Help me troubleshoot another flaky ECS K7S5A based system

[TangleWeb]

System description;

• ECS K7S5A Pro v5.0 motherboard.
• AMD Athlon XP 1800+ (1.53 GHz) running at defaults.
• 512mb RAM = 2 x Simpletech PC 133 (UD256133N38K11091703)
• 20 GB Maxtor 32049H2
• SONY CD-RW CRX215E1
• Nine Visual Technologies Revolution 4 32mb AGP
• Allied PSU Model AL-A300ATX Rev:B in inexpensive, but decent quality mid-tower case.

Symptoms & troubleshooting;

• Owner said that in the midst of doing something routine in Windows the system spontaneously rebooted. When it came back up the BIOS settings were set to defaults, no drives detected, etc. Of course it didn’t boot into Windows which is when I got the system.
• When I benched the system, I re-configured the BIOS, hitting F3 to re-detect all drives & booted into Windows without a problem. I installed SiSoft Sandra & ran the included Burn-In utility overnight (over 10 hours) without a problem.
• In case the BIOS was the issue (I have seen this many times on ECS boards), I flashed the BIOS to the latest ECS provided BIOS dated 08/11/03.
• I also replaced the CR2032 CMOS backup battery with a new one for good measure.
• Now the weird thing – I booted to a Memtest-86 CD & it ran for less than a minute then locked up. No errors were displayed on the Memtest-86 screen, it just “froze” with a flashing cursor at the bottom left of the screen by (ESC)EXIT. When I re-booted the system, the BIOS was re-set to defaults. This happened consistently, every time I re-configured the BIOS & booted to the Memtest-86 CD.
• I pulled the two Simpletech sticks & installed a random 64mb stick of PC 133 SDRam I had laying around & Memtest-86 has run 8 passes on it for 45 minutes so far without error, or freezing.

Background;

• System was custom built 10/30/03 (By me)
• I got the system back 08/10/04 as it was completely infested with spyware. I performed a clean install of the OS & apps at that time. I also ran Memtest-86 & noted numerous errors, so replaced the 1 x 128mb & 1x 256mb PC 133SDRam with the above mentioned Simpletech 256mb sticks of PC133. I ran Memtest-86 on these to verify that they were good & they were.
• Now here it is July 3, 2005 & the RAM seems to be bad again.

Is it possible this setup is killing RAM? Is there something else going on here?

This board supports both SDRam & PC2100 DDR RAM, so I'm going to install a couple of sticks of PC2100 DDR RAM & see if I fare better with it.

~Dave

[MobilePCPhysician]

Either test the psu, or use the bios screen to check the voltage. See if there is a spike or large fluctuation.

[TangleWeb]

Either test the psu, or use the bios screen to check the voltage. See if there is a spike or large fluctuation.

Good call!

I used my Micro 2000 Post Card Reader that lights an LED if the voltages are within +10% to -5%. It also has test points. Here are the results & the almost certainly the problem.

I used my DMM to measure the following;

• 3V line = 2.97V
• 5V line = 4.67 to 4.68V
  
• -5V line = +9.48V This has GOT to be a problem! I'm surprised the system doesn't have even more issues that it has.
• 12V line = 11.97 to 12V
• -12V line = -12.07V

It looks like I need to replace this PSU for sure.

Interestingly, the PSU passes the "green-light" test with one of those inexpensive PSU testers that connects to the ATX connector.

I wonder what effect 9+ V on the -5V line had on the system, RAM in particular?

~Dave

[MobilePCPhysician]

Can you use that and test with the hard drives and CD's disconnected? One or the other could actually be the cause of the spike. Usually, its the cd/dvd.

[TangleWeb]

Well, the bad news is that with a brand new Rosewill 300WATX PSU, the readings were very similar, even the dramatically incorrect -5V line;

• 3.3V line = 2.76V
• 5V line = 4.42V
  
• -5V = 9.28V
• 12V = 11.3V
• -12V = -12.27

Some of the other voltages were even off by more than the original PSU.

So, I took the mobo out of the equation by connecting the ATX power connector to the PSU tester which provides a load & test points for the ATX connector. The egregiously incorrect -5V reading is fairly normal at -5.16V. What does this tell me? I guess it tells me the board is bad, huh? Not a finding I wanted to hear

The voltages read directly from the ATX connector with PSU tester\load attached are as follows;

• 3.3V line = 2.84
• 5V line = 4.59
  
• -5V line = 5.16V
• 12V = 11.3
• -12V = -12.08

In-other-words the readings are only off if the mobo is connected.

What is the -5V line used for? I guess if there was a short between the 12V line & the -5V line somewhere on the board, I might end up with a +9V reading on the -5V line. Does that make sense?

~Dave

[TangleWeb]

Can you use that and test with the hard drives and CD's disconnected? One or the other could actually be the cause of the spike. Usually, its the cd/dvd.

I just saw your post.

