Such thing as too much voltage?

I ran memtest on my recently overclocked memory and it passed 12/13 tests, I consider this pretty stable. OC'ed CPU passed Prime95 for 10 hours. Temps are a inline (+15C over ambient case = 50C under full load).

Played a little UT2003 last night, after about 20 minutes my PC beeps and the screen becomes a colorful and garbled mess. I reboot and everything appears to be fine.

The GPU is also oc'ed, but have played a lot of 3D games with the same overclock in the past with no problems (prior to OC'ing the rest of the system).

I tried raising the vdimm to 3.0v and the PC worked fine for a bit yesterday, then when I turned it on this morning it blue screened. Reduced the memory back to 2.9v and it booted fine. I noticed similar issues with raising the vdimm to 3.0v when initially configuring/testing the overclock, seems to do the bsod every time.

At this stage I'm going to assume the memory's overheating at 3.0v, but I wanted to be sure it's not something else. do a lot of music recording on this machine and walked away from oc'ing because I needed 101% stability (which I achieved by not oc'ing). I couldn't resist grabbing a little extra power from this barton, though, and so now I'm back to dealing with flakey behavior...

Thoughts? Could this be a video issue with the AGP too far out of spec for an all-in-wonder card?

i think 3 volts is a little high, did you touch it and see if its too hot?

Some boards have problems actually delivering more than 2.9V to the DIMM banks. On the Max3, for instance, voltage is very flaky above 2.9, with dips down to 2.7 and spikes to 3.2.

I didn't touch them, and it's possible the voltage is fluctuating some... I have an Antec Smart Power 400W psu, it does a nice job but isn't infallable. It could also be the draw when gaming and loading windows causing the 5v rail to dip too low as well (another possibility I just thought of). Maybe I'll add another .025v on the vcore and see if it stabilizes things (without increasing my temps too much).

Well, 12 / 13 tests in memtest really isn't stable. It means that there exist data patterns for which your memory malfunctions at current settings and returns incorrect results. If some program also has similar data patterns, that program will return incorrect results and function incorrectly.

Chances are your bluescreens and crashes (prior to your voltage increases) are due to these memory problems. Often times, it will come about in just certain apps because only those apps happen upon data operations that your memory malfunctions on. (Or perhaps they use more memory than your ordinary, desktop / OS tasks and start accessing memory addresses more prone to error.)

As for whether or not it's possible to give too much voltage to your memory, I think you've already gotten some great responses here. As with any piece of electrical equipment, it's entirely possible to overdo it, although memory will be less susceptible to damage than CPU's, for example.

It might be worth your while to return to stock speed and timings to do the memtest tests. After all, you probably want to rule out defective memory in the first place. (I've had that happen myself.)

Have fun! -- Paul

Thanks Paul.

I kinda take memtest with a grain of salt. I had another stick of memory that wouldn't even pass a test at default settings but NEVER caused me instability, so sometimes I wonder about that program anyway and really look for multiple errors on multiple passes (usuaully within the first 10 minutes I'll get a bunch of fails when I'm definitely too high on bus, timings, or voltage).

Anyway, you make an excellent point in that I may still be clocked too high on the memory. I think I'm at cas2, I'll drop back to 2.5cas and see how things go overnight.

That's a very good point. memtest86 definitely tests the combined CPU-memory-bus arch., so an error anywhere could potentially cause a problem.

One thing, though -- I've had bad memory test well until a certain memory address range of a certain test. In those circumstances (and it was repeatable), it was most likely a bad chip on the stick of RAM.

But you did raise some excellent points. Best of luck with your continued testing! Perhaps there's another program you could use to provide a "second opinion" for your memory testing? Hmm ... I'll have to google around, because I've seen some other preferred programs as well. (But memtest86 has always worked well for me.)

Again, best of luck, and please do let us know how things continue to pan out for you. -- Paul

Hot cpu has a memtester as well, thanks for reminding me!

Okay, set the memory to cas2.5 and even dropped the vdimm back to 2.8v, tested 20+ passes last night with no errors .

Dropped my vcore back to 1.68v (thinking maybe my instability wasn't from the core) and so far it ran toast for 15 minutes fine and I'll run prime for the rest of the afternoon (for at least 9 hours)...

Thanks for the feedback on the memory, appears to have definitely been the cause.

Hey, that's great news! I'm glad that you've gotten your memory to a more stable place! And while cas2.5 is slightly slower, I'd doubt you'd notice too much of a drop in real performance.

I hope your testing works out well! -- Paul

I kinda take memtest with a grain of salt. I had another stick of memory that wouldn't even pass a test at default settings but NEVER caused me instability, so sometimes I wonder about that program anyway and really look for multiple errors on multiple passes (usuaully within the first 10 minutes I'll get a bunch of fails when I'm definitely too high on bus, timings, or voltage).

Click to expand...

Are you speaking to the win version or the OS independant version? just curious becasue I don't think that mem tests designed to run in windows are any good.

Using the dos-bootable version of memtester (think its version 2.8).

Just as an update, I had to raise my vcore to 1.70v to get stability (prime95 failed after 50 minutes on the first 1.68v run).

Otherwise everything seems fine.