Overclocking Q6600 - system behaving oddly

Artanoon' chaps!

I'm experiencing some odd behaviour from a system I'm trying to do a modest overclock on and I wonder if anyone with more expertise can assist.

It's a Q6600 in an MSI G41M-P28 (I think, might be -P25), cooled by a Corsair H75 (maybe H70) AIO and with 1600MHz DDR3 memory. I initially couldn't boot with an FSB higher than 323, but that freq worked and ran okay for an hour or so in Prime95, with CoreTemp reporting numbers under 52.

When I tried running Prime95 for longer though, I came back to find the screen off and the power light flashing, which I thought was a sleep mode, but I couldn't wake it up other than by turning off the PSU switch and restarting. Oddly, shifting the box around tended to make it turn off, then restart so I thought something was loose, which touching the pump appeared to verify.

Tightened the cooler bracket up, restarted, all good. Even shook the actual box about to test for other loose bits, but all seemed well. Except I started getting hardware failures in Prime which resulted in reboots, so I reduced the FSB to 322, then 321, both with rounding errors on different cores.

Hmm. Thought I'd try closer to stock, just to check it was otherwise okay, so ran the FSB at 280 and...back to the flashing power light again. I have no idea what's causing this - I can't adjust the voltages as there is no option to do so, but according to CPU-Z it was topping out under 1.3v when at 9x323, so nothing major; the RAM is running at 12xxMHz (sorry, can't remember exactly), so under the 1333MHz OC the MB is rated for and well under the 1600MHz the memory is rated for.

I've checked the Prime log and nothing in there indicating hardware failure (other than the abovementioned), I've checked the Windows System log and no errors in there indicating critical failure or shutdown activity, I've eyeballed the temps for a bit whilst under load and nothing over 55/56, I've looked at the highs in CoreTemp but it must reset itself as these say about 45, can't find a log of CoreTemp anywhere and now I'm stuck.

Why is it shutting off? Why is it only doing it sometimes? Incidentally, it doesn't seem to be doing it in normal use (although it did shut itself off with a flashing light a couple of times whilst I was updating Windows, but hasn't since).

Is it RAM related?

Help!

 

Have you run MemTest to check the RAM for errors? How many sticks of RAM are you using, 2 will be fine, 4 puts the memory controller under greater stress and usually requires a NB and RAM voltage increase.

Edit: It could also be that the Motherboard is rubbish for overclocking.

 

That motherboard is dreadful for overclocking, and the VRM is probably overheating.

 

Righto - I've not run MemTest yet, but can do if you think it's worth a shot. It's a twin stick setup on this board ie. only two slots and both sticks are brand new (I realise that doesn't mean much). They're running at the rated 1.5v in the BIOS with no option to change this.

I did go into this motherboard with my eyes half open, not expecting too much in the way of o/c potential but how come it won't even pass Prime at stock speed? Admittedly, the stock run came after a hardware fail at 320 FSB so perhaps the VRMs were still hot...?

To add some colour, the case is a silent one, but does have a 120mm intake blowing in from the rear over the board (back-to-front cooling since the rad would only go on the front). Would a hot VRM result in a shutdown?

Will adding heatsinks to the VRMs and another intake to the front help matters or is that it for this board? And if so, what do you reckon to jacking the FSB back up to 320 for the speed, on the hope that daily use (it's a Mum & Dad machine so nothing heavy) doesn't push the thermals?

 

Have you run it at FSB 266x9 and set the RAM timings manually? If so then I'd definitely try MemTest.

Haven't the VRMs got heatsinks fitted? If they are overheating then a fan might help a bit.

 

A fan directly over them will help a lot, but you will also find that the NB isn't very good for pushing speed either. I honestly wouldn't recommend pushing the clocks on that board. It's probably having a hard enough time at stock.

 

Thanks chaps - appreciate the advice.

In the end, I strapped a couple of fans to the VRMs and the Northbridge and tried again at 320 FSB, but no dice. It was okay for a while, but then a core failed in Prime, I left it running and the system shut down so I assume it's just a no-go.

Will try closer to stock (it is stable at stock) and see where I end up.

 

Had more shutdowns closer to stock as well now, so I'm thinking the earlier comment may hold true - there is little to no chance of stable beyond stock with this combo.

One thing though - if I did an experimental padmod (with a conductive pen) on the CPU to raise the FSB, will that be any different to just raising it in the BIOS normally? I'm guessing not, but just wondered if there was some underlying rule I don't know about.

 

G41 chipset boards are awful clockers, not like the G31/33 stuff at all.

try and pick up a G33 based Gigabyte board, better in everyway and will overclock like a champ.

 

Thanks for your assistance, chaps.

By way of closing this thread off, I ended up re-overclocking to about 320 on the base clock and told my Dad to give it a good bit of normal usage, whilst remembering to save regularly, the idea being that he probably won't ever get close to stressing it out but the extra speed bump might be useful.

I may post back if this fails, just for the record.