I finally upgraded my case and cpu cooler in the form of an Antec 900 and a Corsair H50 a few weeks back, and I finally decided that it was time to overclock my i7 920. I easily reached 3.5 ghz on the stock voltages, with fantastic temps and great stability (62 C max with Prime95 running, HT disabled). I decided to take it further, and found a great point for 3.8 ghz where I could get my ram running right at its rated 1600 mhz as well. I believe it is a 200 mhz base clock, with a 19x multiplier. This was not stable, and took some overvolting. I have it set at 1.35 core voltage, 1.95 cpu pll voltage, and I think a very slight bump on the dimm voltage. This setup ran great, a little hotter at around 66-67 C under full load, but was also very stable, or so I thought. I left a Prime95 test on overnight, but this morning when I checked it, it had bluescreened. My dad said he looked at it before I woke up and everything was the way I left it ("yea it had all of those gray numbers all over the screen" Thanks dad), so the test ran stable for at least 12 hours. I don't even know if the cpu was what caused the BSOD, Its getting time for a wipe and fresh install anyways, and the graphics card is slightly overclocked as well (radeon 5770, only like a 20 mhz bump). So, my question is, should I leave it the way it is, considering 12 hours on Prime95 (I'm talking the large in place FTT, not the blend, the memory subsystem is rock stable at 1600), or should I keep tweaking voltages? And if I need to adjust voltages, what should I next? I won't lie, this is my first time overvolting, and I still am not 100% positive what each one does, any help/ideas/questions? On a side note, I know it is getting old, with now usb 3.0 or sata 6, but dang do I love the BIOS on my EVGA board. I used a Gigabyte board on my friends comp, was completely lost in it. Here are my Specs: EVGA SLI LE i7-920 DO stepping @ 3.8 ghz XFX Radeon @ 920 mhz Coolermaster 500 watt PSU Corsair H50 CPU cooler Western Digital Caviar Blue 300 Gb Antec 900 Case, big fan on high, others on medium Windows 7 Home Premium Images of overclock, big pictures 3.5 ghz EDIT: dang, didn't take one of 3.8, whoops Thanks for reading/help! Have Fun! -SPNKR
You sure it's not something as simple as the PSU? My gut feeling says that kind of overclock with OC'd GPU aswell might be pushing your 500W unit. Just a bit.
Going by the trusty ThermalTake power supply calculator your system pulls 454W. If you have any cathodes it may push it higher or if your PSU is a couple of years old it may drop its wattage rating a touch, but you should be ok really. I'd suggest just trying prime again and seeing how it goes. If you pop into the admin tools section of the control pannel you can look as the Event Viewer and see if it can shed some light on the BSOD, as it could just be a driver error or such.
Prime would stress CPU 100 %, but doesn't it leave the GPU alone at idle? Not sure. If that is the case then the PSU should be able to cope I think. Another matter is if you put both the CPU and GPU under 100 % load in some games for example, then it might be close to crashing.On Johnyguru CM PSUs got miced reviews, some decent units, but their budget ones didn't deliver the stated wattage (eg a 430 or so unit only delivering 280-300). It might be the case here. If you set the BSOD to stay on then doesn't it show you an error number stating what exactly went wrong. Still you didn't really push the voltages that high so power consumption shouldn't have jumped massively.
What's the exact model of the PSU? 500 W is easily enough if it REALLY has the 500 W available from the 12 V rails that matter. The HD5770 isn't a power hog so I'd expect the DC-consumption to hit 350 W max..
Their PSU's are some of the best you can buy. HardOCP swears by them, and they do the most extreme PSU testing...
Thanks for the replies, I'll check the BSOD number in a minute. I haven't had any problems with the power supply before, and its only 8 months old. There are no lights or anything (though the fans are run off Molex's), but you are right, Its on thte top on my upgrade list. Prime95 doesn't touch the graphics card, but that doesn't mean it can't error, as its still presenting the desktop and whatnot. Again, thanks for the help, I'll check the BSOD and get back to you. +rep for all! -SPNKR EDIT: Ok, so according to the event log, it was a critical error 41 Kernel-Power. Now this is not exactly helpful, as it could be a quick power outage at home, as it did rain and such last night, or a power supply stutter. Could that be the result of incorrect voltages? It seems unlikely but I'm not sure.
run the test again. One crash in one test over 12 hours in prime is not yet a reason to panic. If it is repeatable then you might have a problem.
To be honest, I've found that the only way you can really see if your CPU overclock is stable is to use it for a few weeks. I had a 4.2GHz overclock that could run the standard stress tests for 24 hours, yet would crash often. So, use your computer for a few weeks. If the system keeps crashing (say more than once a week) then the overclock is the most likely problem. My favourite method (and one I understand that not many people will have access to) is to set up a couple of FEA models in SolidWorks; one takes about 7-10 minutes (depending on speed) to solve (it's an impossible problem as well) and the other takes a good few hours. As for GPU overclocking, I combine two methods. To overclock the core/shaders, I use OCCT to test stability. I find a point where it runs stable for 20 mins and then put the core/shaders back a couple of notches. I then test the memory by playing Crysis : I find settings that I can play at for ~ 20 mins and then again knock it back a couple of notches again.