New system boot problems

[archae86]

I built a new system a bit over a week ago.

Gigabyte 965P-DS3 motherboard
Conroe E6600 processor
Arctic Cooling Freezer 7 Pro Heat Sink/Fan
Seasonic S12 430W power supply
Asus X300SE graphics card

I've loved it, but form the beginning there have been intermittent failures to boot--usually I'd just never see anything at all on the monitor, though sometimes the sequence of drive lights seemed about normal. This happened on perhaps half or a third of boots, but at worst powering off for a few minutes nearly always seemed to give me a normal boot.

Generally the system has run stably--I've processed, returned, and obtained validation of many, many BOINC work units (SETI and Einstein), so I know it ran Teraflops of computation correctly. It has run hours of Memtest86+ with zero errors reported. Core temps running BOINC flat out ran 45C to 52C, and MB temps 35C, though the infamous MCH on this board is indeed running pretty warm (uncomfortable to touch the heatsink for more than a moment).

This morning, however, the boot symptom has become vastly worse (or perhaps I have a new problem).

On turning on the the power supply switch, and pressing the main power button, I get about two seconds of power on, with approximately the normal sequence of initial lights on IDE hard drive light, floppy light, optical drive light, and sound of the initial surge of the CPU fan. Then the power light goes off for a second, and the sequence repeats. I've let it go up to as many as ten times. Each time seems the same. I've left the system off for as much as twenty minutes--no difference. The room may be slightly colder than it has been on other successful boots--so possibly this worsening of the symptoms is a "cold" boot problem in the truest sense of the word.

Any thoughts as to the prime suspect? I count me at the top of the list, but don't have a guess as to what I've done wrong. My second candidate is the motherboard, followed by the supply. It is hard to see how the CPU could do this, but I guess that is fourth.

As it happens the system this replaces has about a six months old Antec Trupower II 430, which is sitting cold. I'm inclined to try swapping it into the new box, for debug purposes, mostly.

Any advice, comments, and action thoughts gratefully received.

(my Asus p4b5330E recently recapped in Rolla continues to perform just fine--in fact I'm typing on it now).

Thanks

[gonzo0815]

Re: New system boot problems

I would try to reset CMOS with psu disconected or fully drained (switching off by the PSu switch when the sys is runign).
If it boots after that load setup dafults, save & exit restart and apply your customized settings if needed.

Then disconect the IDe drives (how many do you have btw?) and the IDE or Sata cables.

Do not overclock the system during troubleshooting.


Remove the ram and chek if there is an beep code, if yes may be the ram is faulty.
Or just use one stick.
If it is an older Seasonic PSu, may be it is equiped with Ost caps, wich may blow up premature.
The newer model do not have Ost caps anymore (at least those in the black case).
If you have this older Antec handy, i would say it is save to test the rig with it (shure, not for long therm operation...).

How mutch case fans do you have in your rig? itis recomended to have at least one or two, as the PSU did not have that mutch airflow.

The ICH cooling is as usuall on Ggigabyte boards not that good, but probably the chips are on spec with that and it is silent.

[archae86]

Re: New system boot problems

To answer your questions:

I can't check beep code easily, as there is no case or MB speaker. Probably I should pick up on of thse $2.00 clip-on jobs for just this issue.

There are zero true IDE drives in the system, one floppy, two SATA HD's, and one SATA optical.

The case has a 120mm exhaust fan and two 92 mm front fans. For a Conroe E6600 system with a low-power graphics card, this system is extremely well ventilated--that is clearly not my problem.

The Seasonic supply is two weeks old.

The Antec is not old, I bought it in the fall.

Thanks for your comments. I do intend to Clear CMOS and try a boot when I get the case open. If that fails, as I expect, I'll try a power supply swap.

[archae86]

Re: New system boot problems

I'm more puzzled than ever, but currently suspect a true cold boot problem may be part of the picture.

Here is the narrative:

1. Removed PC from usage to work location--attached different keyboard, different mouse, and same monitor--no other change save on side and side panel off.

