i5 Overclock - Odd Issue

[fiish]

Hi,

I experimented with overclocking my Core i5 system today, and saw a funny issue. I basically used the Asus TurboV utility to boost the BCLK to 155MHz, keeping stock voltage, and then let the computer run a Prime95 stress test while I was at work.

So the computer got through 12 hours of Prime95 (hitting a maximum temp of 82)... and I went online, started watching a YouTube video, and the computer resets itself. I tried again with a different video, and it got some way in before resetting again.

I have set things back at stock clockspeeds (BCLK 133MHz) for now and everything seems to be fine again. What puzzles me is why the computer was happy doing a full load stress test and then went unstable with something much simpler like a Flash video.

Anyone experienced something like this before?

[Agent]

Prime95 really doesn't do a very good job of stressing CPUs these days. No doubt it can get the CPU hot, but it doesn't stress a large number of the paths the processor has (like some SSE paths). As soon as an application uses one of these untested paths you can find it's not fully stable.

Try something linpack based. To keep things nice and simple: http://www.ultimate-filez.com/?page=downloads
Grab the IntelBurnTest - considerably better than Prime95 for stress testing CPUs

[fiish]

OK I shall try that tool and see if I get more sensible stability test results. TBH I am not expecting to get a lot out of a stock voltage O/C, but getting instabilities when the ASUS utility says it's fine to go up to 161MHz BCLK isn't ideal

[kalniel]

Anyone experienced something like this before?

Yes. It's not the CPU that's the problem, but a specific memory pattern. Set your ram to 2T cmd rate and I bet it's suddenly stable for flash.

Memtest86+ pattern 8 should show up the same error I think (without crashing).

[nightkhaos]

Prime95 really doesn't do a very good job of stressing CPUs these days. No doubt it can get the CPU hot, but it doesn't stress a large number of the paths the processor has (like some SSE paths). As soon as an application uses one of these untested paths you can find it's not fully stable.

Try something linpack based. To keep things nice and simple: http://www.ultimate-filez.com/?page=downloads
Grab the IntelBurnTest - considerably better than Prime95 for stress testing CPUs

Hmm thanks for the tip. I might give my rig (and my friends rigs) a bang on Linpack... OCCT apparently support Linpack, so I might adapt my guide (see sig) if it is as easy as a long term Prime95 load.

[fiish]

Yes. It's not the CPU that's the problem, but a specific memory pattern. Set your ram to 2T cmd rate and I bet it's suddenly stable for flash.

Memtest86+ pattern 8 should show up the same error I think (without crashing).

I tried that setting and Flash still crashes (although rather than crashing and rebooting it just hangs). So I pulled back on the overclock a bit, a BCLK of 145MHz has been stable with Memtest86+ and the IntelBurnTest so I think I shall go with that.

I know it's possible to overclock a lot further, but I prefer to keep the stock voltages (for power consumption) and would rather not spend on another cooler at the moment

[kalniel]

Okay, well I still think it was a memory timing problem so if you did want to overclock more I'd concentrate on relaxing the memory timings/slowing the memory down while increasing the CPU speed.

[fiish]

Well I am not sure if I changed the correct setting - the P7P55D doesn't have a 'CMD timing' option in its BIOS but there was one which if changed would make the memory more overclockable. Should I desire further overclocking I shall try your recommendation.

[kalniel]

Well I am not sure if I changed the correct setting - the P7P55D doesn't have a 'CMD timing' option in its BIOS but there was one which if changed would make the memory more overclockable. Should I desire further overclocking I shall try your recommendation.

Sorry, it's called DRAM Timing Mode in the P7P55D Bios, set it to 2N.