UnderVolting: voltage ripple to hurt it?!

[m25]

Well, was just trying to further lower the Vcore of my 4200+ with RMClock ; now on full load, the Vcore swings between 1.314 and 1.328, very occasionally going to 1.302 or 1.344 levels. The problem is, that if I further drop the Vcore to 1.300V, after 2-4 hours of Prime the system locks up and I believe it's all fault of this voltage swinging up and down, so how do I solve this?!
Is it just the motherboard or could it even be the cheap PSU; it has a 20 pin power connector instead of a 24?

 

[dragonsprayer]

psu is over heating as it gets hotter efficiency is lower - as the psu becomes stressed the voltage fluctuates - answer is a bigger psu!

 

[m25]

I've got a 450W PSU and nothing more than a basic single HD and a DVD/CDR combo.
To verndewd: I am actually over clocking with the CPU @ 2.4GHz instead of 2.2; the 3000+ I had before had no problems going from 1.8 to 2.0 GHz on just 1.25V, so I am a bit surprised that the 4200+ can't run 2.4GHz on 1.30V. Maybe it's just me, but I think it can do it with a more stable PSU. Don't you think so?!

 

[m25]

Thanks ; would you suggest it to be even a more powerful PSU or just a better one (450W) for my config?!

 

[joefriday]

I notice the same style of voltage drop on my D 805 with the Dell PSU. At full load, the vcore will drop from 1.15 to 1.08v. It still stays stable for me, even with a tiny o/c to 2.8GHz, so I don't care about it. However, my voltage drop is consistant and stable (i.e., it drops to 1.08v and stays there, and doesn't return to 1.15v until the CPU load decreases).

I would say the voltage ripple is nothing to be concerned about as far as CPU longevity. You're actually reducing the rate of electromigration by running the CPU below vcore spec, and even with the positve part of the ripple, you're still under the stock vcore. The rate of change of vcore (i.e., the "ripple") isn't something to be concerned about either; Pentium Ms change their vcore all the time based on load, which is nothing more than a controlled voltage ripple.

With RM clock, doesn't allow you to set vcore and multiplier settings in multiple states as a basis of % CPU utilization? I don't understand why you just don't have the settings as something like this:
80% CPU load: 2.4GHz, 1.3vcore
100% CPU load: 2.4GHz 1.32vcore

That should take care of your stability issues when running the processor at max, while still allowing for max power savings/temp reduction.

 

[m25]

Well, what worries me the most, is that occasional (pretty rare) ripple to 1.30 when I am on 1.320, because if I set it to 1.300, it means it very occasionally drops to 1.280; enough for Prime to fire an error within 2 to 6 minutes.
The P4 I have at work has a good PSU and Vcore only oscillates between 1.400 and 1.408, so if I attained a voltage stability like this, I could make it sit on 1.30V, drop the load temps another 2°C and prepare the ground for a more silent PC when my Arctic Freezer 64 Pro arrives.

 

[joefriday]

Well, what worries me the most, is that occasional (pretty rare) ripple to 1.30 when I am on 1.320, because if I set it to 1.300, it means it very occasionally drops to 1.280; enough for Prime to fire an error within 2 to 6 minutes.

With RM clock, doesn't allow you to set vcore and multiplier settings in multiple states as a basis of % CPU utilization? I don't understand why you just don't have the settings as something like this:
80% CPU load: 2.4GHz, 1.3vcore
100% CPU load: 2.4GHz 1.32vcore

That should take care of your stability issues when running the processor at max, while still allowing for max power savings/temp reduction.

 

[m25]

sorry, forgot about it; there's no way I can make such fine adjustments in RMClock. You can do it with CrystalCPUID, but differently from RMClock, it eats up a lot of CPU utilization for itself and even if it consumed no CPU resources at all; I REALLY run the CPU at 100% uninterruptedly for 8-14 hours when I render some walkthrough or animation, so at the end, I'd still end up not decreasing the voltage