2. Reproduced previously observed two second power on followed by three second power off. Allowed to repeat several consistent cycles. Observed that the power off was almost simultaneous with turnon of the CPU fan.

3. Cleared CMOS (by jumper)

4. Machine then did not self power-down, and got to the POST screens, but did not continue on to boot. Nor did it complain of lack of OS--it just sat there at the second POST screen. Repeated a few times.

5. disconnected all SATA drives (optical and two HD).

6. Reconnected the SATA drive with OS on it.

7. Machine now was able to start Windows.

8. Serially reconnected the other two SATA drives (one HD and one optical).

9. machine continues able to start Windows.

10. Now has been running Windows with a full BOINC load for the last 20 minutes--apparently normally.

I can't make sense of this--in particular don't know under what circumstances the power supply would see something it interpreted as a shut down command just two seconds after power on--on a repeatable basis. I find Web reports of 965 chipset machines sometimes doing this once, then powering up normally--so maybe I have a more extreme case of whatever that is.

[gdement]

Re: New system boot problems

It sure sounds like a PSU issue, especially that hang at POST. I'd think 430W is enough for anything, but I suppose that isn't really true anymore.

However, you also mentioned that it powers off about simultaneously with the fan starting up. Maybe that's the last straw that causes an overload, but I wonder also if it could be a hardware monitor problem. Maybe the board is shutting off because it thinks the fan isn't working, but then it starts at the last second, so the board tries again. Sounds weird how it bounces on and off, but maybe.

Do SATA drives have a way to start them one at a time, like you can do with SCSI? If so, I'd try setting them up that way.

[archae86]

Re: New system boot problems

The cyclic two seconds on, three seconds off condition I now believe was likely a separate issue. Most likely I got the CMOS settings to a bad state during one of the "blind boots" that are part of my main problem. The fact that resetting the CMOS made that go away is supportive, though not proof. My intended CMOS settings turn off fan warn, etc. Maybe the bogus settings I got to indeed thought I had overtemp or fan failure at two seconds and shut it off. Dismaying as that problem was, however, it is not my constant companion.

I know think my dominant problem is bad "hot boot". The machine hardly ever does a successful restart if it has been running. It also usually fails a cold boot if it has only been off a minute or two.

For the last half day I've been taking off the side panel and letting it cool for 15 minutes of off time before booting--and taking good care to be jolly quick in changing CMOS settings and getting to the Save/Restart moment. That recipe seems to be working every time.

My personal suspicion is that the MCH on my board--which on this particular model is infamous for running very hot, somehow behaves a bit differently during boot if it is hot--which wrecks things. I'm strongly tempted to try rigging a fan on it as a diagnostic--if I can switch the hot restart problem on and off with that fan, it would seem pretty powerful evidence (though it could be something besides the MCH, that is about the only component in the system that seems hot at all.)

While it would be interesting to get that answer, I think this is unacceptable. I've filed for an RMA, and shall send the board in Monday. It will be interesting to see whether the next one is any different.

One interesting "feature" of this board is that the North and South Bridge temperatures only show on the PC health screen on first boot after a BIOS load or CMOS clear. I've been there three times so far, and happen to have seen the MCH at 55 to 65 C. If it is scaled properly it is pretty warm--as my fingers tell me when I tap its heat sink.

I've been running the PC most of the time with the PC power cable plugged into a power monitor. Idle in Windows runs about 75 watts. 100% CPUs running BOINC at a 20% overclock is about 145 watts. The highest single second transient I've seen was just over 150 watts. While I can't rule out a spike too short for my monitor to see (it displays once a second, and may smooth very slightly), I have trouble believing I've ever gone much over 200W--and that is input power. My graphics card is trailing edge--no fan, a modest heat sink, and it does not feel at all painful to the touch even after running for an hour. The drives are not much over 10W each, and the fans are less. If the supply was shutting down for overload, it is defective--but I've not been seeing that save the one condition (admittedly one which persisted several hours until I thought to clear CMOS).

[gonzo0815]

Re: New system boot problems

May be the fan is to slow, ....