Worried about memory frequency

[Warpspasm]

This is my first crack at overclocking, so to be honest, I don't know what the heck I'm doing really. All I did was increase my FSB from 200 to 250 and my multiplier ratio stayed at 11.00. I let the BIOS handle the Voltage automatically and it's at 1.265v. My E4500 core speed jumped from 2200 to 2750 Mhz. I ran a program called CPU Stability Test 5.0 and my core temps got as high as 69c (went to 70 for a second or two) under full load. SpeedFan has my idle temps at about 39c. Everything seems to be pretty stable. The thing I'm not sure about is my memory. It jumped to a frequency of 1000 with a voltage of 2.10. I'm running 2 gb of Crucial RAM. Is that frequency number okay? Is there something I should do now?

Any suggestions on what I should do next to either increase performance or lower temps?

Thanks.

 

[Evilonigiri]

If the pc runs stably, I don't see why 1000MHz isn't okay.

You might want to run Prime95 using blend test for 8hours.

You'd also want to manually set the voltages. Auto will set the voltages a tad too high in most cases.

 

[Warpspasm]

Prime95 confuses the hell out of me. I launched it, it failed to make a connection, then I don't know what the heck to do. I guess you have to make a connection to run a test?

If I'm stable now, I guess I can push it a bit more, eh?

What's the maximum I should let the temp reach under load?

 

[Mugz]

Keep on pushing it until it's unstable, or until temps hit ~70ºC under full load (i.e. a Prime95 instance on all cores).

When Prime starts up the first time, it asks 'Join GIMPS' or 'Just Stress Testing'. Select the latter. The former signs you up for something similar to folding@home or SETI@home, only instead of proteins or aliens it goes looking for Mersenne prime numbers. Do NOT even THINK of asking me about Mersenne primes, that was a maths lecture at university I slept through. Basically, a prime number is a number that can only be divided by itself or by 1 - like, say, 3, or 5, or 13, or something.

I usually Prime95 my rig over an entire weekend. Some people recommend about 4 or so hours, others reckon you need at least a week. I find a 30 minute run to test the settings and see if the cooler can handle the load is enough for a test run, while the 48 hour weekend run is to prove it's stable.

To summarise: 'Just Stress Testing' in Prime95, approx. 70ºC maximum under load, and if it's stable, PUSH IT!

 

[Warpspasm]

Well, here's what happened. I started testing with Prime95 at about 10:30 last night. I woke up this morning to find my PC with the PC screen frozen at the Windows boot screen. I rebooted the PC and was able to get to the desktop and look at the log created by Core Temp. I appears the PC rebooted at about 3:30 a.m., which is about a 5 hour run. The temperatures never got higher than 70c anytime during the test. Then, BOINK! the blue "your memory is hosed" screen flashed and the PC rebooted again. I tried rebooting a few times, but could never get back to the desktop without it rebooting. I went into BIOS and set everything back to the old stock settings and that didn't help either. Finally, I shorted the BIOS pins and was able to boot and I'm using the PC to send this message.

What is this all telling me? Did I push too far or was the fact that I got 5 hours out of Prime95 before the reboot a signal that everything is "pretty stable"? Does this mean that I did some permanent damage to my memory and now it will always be flaky? I'm a bit worried that I boogered something up.

 

[Solitaire]

The fact that you had to clear the CMOS manually after said five hours of stability does suggest you pushed it more than a mite too far... 70C is fairly high for modern CPUs anyway, we've come a long way since the P4/D-Series, but the culprit is probably memory - could be CPU though as that was increased prior to the test as well. Check voltages and bear in mind you really should try to set all these settings yourself (BIOS permitting) - OC tools tend to be a wee bit glitchy at best...

 

[Warpspasm]

Read my next thread. I think I damaged something.
http://www.tomshardware.com/forum/245791-29-damaged-something#t1743906