Q.T.Quazar

erm... somehow i failed to notice that my ThermalTake KK 500 was not a 500W, but rather a very pretty 400W box.

I'm currently running a 8800GT 512MB OCed to 666 MHz Core, Champion Fatality Sound Board (or something--the one with the face plate et al.), on a E6750 OCed to 3.1 GHz--ASRock P45TS-R MoBo (not the greatest, I know)--2G DDR2 800 (5-7-7-24). All of this stable in Prime95 (well, stable at core temps of 110-115 degrees Celsius)--Rivatuner/ATITool/memtest86--powered by this 400W--for the last year.

Somehow, all of this is running, although when I tried recently to OC past 3.1GHz, my system crashes in Prime95 (it boots fine to Vista though)

So...

1. What should I buy for adequate power to power all of this--especially what total wattage? I'd like to get my system up to at least 3.2 GHz, and my next purchase will *likely* be Core i5/i7, so I'd like the PSU to make the upgrade. I don't plan on using SLI/Crossfire in the future, but I may bump up from stock cooling (if I can prove that the current OC problem is my PSU).

2. Can inadequate power cause high heat? My Vcore is only set to about 1.36V. After reading through this forum, I'm surprised my system is even booting.

3. A spontaneous shutdown during Prime95 is most likely caused by insufficient power, or by the thermal threshold being surpassed? This was just during the blend test.

4. Suggestions on a reasonable, mid-range price cooling solution?

Short term, I am looking to mod Oblivion as much as possible with this system, and so am trying to drag up all aspects of performance.

Any and all help appreciated.

edit: forgot to mention, also running two Seagate 7200s and a DVD burner.

Quantum Reality

110 degrees C? HOW ON EARTH DID YOU NOT KILL THAT COMPUTER



As for PSUs go get yourself an Antec 650W PSU or something similar.

__________________

nathris

I think he means Fahrenheit, 110-115F is 41-46C, which is reasonable for a low OC. Besides, the CPU would shut down if it crossed the TJ Max, which IIRC is 100 for Conroes.


@OP, its not your PSU. Prime95 crashing has nothing to do with not having enough power. (technically it does, but shut up!)

I'm running a faster CPU on more volts, a GPU that draws a lot more power, plenty of case fans, a TV tuner and a soundcard on 500W, and I still have enough power to run an 8800GT for PhysX if I wanted to.


You're OC is unstable, that's why its crashing. 3.1GHz to 3.2GHz isn't going to increase your PSU load unless you're setting the voltage to some crazy value like 2.0.

__________________

Q.T.Quazar

Unfortunately, no, that's not a typo, that IS Celsius. Highest registered point, according to realtemp, was 115dCelsius on each core. This was on the third series of blend tests. I gave up on HWMonitor and ASROCK's OCTuner because they won't even accurately report the temp after it passed 100. I don't know what to say. RealTemp passes the TJMax and then recalculates it for 200--I actually watched it tick up from 58 all the way to 115. I'm idling at around 60degreesCelsius. My GPU idles at 78dC. Case fan and CPU fan are both stock, CPU fan is a constant ~2000RPM and case fan ranges between 1800 and 2400.

I would be glad to hear that something is lying to me at this point. I'm bringing my system into shop tomorrow to confirm with a tech. I'll admit I'm out of my depth.

And the OC is actually *not* unstable--but I can't push it any further, or else I get errors. 3.1Ghz was overnight. And yeah, it was humming away on Prime95 on a blend test with no problems, ~112dCelsius. I CANNOT get it to work at anything above--only 390Mhz x8 or below. 395 resets during prime, 400 starts tossing up errors, no matter what I set the CPU or DRAM voltages to (currently 1.36 and 1.96, I believe--not at home at the moment). Thermal monitoring of CPU has not been shut down iuun BIOS, so it should shut itself down if it gets too hot.

My understandfing of TJMax was that it was the *throttle* point, not the shutdown point. But, again, I'm happy to defer to the more learned techs. I'm at a loss.

nathris

That's a problem caused by instability. You're stable at 3.1, but you need more volts to get past that.

Past 72C it will begin to throttle. TJ max is a register on the CPU that holds the difference between your current temp and the absolute max. When it reaches 0 the CPU automatically shuts down.

And you're only running a fraction over stock, so unless your stock cooler is mounted wrong, or you just so happen to be the dark lord Satan there is absolutely no way you should be getting those temperatures.

Since it is impossible for a CPU to run at 115C I'm guessing your temp sensor is totally messed up.

__________________

Q.T.Quazar

I am apprenticing to the dark lord, but I haven't made my qualifying exam yet.

OK, I'm perfectly happy to assume you're right. What's my next step? I'l be replacing the PSU (probably with a Corsair in the 500W range) and putting a new cooler on tonight (likely a Hyper 212) but, if I understand what you're saying, my thermal sensor is either broken or malfunctioning. How can I get an accurate reading?

And if the thermal sensor is gone, why is my GPU reading 78 across various programs? (again, three separate programs, including ATITool. is my MoBo fubared?)

And would it be possible to test all this by going from a cold boot to operational and checking the core temp delta? What is a reasonable/common delta from power-up to Windows first being at the desktop?

nathris

Is it safe to assume your 8800GT is a single slot? My single slot 3850 went to 95C under load, which was perfectly normal. The GPU is more tolerant to high temps than CPUs are. XFX's current 4870 model reaches 100C under load, idles at 65, and I've had my voltage regulators at 140C (for under a minute though).

The only way you can get an accurate reading is by sticking a thermal probe either on the IHS, or in the IHS (a la [H]). Unfortunately the temp is being read from a register inside the CPU so I doubt the motherboard is at fault, so it is likely your CPU. If you want to replace it you can get an e5200 for under $70 shipped, which would probably be cheaper than getting a thermal probe.

Have you checked your CPU temps in the BIOS?

__________________

Q.T.Quazar

yes, 8800GT is single slot, and the huge sucker is royally screwing up airflow in my case. CPU temps in BIOS were low 50s if i recall.

can't check anymore because i've done an overhaul.

Replaced PSU with a Corsair HX750W (overkill, but I'm in China, and it's the only decent PSU that was available), and stuck an IceBlade Pro on my CPU. System is now OCed at a light 3.2 GHz, cores at ~38C. Cleaned the hell out of everything. GPU is also 20C cooler at 63 while modestly OCed. As far as I can tell, all reporting on the chips is accurate--took box into a tech here in Guangzhou and he confirmed that my CPU temp was accurate (also said the thermal grease was still 'fine,' though I didn't see this--redone anyway).

Prime95 on blend now gives a modest temp increase to ~52C on the cores, with no further temp increase (unlike before, with the constant rise all the way to 115). No problems at 3.2. Will push to 3.6 tomorrow, though on generic DDR2-800 I'm not sure how much I'll be gaining overall.

Thus ends the saga--I kinda wish I had taken a screenshot of RT for proof, or at least demented bragging rights; I guess it doesn't save a 'best of' file. I still have no idea how I was registering those temps--I re-checked my BIOS before the overhaul and while thermal throttling was disabled, thermal shutdown *was* still enabled.

Go figure.

Thank you kindly for your help, nathris. Greatly appreciated