I give up [i7 920]

M

myounage

Weaksauce
Joined
Jun 4, 2010
Messages
106
I've tried OCing mine, I ran a working 3.30GHZ OC for a long time with very slight instability (BSOD randomly after several days of uptime no matter how much load the system was under. Could prime95 / linx for 72 hours or just leave it sitting, would still BSOD a little after three whole days. Installed another 6GB of RAM, which passes memtest86, WMD, etc and works fine. It's the same model as my original kit. A few days later, I started having very frequent instability. Bluescreens mostly (1A memory management). But memory management isn't exclusively a RAM BSOD. It also covers hard drives and video RAM, and considering I pass memtest I highly doubt that's the issue. I scaled back to stock CPU settings and now they are gone, but my system also started just going black and restarting.

I nailed that down to a loose ATX 12 pin connector on the motherboard. Not too good, but whatever. I tightened it down and now after about 20 minutes playing a game I got a nasty enough crash, the sort where the last sound played loops forever. No BSOD, nothing. I use to get about 20% of those and 80% BSODs whenever the system failed before, now I don't get any BSODs but quite a few of these.

CPU base clock and mult is stock, RAM timings are set as per the manufacturer, heat is not over spec, and everything generally seems to work fine tested independently, but falls to bits when put together. I can run almost all games fine, with some (Bad Company 2, League of Legends) triggering halt within 30 minutes. I can run 3 VMs, 2 Eve Online clients, and another mmo at the same time with no ill effects most of the time. I give up, I need help with this thing.

Voltages:
Vcore: 1.12v (this should power a 920 fine at stock .. I think.)
PLL: 1.8
QPI/uncore: 1.28 (Seems to hang easier if this is lower, and I've heard rumors it's suppose to be 1.3 for my RAM, but that sounds crazy with the effect this setting has on thermals.)
DRAM: 1.64

specs:
cpu: i7 920
mb: Asus P6T
ram: 12GBs (6x2GB) OCZ Gold 1600
gpu: BFG GTX 275 (previous one broke, it's a replacement.)

tl;dr: It worked fine, even with the new parts, now it doesn't. New sticks test perfectly in memtest86, and worked as such for long enough. I legitimately give up. Also the CPU is way hotter than it should be with my cooling, settings, and voltages. The thermal compound application is fine, and the HS mount is snug and perfect. It stays under 60 core temps in linx / prime95, but it's still stupidly hot compared to how it should be with a push/pull H50.
 
Last edited:
try setting your ram to jedec standard speeds 1333 9-9-9-24, lower voltage to 1.5V and see what happens... memtestx86 isnt the most reliable test.
 
86 5.0L said:
http://www.overclockers.com/3-step-guide-overclock-core-i3-i5-i7/

read it throughly
Click to expand...

Nice guide, and I'll refer to it when I feel like pushing forward again, but I'd rather have a system that can handle stock base clock & mult. I'm not concerned with OC, I'm concerned with having the system work at 2.66GHZ again.

I've just started a 10 min Prime95 run, but I more or less know it'll work. What's the deal with them saying to set IOH to 1.3V though? Mine is at the default of 1.10 w/ 133MHZ base clock, why should it be increased?

EDIT: Max for any core seems to be 63 C for prime95 w/ my voltages and stock speed.
 
Last edited:
It goes over very important settings and why they need to be set at certain values. It also starts at the very bottom and makes sure to eliminate any wild variables that can mess things up
 
myounage said:
Nice guide, and I'll refer to it when I feel like pushing forward again, but I'd rather have a system that can handle stock base clock & mult. I'm not concerned with OC, I'm concerned with having the system work at 2.66GHZ again.

I've just started a 10 min Prime95 run, but I more or less know it'll work. What's the deal with them saying to set IOH to 1.3V though? Mine is at the default of 1.10 w/ 133MHZ base clock, why should it be increased?

EDIT: Max for any core seems to be 63 C for prime95 w/ my voltages and stock speed.
Click to expand...

That was the first I had heard about the IOH to 1.3 as well....:confused:

Your temps do look hot for a push/pull H50.

My i7-930 is prime95 stable with 1.03v on vcore with stock settings on clocks stays under 50degrees except for 1st core which has multiplier of 23.. that one goes to 50. I'm running a H50 with only 1 fan pulling.

Your qp/vtti may be a little to high.
 
I raised my vcore a bit and no longer get BSODs, but I also raised QPI/VTT to 2.8 from 2.5. That said I had it at 2.6-ish when the new RAM went in and didn't have problems, so maybe I should try lowering vcore more soon, since it seems like QPI/VTT may have been the problem.
 
Take one step forward and one step back on each then go back and forth till you find the sweet spot.
Raising your vcore a notch got rid of BSOD. I would raise it again and swee what happens. I think your vtt is too high for stock clocks.
 
