AMD X2-5400+ B.E. system unstable

[Amanita]

Hello. Here's the problem child:

System:
Mobo - Gigabyte GA-MA78G-DS3H NB-780G SB-700 BIOS flashed to latest F4
CPU - AMD X2-5400+ Black Ed. ADO5400DSWOF AMD spec: Vmax = 1.35v Max. Temp. = 72*C
HSF - Kingwin RVT-9225 (Xigmatech 923 with another name) with AS Ceramique as the TIM
Memory - OCZ DDR2-800 Platinum
PS - Antec Earthwatts 430 watt (supply rails read very stable throughout this ordeal)
HDD - Seagate IDE 200GB
DVD - Samsung IDE 20x
Case - an old Nimble ATX
Rear Exhaust - Generic Japanese 80mm ~2200RPM. It is quiet.
Video, Sound, Lan, etc. - using the onboard.
OS - Windows XP Pro SP3

I bought the Mobo and CPU a couple weeks ago and have been going crazy over the temps and voltages ever since. What I need advice on is if the CPU is dicey, or if the Mobo is bad, or if the HS-Fan is bad, or ... Any and all advice is welcome, but I really don't want to spend any more money on this thing. I believe the parts should work well at stock and one or more of them is likely bad.

I know that the air flow in the case is far from optimal but the description below is with the case open to room air. I won't put the cover on until the temps are suitable. I lapped the heatsink and have applied it about 6 different times with various TIM application methods from none to more than enough checking temps and coverage with each method. Since the HS is lapped it turns out, obviously, that the best amount is almost none (1/2 grain of rice) manually spread across the HS/heatpipes. That dropped temps 4*C from no TIM (too little) and 2*C from a 1/2 BB sized blob (too much) applied to the center of the CPU. One thing is for sure; I now know how to install this type of direct touch heatpipe heatsink and how much TIM to use. I wish I'd photographed the whole experiment - lap, clean, do not apply TIM, stress test if temps allow (they will on a lapped HS), remove and note CPU contact area on heatsink - Clean, apply TIM, stress test, remove and note coverage pattern - Rinse and repeat until lowest temps are observed.

Here's the testing results with HS Fan BIOS-locked on max and C&Q disabled with the case cover off and room temp 22-24C. CPU is run at 2800MHz stock speed and all BIOS settings are set to auto/normal except Core voltage. The memory is being run at 5-5-5-15-2T with 2.05volt. OCZ recommends 4-4-4-15-2T at 2.0 - 2.1 volt and I have MemTested it for hours at 4-4-4-15-1T @2.05v no problems so the relaxed settings should take this out of the equation:

Note - SpeedFan, Everest, HWMonitor and CPU-Z all agree within 1*C of each other and show the same Voltages. From reading I know that CoreTemp can't read Brisbane core temps properly but it does give a relative reading. In fact, as you look at the temps below notice that if you add about 25*C to the CoreTemp readings you match up with SpeedFan, et al.

*************************

CORE VOLTAGE SET TO AUTO/NORMAL -
Desktop Idle:
CPU-Z Core Voltage = 1.344
SpeedFan Temps: CPU=33 System=35
CoreTemp Core 0,1: 8,8

Prime95 Stress large FFT:
CPU-Z Core Voltage = 1.408
SpeedFan Temps: CPU=57 System=35
CoreTemp Core 0,1: 24,36
Notes - Failed 1st run but passed all subsequent runs. At this point I discovered that the Gigabyte board was overvolting the CPU automatically so I went to manual voltages in BIOS.

*************************

CORE VOLTAGE SET MANUALLY TO 1.300volt
Desktop Idle:
CPU-Z Core Voltage = 1.280 bouncing occasionally to 1.294
SpeedFan Temps: CPU=32 System=35
CoreTemp Core 0,1: 6,7

Prime95 Stress large FFT:
CPU-Z Core Voltage = 1.360
SpeedFan Temps: CPU=53 System=35
CoreTemp Core 0,1: 21,32
Notes - System crashed on exit from Prime95 (black screen = reboot). I was not able to duplicate the crash. This is the Vcore I've been using but I'm still not secure with it. I've had it fail Prime95 on occasion when Vcor = 1.360v and the temps were a couple degrees hotter.

*************************

CORE VOLTAGE SET MANUALLY TO 1.275volt
Desktop Idle:
CPU-Z Core Voltage = 1.264
SpeedFan Temps: CPU=31 System=34
CoreTemp Core 0,1: 5,6

Prime95 Stress large FFT:
CPU-Z Core Voltage = 1.328
SpeedFan Temps: CPU=51 System=34
CoreTemp Core 0,1: 19,29
Notes - From days of experience I know that the 1.275 volt setting is not stable in some way especially on hot days. It will produce a random crash now and then.

