Odd stoppage there Mr... I am pretty much a slave to the same BSOD code lists everyone else is but here's one just for reference:
0x101 = increase vcore
0x124 = increase/decrease vcore or QPI/VTT...have to test to see which one it is
0x0A = unstable RAM/IMC, increase QPI first, if that doesn't work increase vcore
0x1E = increase vcore
0x3B = increase vcore
0xD1 = QPI/VTT, increase/decrease as necessary
0x9C = QPI/VTT most likely, but increasing vcore has helped in some instances
0x50 = RAM timings/Frequency or uncore multi unstable, increase RAM voltage or adjust QPI/VTT, or lower uncore if you're higher than 2x
0x116 = Low IOH (NB) voltage, GPU issue (most common when running multi-GPU/overclocking GPU)
At 1.29 getting up in the 90s with a 920 has got me concerned. Are you talking core temps or dye temps? What kind of cooling are you using on your CPU? And then if your dye temp is 85-90 then your individual core temps on a 920 would be in the 100s... at least if my readings are any indication on the 45nm chips. So I'd caution you about overheating. Whether that's the error code or not- 1.29 and 85-90C temps are not good. Not three mile island, yet... but definitely two mile island.
If you're working a 920 with DO stepping at 3.8 I think you're almost out of the ballpark in terms of vcore. No disrespect to anyone's opinion and each chip is different but I'd be surprised if you even needed 1.2... You'd probably shave 10-15C off your temp if you dropped that down to 1.2 and started again. Maybe you hit a 101.. great! Bump it up a notch or two and try again. At least we KNOW what it is!
Also, I noticed your IOH Vcore was at 1.3? I'm no expert but I think IOH is usually 1.10 or so isn't it? I would return that to auto (or 1.10) because believe you me, a 9800GTX doesn't need extra coal. In fact, people with 3 Fermie GTX cards wouldn't have their IOH up that high. (those of you running tri-sli GTX580s excluded
Additionally, I noticed your QPI is actually underclocked. An average number for your QPI Link Data Rate would be about x36... which = ~6400. I wouldn't be surprised if you had that option or something close available to you. Currently at 1.3qpi I would say that is WAY overvolted (stock is 1.2) and probably also generating too much heat.
To add to that I checked out your memory timings and frequency... and I see you adjusted your timings. Now maybe I interpretted your board wrong but if your frequency is 1159 (can't remember exact number)DDR... meaning 555mhz (underclocked)... volted at 1.65 (max basically)... at 7-7-7-20 timings... I'm pretty sure that could be throwing things off as well. On the off chance I've read it wrong, you could be at 2200mhz frequency DDR and giving you that ~1100 memory (overclocked)... with timings that tight there is NO WAY it's going to pan out for you.
Tell you what:
I'm no pro by any stretch and I invite those who know more about it to chime in- but I think you've gone about trying to overclock too many things at once without getting stability along the way. My suggestion to you would be to hit the reset button. Go into your BIOS, restore defaults and start again. Leave your RAM alone and even if we want to skip a step or two (let's say you start with your current bclock and multiplier) and underclock your ram completely. Leave timings on auto and underclock it out of the way... leave it at stock voltage. Drop your vcore in at 1.25 to be safe, and put your qpi at 1.3.. again, to be safe. (1.3 is a good starting spot if you're going to cheat a bit- it's what I use, anyway
) Drop in your QPI data rate as 36xbclock, and your uncore at 2x your memory frequency.
For testing, it doesn't hurt to turn off turbo mode and speed step, and all those other usual suspects I'm sure you've read about. Run your prime. With memory out of the way and QPI back where it's easier to predict, we limit ourselves to the error codes we get. 101? Sure... bump up vcore. 124? Okay... qpi... probably going to need to bump it up a notch.
Run your tests... your 4 million hours of prime, 1000 lines of LinX, IBT for whatever it's rating is x 20 to make people happy in here that you're defining stability to their satisfaction... and if you're all good then step your vcore down a bit... and do it all over again. Get a 101? Okay, well you were at your minimum voltage... put it back to where it was and bingo presto. Your CPU is stable. NOW you can go after your RAM. Underclock it a bit, tighten up some timings... test with memtest or whatever- now the error codes you get WILL be memory related and you'll know it.
Right now you've got memory over or underclocked (again, sorry... not with it tonight), timings adjusted and at theoretical max supported voltage (1.65), bclock up around threshold limits that often need further tuning elsewhere, an underrated QPI with too much voltage...
And NOW you have this huge email to read through. Anyway... like I said, someone might step in about the RAM thing- but one way or the other, over or under, it's still screwed.
Hope this helps more than hurts!Edited by Eno75 - 2/19/11 at 1:52am