Quote from: gmerk on June 22, 2014, 05:56:35 AM

Quote from: smtn512 on June 21, 2014, 09:06:54 PM

Quote from: gmerk on June 21, 2014, 04:11:39 PM

I found the opencl files in AMD folder inside \Packages\Drivers\Display\WB6A_INF\B171094 (after decompressing the 14.6 driver package without installing it, or with 7zip)
I don't know which are needed, so I copied everything that had "opencl" or "amdocl" in the filename into the sgminer folder (8 files, dlls & exes).
Deleted all previous bins, and now it's even better (+3%) than using the bins from 14.6 (I have 14.4 drivers installed).

Did this... 14.4 drivers with 14.6 opencl files in sgminer folder. Msi 7790s up from 1.17mh to 1.62mh. Sapphire 270s up from 2mh to 2.72mh. If anyone is interested in my configs I'll post them upon request.

Quote from: restless on June 21, 2014, 12:58:08 PM

I use a 7790 (seven-seven-nine-zero, a bit weak I know, but only ~70W).
With 14.6 it went from 1.2MH/s to 1.8MH/s but with locked memory clock (core clock not locked).
I removed 14.6, kept the .bin files, reinstalled 14.4 and everything is fine.
If I recreate the .bin with 14.4 I get my old low hashrates again...

Yes these little things are great power savers, and they run so cool...wish I had bought more last December when they were .1btc each. I'm very curious to know whats in your 1.8mh config...

I have a gigabyte 7790 with 1Gb GDDR5. Core @1150 Mem @1200 (for less heat)
I'm getting 1.830 MH/s and 0.027 WU/m (currently at fractalcoin)
This is my conf. It's nothing special...

Code:

]
,
"xintensity" : "100",
"vectors" : "1",
"worksize" : "128",
"kernel" : "x11mod",
"thread-concurrency" : "8000",
"lookup-gap" : "2",
"gpu-threads" : "4",
"gpu-engine" : "1150",
"gpu-memclock" : "1200",
"queue" : "2",
"scan-time" : "4",
"expiry" : "8",
"no-pool-disable" : true,
"failover-only" : true
}

TC doesn't play such important role as in scrypt. I also use 8192 or 2048, without any significant change.
Memory clock also doesn't affect my rate much, so I keep it low to drop the temps.
You can also try more threads (6 or 8 ).
Q/S/E based mostly on the coin and pool and my experience/hope/illusion/paranoia...

Update: After a full day of mining, the memory clock is again stuck at stock speed (1500).
I restarted the miner but its still stuck. I'm using djm34 version.
I switched to kozmos version (with bins from 14.6), and the clock is again working ok.
I think this is getting too confusing for me to follow...

From what I've read so far, it seems series 5xxx & 6xxx don't benefit as much by 14.6 or the new miners.

Update2:
If the 14.6 opencl files are in sgminer folder, when I run sgminer, mem clock gets stuck @ stock freq.
If I remove them (after I make the bins), then mem clock acts normal. Weird or not, it works.
I've tried this with both djm34 & kozmos versions for both x11 and x13...

Thanks for the follow up. The MSIs have never liked multiple GPU threads, on any algo... Tried it today with 2 threads and most of your settings and I did get into the 1.8mh range, but after a few hours a couple of cards just slumped down to 1.4mh for some reason. Clocks still the same, no proof of a driver crash but yet they slowed down, had to reboot to get full hashrate back...odd. I did change to Xi 100 and moved my clocks around some... Kept 1130/1400 1 thread per gpu which got them to 1.75mh each just fine holding 66C in a Texas garage with AC maintaining 30C (got tired of the rigs fighting my central air) each card pulling 65-70w total. Good enough for me!