Epox 8RDA+ Thermal sensing and shutdown.

Walrusbonzo · #1 (**permalink**) 15th March, 2003, 11:17 AM

*NOTE*
Post Edit:

Please read our announcement at the start of our EPoX Topic as they include restrictions which do not apply to other AOA topics.
************************************************** ***************************

Now I don't want the Epox techs to jump down my throat, I just want them to explain!

From my findings the Epox 8RDA+ appears NOT to use the CPU's internal diode for anything!

It would seem that everything is done using the boards really inaccurate temperature reader

I know this for a few reasons. I have a **** as well, this has a thermal sensor that is nearer the backside of the CPU. When I popped my 1700+ in it @ 2.083GHz @ 1.8v the temps were 5C higher than they were on the Epox 8RDA+. Same CPU settings, same cooling!

I also tested a Volcano 9 with a thermal sensor to control fan speed. It does 4800rpm at 55C. So I set it up right next to the CPU core on the topside of the 1700+ @ 2GHz @ 1.7v. At full load while the Epox 8RDA+ was stabily showing 45C my fan was spinning like crazy at 4900rpm! So the thermal sensor on the Epox 8RDA+ is very inaccurate. About 10C too low.

So I thought, lets try this thermal shutdown feature, set it to 60C, and knowing how inaccurate the motherboards sensor is I thought it would shut down at around 50C, did it? No, not even at 55C, then upto 60C and then it shutdown!

So it really does NOT use the CPU's internal diode at all, did Epoc lie?

ALSO, don't test by removing the heatsink, the motherboards sensor responds far too slow. I could be in Windows, idle with the V9 fan at 3500rpm, and as soon as I started Burn K7 it would fly upto 4000rpm, straight away. While the Epox 8RDA+ temp sensor still showed the same temps for about 10 seconds while the air in the socket heated up.

Please explain why we have been told for so long that the 8RDA+ uses the CPU's diode for thermal protection when in fact it appears not to?

EDIT: Daniel, I see no point in editing out that bit about the other board I have, it says it in my sig too

Áedán · #2 (**permalink**) 15th March, 2003, 11:39 AM

Quote:

Originally posted by Walrusbonzo
Please explain why we have been told for so long that the 8RDA+ uses the CPU's diode for thermal protection when in fact it appears not to?

The 8RDA+ *DOES* use the CPU's thermal diode for protection.
However, this part is a *HARDWARE* solution that cannot be adjusted in any way. (Look for the Attansic chip - that's the device that does it.) This hardware solution will operate even if the CPU has locked up.

The shutdown you tested is based on the BIOS reading the thermal sensor and then shutting down the PC should it reach that point. This requires a) the BIOS to be active (not all OSes use the BIOS), and b) the CPU to be running correctly. This is completely seperate to the thermal solution provided by the Attansic chip.

So, you have TWO safety devices, not just the one you've been playing with!

Áedán

Walrusbonzo · #3 (**permalink**) 15th March, 2003, 11:45 AM

So when does it kick in?

I guess this other Attansic chip reads the CPU's thermal diode?

sirtoby · #4 (**permalink**) 15th March, 2003, 12:06 PM

*Post Edited*

How do you know that 8rda+ is too low, it might be the other one that is too high..
And last but not least.. good luck on finding ONE thermal sensor that is accurate..

Walrusbonzo · #5 (**permalink**) 15th March, 2003, 01:15 PM

Quote:

Originally posted by sirtoby
*Post Edited*

How do you know that 8rda+ is too low, it might be the other one that is too high..
And last but not least.. good luck on finding ONE thermal sensor that is accurate..

I know it's too low because all motherboard sensors read low as they are external to and don't touch the CPU! I had used that thermal sensor for the fan too!

Áedán · #6 (**permalink**) 15th March, 2003, 02:57 PM

Quote:

Originally posted by Walrusbonzo
So when does it kick in?

When the thermal diode reads approx 110C, as stated in the manual (page 3-15 in mine)

Áedán

sirtoby · #7 (**permalink**) 15th March, 2003, 04:19 PM

My apologies for my slightly aggressive post..

I am just a bit tired of these "Epox are lying" threads..

Well.. this will be my last post in this thread.

k1114 · #8 (**permalink**) 15th March, 2003, 04:27 PM

Quote:

Originally posted by Áedán

When the thermal diode reads approx 110C, as stated in the manual (page 3-15 in mine)

Áedán

What good is that! Most processors die if they reach 110c, and i've heard of thermal death at temps as low as 85c.

Holst · #9 (**permalink**) 15th March, 2003, 04:37 PM

The OTP is accurate to +/- 5*c and has a reaction time of under 1us (pretty damned fast)

So you wont be at 115*c for very long at all.

Plus I doubt you the people you have seen who's CPUs died at 90*c had an accurate temperature reading... certainly not unless they had a diode reader attached to another PC when the chip died and even then you have to calibrate it... getting it wrong by 20*c when your up at 90*c + would be easy to do.

Warusbonzo.... yo urealy need you read your manual, as the OTP is explained in there, including how its a pure hardware solution that has NOTHING to do with the bios overtemp that works off the insocket doide, thats only used for Thundrbirds and Durons.

Áedán · #10 (**permalink**) 15th March, 2003, 04:39 PM

Quote:

Originally posted by k1114
What good is that! Most processors die if they reach 110c, and i've heard of thermal death at temps as low as 85c.