Well, scratch that. Halted within minutes of starting LinX. I use to blast through hours of it at these voltages and 3.3GHZ with 160 base clock. The RAM manufacturer flat out told me to set QPI/VTT to around 1.30. It should also be noted that I have not seen a real BSOD in some time. Every single halt within the last week has failed to produce one. Whatever is happening here absolutely does not produce blue screens, nor does it produce any form of memory dump. Everything simply stops, looping any present sound forever.
 
Oh cool .. running it again and now my temps are pushing 70 on two cores .. wow. I think it actually ran a few degrees cooler when it was OCed .. using the stock H50 fan. Something's very wrong here, and I think it's the actual chip, because the TIM and mount are fine. I've done it all before and they are fine.
 
My guess would be your initial culprit was your vcore was set too low and that was causing the random instability. Why it would run fine for a few days and randomly crash is beyond me. Try setting your vcore at 1.3 initially and after getting things 100% stable (either overclocked or stock) gradually drop that down. Keep your QPI in the same range as the vcore. It needs to be .5 from the RAM voltage. Don't overclock your RAM until your CPU is completely stable.

The low 70s are probably where the H50 is going to keep your CPU under load. I have a custom loop with an external 2x120 Rad and under load my 920 gets into the low 60s.

Just to eliminate another possibility what PSU are you using? If its a legitimate brand and powerful enough it is probably not the PSU.
 
OCZ 700W PSU. The rails suit the hardware, and I could add a second GPU with what's left over. Also, isn't 1.3 a little high for just 2.6GHZ? Isn't 1.3 4GHZ territory?
 
myounage said:
OCZ 700W PSU. The rails suit the hardware, and I could add a second GPU with what's left over. Also, isn't 1.3 a little high for just 2.6GHZ? Isn't 1.3 4GHZ territory?
Click to expand...



1.3v does seem high to me. I'm use to i7-930 though.
 
Alright, voltages have been altered again. Now running 2.6 QPI/VTT and 1.15 Vcore. Have been stable in Linx so far, and I am running Furmark next to it to get maximum system stress, meaning 100% CPU load, all RAM used, and 100% GPU load. Been running that a good 20 mins now, seems stable. Highest recorded core temp was 75 though .. yeow. My "CPU temp" reported by the board is a full 10C below any of the core themps though. I've heard 4-5 is normal. Perhaps it is remount time?
 
I see 10 to 15C differences between the onboard readout on my eVGA SLI3 and what coretemp reports. I believe the coretemp number over the motherboard sensor.

Highest recorded core temp was 75 though .. yeow
Click to expand...

Still got 20C to push it.. :)
 
I used 1.375 on my 4Ghz i7..

1.3 seems low to me and from what I've read.

Could be 100% wrong though.
 
drescherjm said:
I see 10 to 15C differences between the onboard readout on my eVGA SLI3 and what coretemp reports. I believe the coretemp number over the motherboard sensor.



Still got 20C to push it.. :)
Click to expand...

AFAIK isn't the max safe sustained temperature 67? TJmax is not "feel free to run 5 degrees less than me." it's "shut down immediately. From what I've read you do not want to consistently go over 67-70.
 
Around 100C is the maximum safe temperature on most Intel CPUs core2 and above.

Over that they automatically throttle to prevent damage. My last two quad (Q9550 and i7 920) core intel chips both hit 75C at stock voltage and the stock fan when all cores were maxed out.
 
myounage said:
AFAIK isn't the max safe sustained temperature 67? TJmax is not "feel free to run 5 degrees less than me." it's "shut down immediately. From what I've read you do not want to consistently go over 67-70.
Click to expand...

haha let me show you what a clogged H50 does, and I had been folding at these temps for almost a week, till I cleaned it out

89419909.jpg
 
drescherjm said:
Around 100C is the maximum safe temperature on most Intel CPUs core2 and above.

Over that they automatically throttle to prevent damage. My last two quad (Q9550 and i7 920) core intel chips both hit 75C at stock voltage and the stock fan when all cores were maxed out.
Click to expand...

Why does the Intel website list 67C as the maximum safe temperature for Nehalem? Also my chip liked to hit 90 on stock cooling / voltage .. thus the H50.
 
Well. It just happened again. I think something in the case is permanently screwed. This time my mouse and keyboard shut down (the LEDs went out, other USB devices kept going.) My voltages should provide CPU stability, and maybe they do, but something isn't working. I successfully stressed every last bit of RAM with linpack for two hours. Played three games of League of Legends and BAM, screwed. Wonderful.
 
SonDa5 said:
Psu failing?
Click to expand...

Shouldn't be. In hardware monitor my PSU voltages are all stable at where they should be. 12V is 11.82-88, 5V is 5.00something, and 3.33V is 3.33something. Generally speaking a failing PSU will have erratic or out of tolerance rail measurements.

EDIT: I can also run linpack / prime95 with furmark on the side. I think a PSU would trip over that almost immediately, not on some game I can run at 130 fps that hardly touches the CPU or GPU.
 
Bent pins on the motherboard under the CPU can exhibit that behaviour, had it myself before.

