For 3 years my PC has worked flawlessly, hardware below:
Antec 650W PSU
Gigabyte GA-P55A-UD3
Intel i5 750
Fujitsu 500G Sata II HD
ATI HD5770 PCi Express 1G DDR3
Memory Crucial Ballistix 4G DDR3 Kit 1333 (2 pieces 2G each stick)


A few weeks ago after several hours I was working in KDE using Linux Mint 14 and system just shut down. When I restarted the computer would power up for about 1/2 second then shutdown for 1/2 second and this "reboot loop" would repeat. I managed to get out of it by leaving it unplugged for an hour before switching on.

I checked everything over and found a small/medium amount of dust around CPU fan. Cleaned PC thoroughly and tested with 1 stick of memory. All seemed well but noticed the memory voltage was set at 1.5V automatically, BIOS settings are standard no overclocking.
All BIOS voltages were normal and temperatures about 34C

My Crucial Ballistix memory is rated
1333MHz 7 7 7 24 1.65V

For 3 years this has worked well, but wonder if the tolerance and voltage on the PSU changes over time? To make things stable I increased DDR voltage in BIOS to 1.56V and all seemed well.

Today, (about 10 days later) I am experiencing a similar problem.
KDE shutdown randomly and went into the "restart loop" on 1/2 second off 1/2 second, never reaching BIOS.

I managed to get into BIOS and adjusted memory voltage back to 1.5V which is Auto setting. This made things worse and the reboot loop would continue restarting before I could enter BIOS. I left it unplugged for a while and did mange to access BIOS, I manually adjusted DDR voltage to 1.6 Volt and so far, uptime is 33 minutes.

Something is unstable, unless PSU voltages have drifted. Could it just be a case of DDR memory undervoltage?

Output from sensors:
Vcore: +0.86 V (min = +0.78 V, max = +1.50 V)
Vdram: +1.62 V (min = +1.42 V, max = +1.73 V)
+3.3V: +3.39 V (min = +3.14 V, max = +3.47 V)
+5V: +5.16 V (min = +4.76 V, max = +5.24 V)
+12V: +12.43 V (min = +11.41 V, max = +13.19 V)
Vbat: +5.23 V
CPU Fan: 952 RPM (min = 550 RPM)
CPU: +38.0°C (low = +127.0°C, high = +127.0°C) sensor = thermistor


Sorry for long message, thanks for taking time to read and for any insight.

Run memtest86+ (or whatever is the latest version) to test the memory. Let it complete the cycle; there should be no errors.

Memory voltage can be a factor if you add more sticks of ram (you haven't) or if you ramp up timings (you haven't). Either of those can call for a little more voltage. Conversely, slackening timings might help with a memory issue. Memtest should tell you if memory is a problem. You can also test each stick individually if you suspect it might be one only. Some linux live disks have Memtest among the boot options. I run it from the Ultimate Boot CD, which contains all the freeware test utilities, notably including all the manufacturer hdd tests.

As you suspect, it could also be psu or motherboard, or less likely, cpu. Parts swapping can tell, when you have suitable alternatives available.

without the ability to look at voltages like you have i recently installed linux mint 15 as a newbie. i ran into a similar problem i had 2 512mb sticks from different manuf. computer had sat round for at least a yr unused. mine would shut down and bios would "beep" intermittenly 1 high 1 low. in a rudimentary test i removed 1 that i had added when it ran windows xp. that was a day ago and now i plan to dl the suggested ultimate boot cd to test the sticks themselves. i hope to use this pc as a tv in my apartment. i dont subscribe to cable tv. good luck. that was my original intent now i havent picked up new laptop in 2 days wtg microsoft. lets hope its just a memory stick and not something else. newbie here

In my experience, there can be a few causes for voltages drifting
1. Ohm's Law states that current x resistance = voltage,
so, current (through pcb tracks) x resistance of said tracks = voltage (lost across pcb tracks). You have very little voltage there to begin with, and not a small current. If some power track goes, thoings are often fed, but by unexpected routes, which causes a larger voltage drop.

2. If there's a dirty/aging plug in the way, that makes it much worse, and I'd check for that. That's a prime suspect in this situation. The 3.3V pin on the power supply plug.