According to what? You're unlikely to see semiconductors fail until you start to hit the 110C mark or so. The whole idea of the thermal protection is to prevent damage to the CPU should the heatsink not be in proper contact. If you're relying on thermal protection because you want to run your CPU at 90C, then you probably need to rework the cooling solution.

A processor dying at 85C internally on it's die was either already damaged, or was running at a voltage rather too far in excess of the CPU's abilities. (Or the device wasn't measuring the internal temperature. Besides, how did you know the processor was at 85C?)

Áedán

Holst · #11 (**permalink**) 15th March, 2003, 04:48 PM

yep, running 109*c for months on end might kill your CPU.

As aidan says the OTP is there for fan/heatsink faliure not long term protection.

We assume that people will use a heatsink suitable for there CPU. plus your unlikely to be stable running 109*c
(unless you have a super duron like one I have that will run at any temp and not crash)

Áedán · #12 (**permalink**) 15th March, 2003, 05:01 PM

Quote:

Originally posted by Holst
(unless you have a super duron like one I have that will run at any temp and not crash)

How long can you run your super duron at 109C for then? I'll put a pint on less than a month!

Áedán

Holst · #13 (**permalink**) 15th March, 2003, 05:07 PM

ive never run it at those temps for a long time.

One day I forgott to run my CPU fan on it and it didnt crash without a fan, it must hae been at 90*c + for an hour or two playing games before I noticed the smell.

This was when I was aircooled with a 120mm fan on the CPU, the fan was on a rheostat and was on too low a voltage to spin itself up, so you had to press a button to short the rheostat to spin the fan up. This time I forgott to press the button.

I think that chips dead now, died in a motheboard testing accident.

Walrusbonzo · #14 (**permalink**) 16th March, 2003, 02:04 AM

Well according to AMD's tech sheets for Athlons the maximum temperature they can handle is around 90C, so I don't see how 110C is any good!

I've seen Athlons fry up at 80C(Very accurate reading) before, and that was undervolted.

k1114 · #15 (**permalink**) 16th March, 2003, 02:15 AM

I've killed at tbird at 85c. Ran fine for a while and then locked. (temps stabilized at 80-90c) but was dead upon reboot. Took a look at the chip and it was well melted. I spent quite some time researching why this happened, and I've discovered the following:

Part of the problem is that all the heat from the core dosen't instantly dissipate when the system is turned off, but the fan cooling does, and thus temps can actually rise for a few moments after powering down. 110c shutoff dosen't mean that the chip will automatically start cooling, and it's likely that a siginficantly higher temperature will be reached.

Walrusbonzo · #16 (**permalink**) 16th March, 2003, 02:21 AM

Maybe it's time Epox lowered the OTP temperature to 85C or 90C.

I know I'd feel a lot safer should my heatsink fall off, not that it will.

jcool · #17 (**permalink**) 16th March, 2003, 02:50 AM

I've had a problem with that auto shutdown, too, but it was a bit different, oc-related: When I raised the Vcore up to 2,1V, the system would reboot and/or shutdown randomly no matter if I clocked it to 2,5Ghz or only to 2,3Ghz.
Its simply impossible that the Core-Temp went over, lets say, 60 Degrees, because I have a very strong watercooling with 2 Radiators and Copper Coolers.
So I did this mod to disable the auto shutdown and now everything is fine.
Why should I have this thing enabled when it doesn't work properly?!

Lord Vetinari · #18 (**permalink**) 16th March, 2003, 03:43 AM

Are we talking about Fahranheit or centigrade on this 110.C temp mentioned?

Just as a matter of interest- I had a faulty connector on the potentiometer I use to alter my CPU fan speed. When I went to adjust it, the machine instantly shut off. It was not shorting but think the fan ground lost its connection.
Not sure if the instant shut down was on fan speed or Temperature but it was fast.

jcool · #19 (**permalink**) 16th March, 2003, 03:54 AM

I'm talking about Celsius

Áedán · #20 (**permalink**) 16th March, 2003, 04:08 AM

Quote:

Originally posted by k1114
I've killed at tbird at 85c. Ran fine for a while and then locked. (temps stabilized at 80-90c) but was dead upon reboot. Took a look at the chip and it was well melted. I spent quite some time researching why this happened, and I've discovered the following:

Part of the problem is that all the heat from the core dosen't instantly dissipate when the system is turned off, but the fan cooling does, and thus temps can actually rise for a few moments after powering down. 110c shutoff dosen't mean that the chip will automatically start cooling, and it's likely that a siginficantly higher temperature will be reached.

If your tbird was really at 85C, then you should consider returning it to AMD. AMD are quite serious in the maximum ratings - running the chip below 95C(90) should not cause any damage.

I would disagree quite strongly with your statement about the thermal overshoot. The thermal mass of the CPU is quite insignificant next to the thermal mass of the heatsink. You can test this quite simply! Stop the fan on your CPU, and see how long it takes your CPU to overheat. That length of time is based on the thermal mass of your heatsink. If you have a water cooling setup, the thermal mass will be even larger. (In a rough rule of thumb test, I was able to run my PC with no CPU fan for around 3 minutes. )

When the power is cut to the CPU, no further heat will be generated, but heat will continue to be absorbed by the heatsink, even without a fan. Hence, at the power cutoff point, the CPU temperature is unlikely to rise significantly.

Áedán