New mobo, the CPU temps went down.

Have you installed Whoscrashed to see what the BSODs point to?
 
My i7 920 c0 has hits 90c ish full load at max overclock of 3.96GHz, It's fine, has lasted over 2 years so far, these things don't mind the heat :)
 
Arthur Hucksake said:
Bent pins on the motherboard under the CPU can exhibit that behaviour, had it myself before.

New mobo, the CPU temps went down.

Have you installed Whoscrashed to see what the BSODs point to?
Click to expand...

I haven't had BSODs in a while now. Now it just freezes and loops sound forever with a still screen.
 
myounage said:
OCZ 700W PSU. The rails suit the hardware, and I could add a second GPU with what's left over. Also, isn't 1.3 a little high for just 2.6GHZ? Isn't 1.3 4GHZ territory?
Click to expand...
Depends on the chip. My C0 i7 920 needs 1.3v to hit 3.6GHz and 1.4v to hit 3.8GHz. I can't run stock below 1.2v. Per Intel the voltage range is .80 - 1.375, so 1.3v is within that range.
 
myounage said:
Why does the Intel website list 67C as the maximum safe temperature for Nehalem? Also my chip liked to hit 90 on stock cooling / voltage .. thus the H50.
Click to expand...

That is TCase.

TJunction is what the thermal diode measures and the upper limit for TJunction is at least 100C on all Intel CPUs core2 and above. I have seen 105C max on some CPUs.
 
Wow, 1.3Vcore is what my 950 needs for 4.1Ghz

My understanding was the 950s just a 23x 920.
 
It's your PSU, I'm almost 100% sure. I know because I had the exact same PSU and had the exact same problems. Its a very popular power supply but its not actually that good when you push it. I tried so many things and then bought a Corsair 850HX on sale just on a whim and ALL of my problems went away. That and I got 4+GHz OC's reliably once I switched and I could never get that when I had the OCZ 700w. I thought I just had a bum chip, really opened my eyes to how important a great PSU is.

Seriously, go to MicroCenter or BB or something and buy a good one, if your problems go away then great, if not return it. MC has 30 days no questions. Worth a try.
 
The OCZ ModXstream Pro 750w is the reason i can't overclock my GTX470 higher than 750core and i think it's the reason why i hit a wall at 3.96GHz. The split rails screw overclocking for me right up.
 
Alright. This time I had an uptime of 2 days before it happened. I went into my case and examined the main motherboard power connector. Noticed it wasn't fully clicked on and pulled it off before putting it back on, pretty sure that was the source of the issue. Checked all other cables as well.

Now I'm wanting to OC again. My current QPI/VTT is 2.75 and I just got a 0x124 in Linx after a few minutes. Up it? Down it? My RAM manufacturer suggested a setting of 1.30 but that seems slightly insane. Since 124 can mean too high or too low I really don't know where to start with this. I just wish linpack were a bootable utility so I could fail stability tests without putting my OS and filesystem at risk. at the same time it wouldn't give us all those useful BSOD codes though .. sigh. RAM is running at same speed as when I ran stock settings, and I was able to pass memtest86+ with it there, but haven't tried that at these settings. Shouldn't matter, but it might.
 
Alright, lowering to 1.200 caused an immediate 0x9c hardware_check BSOD before the login screen. 1.225 caused a 124 in Firefox (at 1.28 I had to stress test to get it). It would seem it was too low right? At 1.375 right now, will stress.

EDIT: Up to 1.29something now. The system feels more usable while Linx is running than it did at 2.6-2.8. I think up is the right direction here.

OC stats:

Base clock 160
CPU mult 21
vCore 1.175
QPI/VTT 2.9~~
DRAM 1.64

Cores load around 77 in Linx as is, but Tcase maxes out at 65 so it's considered in spec.
 
Last edited:
Alright, now it crashed while gaming with a page fault in non_paged area bsod, and it's less stable than it was a few hours ago. I think something's in the process of dying but can't tell what.
 
Your qtt/vtt shouldn't go over 1.5V. I've read at 4.0Ghz, 200x20 it'll be around 1.36ish.

If you had you vtt/qtt at 2.6 you've probably fried your board.

Are you using any guide?.
 
Patonb said:
Your qtt/vtt shouldn't go over 1.5V. I've read at 4.0Ghz, 200x20 it'll be around 1.36ish.

If you had you vtt/qtt at 2.6 you've probably fried your board.

Are you using any guide?.
Click to expand...

Right, I meant 1.275. Not sure how people mess up voltages like that when talking about it. And no, not really. Not apart from general knowledge about the platform and what some of the BSOD codes mean. Anyway, my QPI/VTT is now 1.30V, and I managed to catch a page fault in non-paged area BSOD, which I've never seen before. Up it? Down it? Ran memtest86+ and still could not find any memory issues. Obviously I ran it longer than this, but maybe it helps some.

2q3p6op.jpg
 
You must log in or register to reply here.
Back
Top