I'll disconnect the hard drive, optical drive, floppy & pull the modem, leaving just the video card & motherboard with integrated sound etc. I will test the voltages using the Post Card Reader & then compare them at the actual ATX power connector (connected to the board of course)

[UPDATE] - I removed everything & tested voltages at the test points on the Post Card Reader they are the same as with the drives connected.

[UPDATE] - I found this in "Upgrading and Repairing PCs" 4th edition;

f you look at a specification sheet for a typical PC power supply, you can see that the supply generates not only +3.3v, +5v, and +12v, but also –5v and –12v. The positive voltages seemingly power everything in the system (logic and motors), so what are the negative voltages used for? The answer is, not much! Some of the power supply designs, such as the small form factor (SFX) design, no longer include the –5v output for that reason. The only reason it has remained in most power supply designs is that –5v is required on the Industry Standard Architecture (ISA) bus for full backward-compatibility.

Although –5v and –12v are supplied to the motherboard via the power supply connectors, the motherboard normally uses only the +3.3v, +5v, and +12v. The –5v is simply routed to the ISA bus on pin B5 so any ISA cards can use it. Today, though, not many do. However, as an example, the analog data separator circuits found in older floppy controllers do use –5v.

The motherboard logic normally doesn't use –12v either; however, it might be used in some board designs for serial port or LAN circuits.

I'm at a loss to know what is going on with this system. The other readings are within 10% of target values & the -5V is within 10% IF the motherboard is not connected.

~Dave

[TangleWeb]

Another user had this to say;

I had the exact same problem last December with the exact same motherboard purchased around the same time. I also noticed the voltage irregularity and bought a replacement power supply. I also checked the capacitors visually and couldn't find a problem. After a few weeks more problems developed, as the CPU would no longer run at stock speed but seemed stable at lower speeds. I thought the CMOS battery was bad so I replaced it, but that didn't fix the problem. Eventually the system stopped booting at all and I bought a replacement CPU. After installing that and not being able to start the system at all I finally realized it was a problem with the motherboard.

Eventually you might begin to notice some strange CMOS errors and the system might refuse to start on occasion.

I have built, or worked on 20+ systems based around the ECS K7S5A & many of them have developed problems as they age. Several lose their hard drives on a regular basis, my wife's system based on this same board does this almost every re-boot. I haven't noted the voltage irregularity before however.

It appears I will be searching for a suitable Socket A (socket 462) motherboard to replace this failing board.

Newegg has the PC CHIPS M811LU Socket A (Socket 462) VIA KT266A ATX AMD Motherboard - Retail for $24.99 + $2.99 shipping, less than 30 bucks to replace this failed board. I would use the techniques in the Swapping your board without so much as a reinstall thread to eliminate the need for the OS to be installed clean with the new board (since I just did a complete software rebuild on this system not long ago).

I know some will scoff at the use of a $25 PC Chips motherboard in any system, but remember it’s replacing a failed ECS motherboard, essentially a PC Chips motherboard anyway

This board has 189 user reviews on Newegg’s site, not that I base my entire buying decision on those reviews, but it does tell you that people are happy with this board for the most part & it beats a non-working board.

The only drawback is that it takes PC2100 DDR RAM, so I can’t use the pair of 256mb sticks of PC133 SDRam. This RAM is quite possibly bad anyway, Memtest-86 did lockup testing these sticks in the system with the bad board. I tested them by installing them in a Dell Optiplex GX1 PIII 600 with the i440BX chipset & it wouldn't even boot. I think this RAM is compatible with the i440BX chipset, but I'm not sure. I'll try them in a newer system shortly.

[UPDATE] - The RAM from the failing board passed Memtest-86 in a PC based on an Intel i815 chipset, so at least it’s salvageable.

~Dave

[Ruslan]

• 3.3V line = 2.76V
• 5V line = 4.42V
  
• -5V = 9.28V
• 12V = 11.3V
• -12V = -12.27

I can bet you have defective capacitors on that board!
And also most likely few bulged caps inside an old PSU... Don't be fooled by absence of leakage or cracked tops of capacitors - any capacitor with shape of the top other than ideally flat is already defective!

+5V line voltage is way below normal (4,42v) ....should be at least 4,8V for normal work (IMHO). Manuals for MBs and others are usually saying "5V -/+ 5%" or even +/- 10%, but from my experience some harddrives (and CDROMs) can start work erratically with tendency to reboot without any reasons even if voltage is just only lightly below 4,8v... as a rule - 5,2V is much better for stable work than 4,8v...take also in into consideration possibility of measuring errors...

I had had exactly the same motherboard and the same problems - solved by replacing almost all capacitors (even those without any visible defects, but neibouring ones were bad)...
Probably something is shorting (overloading) 5V line, and PSU cannot handle the load... memory errors could be also as result of all above, by the way...
...(my 2 cents)...

[TangleWeb]

Ruslan, you are almost certainly correct. I don't have the time, equipment, or soldering skills to replace the caps on this board. They do all look clean & flat, but I have not checked inside the PSU. I will be replacing this motherboard entirely. I will see about the PSU.

You are correct about the memory problems; when I installed them in an Intel i815 chipset based system they passed Memtest-86 perfectly.

~Dave