Windows 7 ntoskrnl bluescreens

[cybercore]

TDP is not the only factor to bear in mind when overclocking. Just in theory, you can keep the temps below 0 and yet the chip will be unstable. Set all the clocks in strict accordance with the manufacturers', no overclocking untill BSOD's are gone.

 

[scruffy1]

TDP is not the only factor to bear in mind when overclocking. Just in theory, you can keep the temps below 0 and yet the chip will be unstable. Set all the clocks in strict accordance with the manufacturers', no overclocking untill BSOD's are gone.

where's the fun in that ?

i'm approaching it from the overclocker's perspective - unstable ? goto more volts - given that i am undervolted anyway

tdp is something i wouldn't even approach - max clocks at stock volts is my modus operandi

either way, it'll be success or reinstall if i can't get it sweet; meantime 90 minutes priming without incident so far

when i'm bored enough with bsod it's start from scratch

but in 4 years of very considered overclocking i have never suffered consistent bsod like this, so it's either this combo of hardware (mebbe), too wimpy voltages (most likely) or win 7 having corrupted during the video issues i had with catalyst ....

in any event, thanks for your persistence and counsel, even if i am ignoring the "go stock" option

 

[cybercore]

where's the fun in that ?

i'm approaching it from the overclocker's perspective - unstable ?

What fun in it?

- To stabilize the system. If you set the clocks and voltages as the manufacturer says, you'll get a clear picture if your hardware is actually in working order. Then, when it's been running without a single bsod for 2 weeks, again without bsod's at all, go ahead start overclocking. But for now it should be neither under nor over.

~~~~~

I understand your needs perfectly, overclocker myself, but from the bsod troubleshooting perspective I definitely recommend you stop overclocking for a while. Cuz flash and other updates you mention can hardly fix you if something's not well with the hardware settings.

 

[scruffy1]

What fun in it?

- To stabilize the system.

that's a good point

but if i last 2 weeks on the new settings then my way will have proven my suspicion , and if i get another bsod then i will take your hint, and report back in either eventuality

watch this space ~o)

btw had ~ 2.5hrs priming without complaint before i had to go to bed last night, fwiw

 

[cybercore]

but if i last 2 weeks on the new settings then my way will have proven my suspicion , and if i get another bsod then i will take your hint, and report back in either eventuality

Sounds good.

btw had ~ 2.5hrs priming without complaint before i had to go to bed last night, fwiw

Promising.

watch this space

Okey doke, that's the first thing I do, go check this one first.

~~~~~~

The nature of BSOD is either software/drivers or hardware/settings such as clocks and voltages. Looks like you've sorted the software and now it's actually up to you to clock things right.

Fingures crossed.

 

[scruffy1]

mmmm... more of that blue

now at stock ram, but 3600 (9x stock 400)

i reckon the ram is the problem... and if it sits happy there i might swap it out for my old geil which seemed more inclined to go well, or for the lesser 2x1gig of 1066

maybe the fsb on ds3 is not as capable as the ds3p ? missing the extra x0.5 multi of the e3120 hurts

but as before, it was the flash plug-in in ff that seems to have broken the machine - it threw a crash report and when i persisted twice to replay the clip - whammo ! whatever the glitch is hardware wise, flash is the trigger that blows it up

more reports later

 

[scruffy1]

well i have to say i expect the ram is the bad guy

i clocked it to the 800 5-5-5-18 as stated in cpu-z, but left the 9x multi, and dropped the voltage of the cpu to 1.20 in the bios, and left everything else alone

stable at 3.6 undervolted, but a little disappointed, because i recognise the motherboard is not great for clocking, as the ram / mobo combo spits if i exceed 1600 fsb - though i haven't dug out my ddr2 2x1gig 1066, but the geil is equally unhappy to go much further

on the plus side it idles at 34c here in late summer, and primes at not much over the mid 40's

and no bsod for days

so your "go stock" instruction was the winnah ! for the ram anyway

looks like i need a new motherbard and some better ram, or a platform upgrade
or maybe to just swap back in the e3120

or maybe to just stop doing adjustments and rebuilds for a bit

 

[cybercore]

i clocked it to the 800 5-5-5-18 as stated in cpu-z, but left the 9x multi, and dropped the voltage of the cpu to 1.20 in the bios, and left everything else alone

stable at 3.6 undervolted, but a little disappointed, because i recognise the motherboard is not great for clocking, as the ram / mobo combo spits if i exceed 1600 fsb - though i haven't dug out my ddr2 2x1gig 1066, but the geil is equally unhappy to go much further

There's gotta be something I miss: you mention undervolting each time, and I don't quite get it --> why not setting CPU and RAM as meant by the manufacturer, without going over or under? Is it not possible?

 

[scruffy1]

There's gotta be something I miss: you mention undervolting each time, and I don't quite get it --> why not setting CPU and RAM as meant by the manufacturer, without going over or under? Is it not possible?

yep, it's easy

but if i use "default" vcore then i am adding temperature without discernible improvements in stability

my standard approach is to test the max overclock available at stock volts; then i see whether a notch up in volts produces >50mhz improvement - if not, why bother i reckon

then, i reduce vcore to the stable level that holds the "original" best speed - cooler, less stress for the cpu, more leet for me

for my 939 opterons, it was 2.7 (150% stock speed) at 1.4 stable
for the single core 939's - a selection of san diegos and venus - the sweet spot was 2640 (from memory) at various volts, none over recommended vcore, several well under

seems to me there is a resonant speed for silicon from a given iteration, however binned - the difference is in the manufacturer's labelling - thus a denmark is a denmark, but might be an opty 165 up to a 180, but the stepping is more indicative of potential than the arbitrary stock number

i know the e8400 was good to go in the ep35-ds3p at 4gig, and i know the xeon e3120 (an e8500 by any other measure) hit 3750 stable in this rig (ep35-ds3, not the "p") but that the ram was at 395 (790) not overclocked, due to the extra x0.5 multi i had in hand

i reckon the motherboard and / or ram don't like me exceeding 800 5-5-5-18 and i'm not keen on pushing volts to force it - either vdimm, or northbridge, although the team dark 1 gig x2 might offer experimentation at higher fsb and a low multi to hit the same 3.6 and prove a point (9x400 versus 8x450) on stability

happy for any input, but stock vcore is not required for cpu stability from my perspective, and i am quite certain that the rig is stable as set at 3600, but pushing the ram past there (or maybe the fsb) is the issue

ntoskrnl is just the symptom, but i am still intrigued as to why embedded flash is the trigger, when much fiercer loading in games is never unstable

thanks for your ongoing interest

 

[scruffy1]

yep, 4+ days of no problem at 3.6 undervolted, including some foul hot weather

the ram is no good at overclocked settings i am more convinced; still haven't got around to trying the other 1066 ram, but may not bother

additional info :

did bother, and it runs 2x1gig ddr2-1066 team xtreem dark happily at 445 fsb (=1068, @ 5-5-5-18 at 2.1vdimm)

hmmm... ram is the problem

so i just purchased a 2x2 pair of corsair dominators to try instead

 

[scruffy1]

Re: ntoskrnl bluescreens - solved at last

last word on the matter - turns out the problem was likely just me wimping out on northbridge and gmch voltage for the amount of ram (4gb) and fsb speeds i push it to go

since adding +0.2v to both, and the requisite +0.3v to the corsair dominators for 1068mhz @ 5-5-5-15, the cpu is good to go at 4ghz reading 1.2v in cpuz under load

take that, ntoskrnl