3. If the current drawn is too high, the voltage may reduce - often a sign of something wearing.

Check for heat where it should not be(e.g. plugs). That's often the best way forward.

According to the ATX power supply specification the allowed voltage range for the +12V lane is 11.40-12.60V. Your value is dangerously close to minimal allowed value and, more dangerous, way above the maximum allowed value. I recommend to test the machine with a different PSU.

Regarding your RAM issues, for optimal stability you should always run your RAM at the voltage it is specified for, so in your case setting it to 1.65V is the correct option.

I had a similar problem which when running memtest and cpustress test i could reproduce the problem. since it wasnt consistent to 1 subassembly of my pc. i went to heat problem and treated it with a simple old boxfan to help cooling. evrything looks clean and fan works. with covers on bios reports 30c. with this fan i have been running without incident for 24 hrs. i will put both 512mb sticks in tomorrow with fan to help cool and see what the results are there. i recommend getting the ultimate boot cd. should be able to come across a free or version with a donation attached. i slapped down 5 bucks for donation and it was well worth it. many sources available.
good luck

Just a note: The Ultimate Boot CD is a free download.

11.43V on the 12V line? I missed that.

It's overloaded, or something is leaking. Have you any noisy capacitors? Some will make noise when powered and it causes digital madness.

Thanks for all the replies.

12V line is actually 12.43V (within limits).
It is my /etc/sensors.d/sensors3.conf line that has set these dangerous limits so I will modify
the limits on sensors3.conf
Yesterday I set memory voltage to 1.6V and all was stable for about 10 hours.

Today I have a "reboot loop" I turn PC on, it starts for 0.5 second then goes off, then starts up 0.5 second, then goes off. This is bad as I cannot even access BIOS let alone any OS.

Luckily I have a friend who built exact same PC with identical parts. I am going to swap memory first the try PSU.
I hope its not motherboard as Gigabyte no longer make the GA-P55A-UD3. I have inspected board closely, all capacitors are solid state and hopefully not prone to splitting apart and electrolyte as in the can variety.

I will also try 1 piece of RAM at a time in case its bad memory, and post update soon.

Update.
I have a friend who built the same PC with exact same components. I have tested my memory in his
computer and my memory works ok in his machine even at 1.5V.
Comparing BIOS settings I see the Vcore on my Intel i5 750 is reported as 0.836V in BIOS, on his
machine the BIOS reports Vcore as 1.2V This is quite different so I'm hoping its the PSU and not
voltage regulator on the motherboard (as motherboard is no longer available).

I'll substitute PSU's next.

The voltage difference could simply be due to the fact that modern cpus behave like laptop cpus, throttling speed and voltage with load.

If the psu swap produces no change it's likely the motherboard. In my experience they fail far more often than cpus do.

So you know now that your memory seems to work correctly in your friends machine. Sadly, that is no indicator if it works correctly in your machine, even if you drive it with the same parameters. Test the RAM in your machine.
Core voltages differ between different power-states and different CPUs, even if they are of the same type. That your CPU core voltage is different from a the core voltage in a different machine is totally normal.

1 members found this post helpful.

@TobiSGD thanks for that information.

I am now using, my memory, PSU and graphics card on my friends
GA-P55A-UD3 motherboard and Intel i5 750 CPU.

Looks as though my problem is possibly my CPU or motherboard.
I ghave checked CMOS battery it measures 3.16V so is fully
charged.
Running prime95 in linux as a stress test, and I note that
under dynmaic loads the Vcore on the Intel i5 does vary.
Its about 1.14V under full load all 4 cores dropping to about 0.85Volts
under idle conditions.

I once had a fault ob a previous motherboard which was bad capacitors,
the gigabyte boards use blue solid state caps, no signs of bulging
or electrolyte, so looks like faults is either CPU or motherboard now.

I've messed around with a lot of used computers, though not in the last couple of years. In my last years I did not come across one instance of a dead or damaged cpu. All the dead computers had defective motherboards. If I were in your position, I'd look for a new motherboard compatible with your cpu and memory.

I'm looking around for new motherboards now, I did find this article:


http://www.ehow.com/info_8085489_sig...al-grease.html

It does mention restarts when overheating, but in my case the PC powers
on for about 1/2 second then shuts down and repeats again, so I think
its unlikely its overheating