The Mystery of the Unstable EP-4T4A Series

I replaced the power supply with a similar Antec TruePower 430W unit. The voltage reading between the new and old units were within 1% of each other, with new one being slightly higher on the 12v rail.

Readings recorded with a Fluke: 12v rail reads 11.98v; 5v rail reads 5.06v. USDM shows 11.673v and 5.002v respectively and is solid at idle or full load. Memory and AGP voltages are solid and never seem to bounce, only Vcore.

I went up to 1.600v Vcore in the BIOS and USDM now reads 1.55v idle and drops to 1.52v under load. It will bounce between 1.52v and 1.54v while running under load.

With these voltages, FSB at 133 and memory multiplier at 4X, system still hangs under heavy load. So I took a leap of faith and tried adjusting Vdimm, though the reading in USDM of 2.50v seemed dead-on within tolerance.

I bumped it up two clicks to 2.72v; it reads 2.74v in USDM which is outside the voltage limit for my memory (2.5v +/-5%). Interestingly enough the system is now running stable. I have overclocked to 151 FSB with the memory multiplier at 3X; system is moderately stable with this A0 stepping P4 processor.

My conclusion is that this board has high resistance in it's internal power rails, weak regulators or both. Basically it severely undervolts, and has an erroniously high reading for Vdimm and erroniously low readings for the 12v rail.

I also suspect that many users having trouble with two RIMMs installed is a direct result of memory current draw pulling the Vdimm power rail below tolerance. With a single 256MB RIMM they are likely to be just barely within tolerance and with two or a single 512MB stick they are not, causing system hangs as I have been experiencing.

HTH

After running for several weeks with the RIMM voltage cranked-up, she's back to random lock-ups. I tore the board out of the case and benched it--time to make some measurements.

I tested with the meter the power pins on the RIMMs and they match the settings in BIOS. So, the regulators are not undervolting, but the sensors are.

Okay, I put all settings to default voltages and speeds--try it again. Running a combination of Folding@Home and playing a DivX encoded movie, locks the machine within an hour.

Maybe it's Windows. Maybe a driver or something just isn't right. So I installed a fresh download of Redhat Linux 9. Download the Folding@Home Linux console, download the DivX Linux codec, download MPlayer and try the same stuff under Linux. Boom! Within just a couple of minutes, complete system freeze, no kernel panic, no error messages, notta, just locked-up tight (something I've never witnessed before with Linux, ever).

Now I'm convinced it's a hardware problem, no question--either both of these RIMMs are hosed, or it's the board; I guess the later.

http://wattscomputers.com/working/bench_4T4AU.jpg

[flytek]

run the rimms below max performance? 300mhz for example. that is the only solution i can think of. maybe the nothbridge would like a fan?

edit: interesting pic. are you running the psu without any fans? that could be excellent for an ultra silent system. if the psu was in the case and vents were say cut out above the psu could it be run safely?

[EPoX Tech]

Are you running the latest BIOS from http://www.epox.org/bios/4t4a/4t4a.zip ?

flytek: When I run at 3X, it will go longer before a failure, but will still fail.

Northbridge fan--like this:
http://wattscomputers.com/working/active_cooler.jpg

I knew I kept those old 486 accessories for something. Only, it doesn't make any difference on the failures.

For the power supply, I took the large fan off my Antec 430W; it still has the 80mm on the back. The thing gets a little warm, but not dangerous even with both fans shutdown. I probably wouldn't run it that way inside a case though. BTW, if you ever decide to touch the heatsinks inside this PSU while it's runing, don't. One of them is highly energized--ouch!

EPoX Tech: Yes, I'm running the January 9, 2003 BIOS. It is the best of all the BIOS updates so far. It would appear that the firmware guys are on the right track, especially with this latest release--the failures take longer between occurences, but they haven't gone away.

The only thing I can put my finger on is that this board needs really clean power. The first question EPoX support asked me was what kind of PSU do you have. I'm using an uninterruptable power source but maybe it's not good enough. The 7-8 weeks where the machine ran fine has been during the comfortable weather here and no one is using their heat-pumps; now it's getting warm and they're running again. I'm sure those units put a lot of electrical noise on the line. I'm tempted to get me an inverter and 12v car battery and see how things run powered that way.