changed for X13 hashrate.
I'll post up again if I find any better settings.

Trying to avoid win 8.1 for as long as possible,  so thanks for the .bin file. Win8 is bad enough.

utahjohn, I know that you are collecting 14.6RC2 data in this thread,  but is there a .bin file available for x13 from 14.6 drivers that I can drop into my mining folder so that I can stick with my 14.4 modded driver, like I have been doing with x11?
(I'm running six hd7950 cards on one Win8 rig.)

My x13 stats right now are that I'm seeing about ~2.33 Mh/s per card,  ~14Mh/s total using lasybears x11/x13 miner. I tossed in the modded kernel and saw the same hashrate with either one. Haven't checked amp draw, but I'll estimate around 8A based on card temps.

Update on the issues that I'm having:

The problem kept persisting even after reloading Win8 and reloading all of the drivers and everything. I then noticed a couple of times when rebooting that the screen would pop back up and I could access the desktop/move the mouse around before it would actually reboot.  I noticed that the problem would occur at more or less exactly the same time every time.  Then it happened again and I realized that I could get back to the desktop and the miner would start working again if I smashed on the keyboard a couple of times with my fist. I'm thinking that the rig is going into hibernate or some other sort of power saver mode.  I guess that upgrading to 8.1 reset those settings, and the reason that my old "test" drive was working,  and main drive was working before my 8.1 update is because I had set those power saver settings to not be an issue.
I have to check those settings now but I'm pretty certain that that's the problem.

Thanks for everyone who chimed in with tips and ideas, and to WP for allowing me to post up in his thread with my issues. I feel pretty stupid right now,  but I think I see the light at the end of the tunnel, hopefully.

  

Pretty sure something got corrupted somewhere in my software....I have an old SSD with a copy of Win8 and amd 13.1x modded drivers on it,  and have been hashing Scrypt-N for the past hour to give the gpus and psu a workout, and it has not crashed or acted up yet.  Disc is small and has no space left on it,  so it is temporary but good for diag.

When I upgraded to 8.1 on my main drive,  that actually seemed to rip out my 14.4 modded driver and I had to reinstall it basically as described above.  You're forgetting a couple steps there, but yeah, it was running x11 fine until I started playing with the modded .bin file to gain some hash.  I also had loaded the sgminer"fixed" version for a minute around that time when I hadn't yet matched TC in my config to what was in the .bin, so maybe that screwed something up....trying to reset to zero at this time and re-load win8 from my purchased valid copy of win8 and 8.1 upgrade is fighting me all the way.  Invalid disc this, doesn't want to format that, etc. etc. I miss DOS.  I should be able to figure this out with google and my laptop though. Maybe.

You might want to try it without the amp meter inline.  I had a problem with one a few months ago that I believe fried a psu.
[/quote]

I pulled it out after I started having issues, and I have used it before on this rig for much longer periods of time with no issues.  PSU could be bad,  but it looks/sounds/smells ok.

Update:  After about 2.5 hours of hashing at increased rates last night,  I shut down to plug in my amp meter and restarted the rig.  When I restarted it,  I noticed that I was only hashing at about 3.2 Mh/s per card instead of the 3.7 that I had just been seeing.  So I restarted the miner from within itself and it hashed for about 10 minutes back at 3.7 Mh/s per card, and then the screen started flickering and sgminer shut down and I lost video.
Rig would re-set if I hit reset,  miner would auto-start through my .bat file in startup and I would see 3.7Mh/s per card, but the flicker/black screen scenario would repeat itself after about 10 minutes and I would have to reset.  I tried and went back to not using the custom .bin file, and reloaded my old config,  but always the same scenario after about 10 minutes.  Tried a bunch of things such as removing and reloading 14.4 and modded .inf files,  same scenario. One time I got a message that I had unplugged an audio device,  which was not true,  but got me to thinking, so I deactivated the amd hdmi audio drivers after remembering that moving to win 8.1 caused me to reload amd 14.4 and I forgot to shut them off,  but still no joy.  Just did a system restore to right after loading 8.1, and I reloaded 14.4 + modded drivers,  and it just crashed again after about 15 minutes.

I did discover that even with the 20% increase in hashrate,  my power consumption only increased to 8.5A which is nice,  but I might be heading towards a F & R since 8.1 is already pissing me off on its own for numerous reasons,  but idk what the fuck I did to cause the miner (?) gpu driver (?) OS(?) instability.

I would like to know too.

Ok,  changing the TC in my config to match that of which that was specified in the .bin seemed to help out a lot.  

I would not have known to do that about matching the TC numbers,  thanks man.

Now I'm seeing around ~3.7 Mh/s per card.  I just kicked it off, so the hashrate may taper off a little.  No HW errors,  and we'll see what the WU settles at in a few hours.

Win 8.1, six Sapphire hd7950 non-reference, amd 14.4 modded drivers, lasybear x11 miner, Penchas .bin.

My mem and engine clocks are running where I set them to in my config.

The cards are running a couple of degrees hotter, but still cool and much cooler than mining scrypt & N.
I'll plug in my amp meter to see the rigs total draw a little later.  It was ~ 8 amps before.

I may have to test out a couple other miners,  I think there is one that is supposed to kick out some more hashrate.


Here is my config that I am currently using,  for better or for worse.  I spent some time dialing in the clocks.
{
"pools" : [
   {
      "url" : "stratum+tcp://useast.wafflepool.com:3331",
      "user" : "VR6'S LOVE FORCED INDUCTION",
      "pass" : "X"
   },
   {
      "url" : "stratum+tcp://uswest.wafflepool.com:3331",
      "user" : "VR6'S LOVE FORCED INDUCTION",
      "pass" : "X"
   },
   {       "url" : "stratum+tcp://eu.wafflepool.com:3331",
      "user" : "VR6'S LOVE FORCED INDUCTION",
      "pass" : "X"
   }
]
,
"gpu-platform" : "0",
"worksize" : "256",
"thread-concurrency" : "8192",
"lookup-gap" : "2",
"kernel" : "x11mod",
"kernel-path" : "/usr/local/bin",
"shaders" : "1792",
"gpu-threads" : "4",
"intensity" : "18",
"gpu-engine" : "1100",
"gpu-memclock" : "1500",
"gpu-powertune" : "20",
"gpu-vddc" : ".900",
"temp-cutoff" : "80",
"temp-hysteresis" : "1",
"failover-only" : true,
"failover-switch-delay" : "60",
"log" : "5",
"no-pool-disable" : true,
"no-client-reconnect" : true,
"tcp-keepalive" : "30",
"queue" : "0",
"scan-time" : "1",
"expiry" : "1",
"shares" : "0"
}

Interesting,  I will give that a shot.  Worksize was set at 256 but TC was 23,xxx ,  and then I deleted TC because I didn't think that it was necessary with these x11 miners.  I'll try out what you are saying a little later though.