*************************
What is driving me nuts is 2 things:
1) The propensity of the system to crash as soon as the hotter core gets over 33*C (+25 ~ 58*C?) as measured by CoreTemp @1.360volt (BIOS manually set @1.300v). The CPU is then at 58-59*C. Is it normal for this CPU to have a 12*C spread? If there were only a few degrees spread from the cooler core then you could knock off ~ 8*C from the CPU and this thing would be golden. Technically it specs out at 1.350v max so this is as high as I can go and it should work. Plus it is nowhere near Temp. Max. of 72* and it still isn't stable. Is this a manufacturing defect and I got a defective chip?

2) The lack of a stable Vcore voltage regulation even with a manually selected voltage from the Gigabyte board. Is this normal and do other brands exhibit the same behaviour? Is it fixable via a BIOS update or is it hardwired into the regulator circuitry? The system crashes that have occurred have all been the screen blanking out to black requiring a system reset. Is this perhaps not a total system crash but a crash of the integrated GPU, which is set to default/normal voltage in BIOS? Is this board defective?

Remember, this is all at stock speeds with most everything set to default in BIOS. I had hoped to overclock this a little to 3000 or hopefully 3200 to get full speed DDR2-800 but the way it's going I'm having an hard time getting it stable at 2800.

In frustration I took my old Palomino 1700+ apart and cleaned the Arctic Silver Original off the exposed core, relapped the stock HSF and applied a miniscule dab of Ceramique to the core. Buttoned it up, put it in an old Antec case with *no* provision for exhaust (just the PS Fan) and reloaded XP fresh. Ran the same tests as above with the case closed. Results? It idles at 41*C and hits 49*C stressed with Prime95. This is kinda' what I was expecting from the 5400+ B.E. since both are about 65 watt dissipation devices and the Kingwin RVT-9225 is a massive honking heatsink compared to the puny stock AMD XP 1700+ heatsink.

An aside: (((This thing just crashed on me again while I was writing this. Black screen. I'd manually set it for 2.750volt. It had been idle for quite a while and the SpeedFan log shows the CPU temp was 29*C and Vcore 1.264. I reached into the case and felt the North Bridge and South Bridge. Both were almost cool, just a few degrees above ambient, maybe 35-40C. Just upped the Vcore to 1.300v.)))

If I reach in and touch the HS on the XP1700+ when it is 49*C I can feel the warmth and it does seem to be about the temp of the hot water from the tap (~120*F).
If I touch the HS/heatpipe junctions on the X2-5400+ when it is 54*C they feel cool.

Okay friends, what do you think? Bad board? Bad chip? Bad HS-Fan? Bad builder? Something else? Suggestions? I don't pretend to be an expert at this but I have built a few systems in my day and I'm stumped. Be honest, I won't take offense, I'll be appreciative.

Sorry about this long missive but it's been a long 2 weeks and I thought I'd share it.

Thanks.

[mattaggie]

this model motherboard does tend to run hot. For more info, go to avsforums. There is a whole sub-fourm for this motherboard (under home theater).

[Amanita]

Thanks for the reply mattaggie. I can't find the subforum you're referring to at avsforum. Could you post a link?

For anyone reading my above screed, what is boils down to simply is that despite my best efforts I cannot get the described system stable at stock speed.
I think I've got one that presents a very thin paradox. It apparently isn't stable at Vcore less than 1.280v @idle and 1.360v @100% load or over ~59*C CPU. As ambient temps rise it needs more volts which increase temps and so it spirals upward until a crash. It is like it is a pre-overclocked chip that is on the razors edge and *very* sensitive to any slight changes in temps or voltages, but I can't back down the voltage since that, too, leads to instability. If I up the voltage I hit the magic 59*C and it fails.
I'm stuck with stock speed and manually setting BIOS at 1.300v for a delivered 1.360v at 100% load. On a cool day with the case open and the HS-Fan at max it will be stable, I think, but I would like to close the case during the summer when ambient is 78-80*F and turn the fan on auto for a somewhat quiet system. I don't think this can take it.
I don't understand how this thing is unstable 13*C below its spec'ed max. and why it needs so much voltage even at 29*C idle. I don't understand why this thing runs at the temperatures it does, see my note above about the XP1700+. I would normally diagnose this as lack of cooling. I would expect it to be cooler considering the size of the HS, but it isn't *that* hot. I still can't figure out why the HS is cool to the touch at 50+*C, but the ramp-up under load to max. temp takes 10 minutes and the cool-down on exit takes 1 minute so it appears to be performing. If I have to buy a $60 heatsink to get this thing to run stable at stock... you've got to be kidding me.
I need some wisdom and clear thinking as I'm ready to send it all back and go Intel, but I really just want it to work since it is more than enough power for my purposes.
Tomorrow I'm going to swap the PS and try the old Enlight 300 watt that is now powering the XP1700+. The system isn't a power hungry monster as you can see. It will be dicey, but it has 150 watts available on the 12v rail I think, so it should work. The new Antec has 360 watt max on the 12 volt and it shows stable but it may be throwing transients that I can't see.

As you can see, I'm getting frustrated since the monetary hit I'll take returning the CPU and moboard will probably be up to 1/2 the purchase price and I've lost many times that in hours spent on what appears to be a piece of junk.
I have *never* had an experience like this before. I can put that HS on in my sleep.

Any help would be great.