Topic: Sapphire HD7950 DualX OC (11196-19-20g) made 50HW error a day (Read 1697 times)

I have the same cards and same problem.

The solution I found is that running the rigs headless, i.e, without connecting the monitor. The important part is that when I turn on the rigs with a monitor connected, hashrates drop about 100-150 kh/s per card even if I disconnect the monitor later. But if I disconnect the monitor then reboot the machine via ssh (I use linux) every card produces full hashrates they are capable of. I have no explanation for this behaviour but all 3 of my rigs occasionally do it and I can consistently solve it with procedure I described above.

I've run into a similar issue. Running the same 19-20g cards.

I'd been running a single card at ~630kh/s for a few weeks w/out to much trouble. Then I added a second card on a 1x card riser [external powered] and my hash rate has suffered.. from 620 now I'm down sub 575 and most of the time it's sub 550.

I decided to monkey w/ it this weekend and see if i could improve this and just made it worse. I started to get hardware errors one after the other [like one every 10 seconds].

I decided to do a clean install of everything as I'd tried to roll back my Catalyst from 13.x to 12.x and wondered if that had caused my grief.

So a full reinstall of Win7 Pro and I put Catalyst 13.4 on and I'm still getting hardware errors.. either that or the cards will just freeze the whole system after 5-10 minutes.

I've tried every overclock and under clock and over-volt and under-volt known to man w/out much luck. All my prior settings [.bat files] are give the same results.

Running these cards stock w/out any OC and I still get the same issues.

Intensities from 18-20 give the same results, TC's from the 16000 level up to the 24000 level [in steps of 64] give the same results. Not sure if I've fried something on the cards or my PSU or what..

One thing I didn't do was install any kind of SDK.. I just put the full 13.4 Catalyst software on and ran it from there [same as i did the first time i installed these cards].


Any suggestions or guidance on what to check to narrow the problem down to a specific area? I've got two more of these cards in the mail so.. hoping I'm not frying these things...

Thanks all

JFG

Try intensity 18

do your rig run on windows 7 OS?

edit: whit your conf, i ryun only at 380kh/s not good for me

I have the same cards on my 4 card rig. 11196-19-20g I found are sensitive cards. My stable settings ---> 614-619 khash, 950w from the wall (voltage adjusted to 1093 using trixx), no hardware errors, 70-75C are:

-s 2 --expiry 1 --queue 0 --lookup-gap 2 --gpu-engine 1025 --gpu-memclock 1500 -I 19,20,20,20 -w 256 -g 1 --gpu-powertune 20 --thread-concurrency 20400

• 20 intensity on first card causes lower hashrate on my rig, no idea why
• fan settings is your preference

If anyone find this useful, please donate BTC to 1DDabQUQCRLUjx7fXar7GztD68KHDmvRNM or DOGE to D9f58v6s2KSbpAneQFGEaFqTwKMt8JZj25

THANKS!

Hi all

like you read on title, i have a dualx on my desktop, run 24/7 for minig.
i'm pretty satisfy of hashrate, that is about 610/615 Kh/s.
this is my setting: -I 19 -g 1 -w 256 --thread-concurrency 18816 --auto-fan --shaders 1792 --gpu-memclock 1500 --gpu-engine 1090 --gpu-powertune -10
temp is stable at 72°C and fan (auto) run at 60%
voltage is loked at 1.169 and i dont know how to undervolt. (i tried with triXX but gpu crash after few and on cgminer not see difference on V value). bios n. 015.029.000.002.000000 (113-1E249100-S41)
here i have a question : GPU-Z sensor said: vddc 1.087V (54.0A), cgminer said 1.169V. what the real?
consumption is about 330/340v at the wall. (idle 80/85w)

but the big problem is that card made at least 50 HW errors every day. at this moment 2 days and 18 hours run it made 153 HW
(card have already a hdmi cable to connect my 32 inch monitor )

any advice for fix this?