AMD's thermal protection REALLY works!!!

[RanDum72]

Whew! Almost lost a perfectly fine 1600+ XP CPU. I forgot to hook up the HS fan to the motherboard, an Asus A7N-266E with the nforce chipset and 'overheat' protection. I was running several loops of 3dmark2001 just to see the stability of the system after upgrading the CPU from a 1ghz Duron. It was really choppy; the performance of the built-in graphics with the Duron was much smoother. So after running several loops, I stopped it for a few minutes and tried another few loops. I left to grab something to drink and when I came back, the system turned itself off! I took the cover off and lookd around. Iwas trying to see if the HS was attached fine when I almost burned my fingers on the heatsink. That when I discovered that the fan was not powered up. I attached the fan, powered up and went inside the BIOS to see the temps. They were around 160deg F. After a few minutes, it went down to a nice and steadt 110. Exited the BIOS, booted to Windows and now its chugging along as if nothing happened! Duration without the fan was about 15 minutes.

*EDITED*. I forgot to put the one before the 5. Its actually around 15 minutes duration where the heatsink took the brunt of the heat.

 

[CrazySaint]

Sweet Good to know it works.

 

[tritium4ever]

Hey, at least you got your heatsink on properly! I wanted to test my unlocking attempts on my 1600+ AGOIA Y and the Alpha PAL8045 was getting be a pain to put on and off, so I bought a Volcano 6Cu just for testing purposes. In my overzealousness to test my unlocking job, I didn't pay attention and actually put the heatsink on backward. It ended up being tilted and only 20% of the core touched the heatsink. Needless to say, the CPU quickly overheated and the thermal protection on my MSI KT3 Ultra2 kicked in (I set it to shut down when the core gets to 70C) and saved my precious CPU. Right after that, I destroyed the clip of the Volcano 6Cu when I tried to install it the right way (the clip just gave under the pressure). I now have a newfound appreciation for heatsinks that mount through the holes.

Thank goodness AMD finally got its act together and demanded thermal protection on all Athlon boards. It's about time.

 

[dszd0g]

I thought the thermal protection was supposed to kick in at 70C/158F? I am glad to see that it works to some extent, but I am not happy that you saw a temp higher than that. That after attaching the fan and powering up the system was at 160F indicates to me that it was probably higher than that when it shut down. 90-95C/194-203F is the point of no return, I would like to know that it comes nowhere near there. I guess there still is a bit of a difference between 160F and 194F, but I don't know how high it actually got.

I would have let the system cool down a little before powering it back up.

 

[Dulanic]

Originally posted by: dszd0g
I thought the thermal protection was supposed to kick in at 70C/158F? I am glad to see that it works to some extent, but I am not happy that you saw a temp higher than that. That after attaching the fan and powering up the system was at 160F indicates to me that it was probably higher than that when it shut down. 90-95C/194-203F is the point of no return, I would like to know that it comes nowhere near there. I guess there still is a bit of a difference between 160F and 194F, but I don't know how high it actually got.

I would have let the system cool down a little before powering it back up.

Depends on the MB, different motherboards have different settings. Especially if it reads the diode, 70C isnt all that high for diode reading. A 1600+ running at 1.75Ghz with 1.7Vcore hits 57C under load according to the diode.. and thats with a swifty 462, socket reading is only 42C. If that was a socket reading then yes it got toasty warm. But AMD specs the CPUs to 90C and while a socket reading of 90C is insanely high, if it was 90C on diode reading it would prob still be stable, at default speed of corse.

Anyways 90-95 isnt the point of no return for sure... I had a CPU hit 120C when I forgot to plug in the fan, it still run fine.

 

[JavaMomma]

A friend of mine has a 1400TB (no termal diode) his motherboard reports 60C at heavy load. His system is very stable in Win2K.
Maybe his MB is wrong, reporting to high? or maybe its just right? hard to say. hope its not to low

 

[Markfw]

The tbird 1400 at 60c sounds right. They run hot.

I also left my fan unplugged for a couple of minutes on an XP1800, on a ASUS A7V333, and it shut down. I waited 10 minutes before booting up, and all was fine. (stupid me!!) Thanks AMD and ASUS !!

 

[dszd0g]

Originally posted by: Dulanic

Anyways 90-95 isnt the point of no return for sure... I had a CPU hit 120C when I forgot to plug in the fan, it still run fine.

Agreed that it is not a sure thing at 90-95. But AMD says that is where it says the point of no return is. At that point your chip could be dead and its not their fault. Somepeople misunderstand that as the maximum AMD says that it is safe to operate at, and that is not the case. AMD seems to recommend running somewhere in the 40-60C range (It seems pretty hard to get them to say exactly what temp each chip should run in). I would not operate an Athlon above 70C.

You were lucky if that 120C reading was accurate. Some Asus mobos have had problems giving jumping up really high for one reading (i.e. 44, 44, 45, 72, 45, 44, 44). The 72 was not a real reading. I haven't seen the issue on my A7V8X, but it was definetly there on my P3B-F. If that 120C was a real reading, you are very lucky as that is quite outside of spec.

I am really glad that AMD and Asus have put in this thermal protection. It still is not as good as Intel's so I hope they keep at it.

 

[mrman3k]

Wow, very nice to hear that AMD has got their act together with thermal protection. Now I bet Tom from THG will say that he was the one who was responsible for this technology.

 

[dszd0g]

Originally posted by: mrman3k
Wow, very nice to hear that AMD has got their act together with thermal protection. Now I bet Tom from THG will say that he was the one who was responsible for this technology.

In my opinion, he probably was the one who got it to actually happen. Good for him in my opinion. He is full of his opinion, but overall reviewers do us a lot of good.

I really hope Anandtech adding the section about e-mail support does something to change it. I went with an Asus mobo, even though I have found that their tech support stinks. However, every other mobo company I have dealt with (I believe I may have had a good experience with Tyan a while back, but it would have been 5+ years ago) has stunk worse.

 

[PH0ENIX]

Uhm,

Please, correct me if i'm wrong, but doesn't this actually say SFA about AMD's thermal protection?

I thought the situation was that the thermal management could deal with STEADY increases in temperature - such as a low-quality heatsink or a failing/failed fan?
Whereas sudden phenomenal increases in temperature such as complete HSF removal, where the temp would go from say ~50c to ~300c in about 1 second, the system simply cant react fast enough to shut itself down and save the core...

If this is the case - then I still dont have any faith in AMDs thermal management (even if it's not their responsibility), especially when considering a shipped system could arrive at a customer's location; and promptly start smoking upon turning it on...

 

[RanDum72]

Asus boards do have three-year warranties though, which is a real plus. I have e-mailed Asus a few times and their response times are usually long but they ALWAYS answer back. I have e-mailed ECS a few times and not once have I gotten a reply.

 

[BFG10K]

left to grab something to drink and when I came back, the system turned itself off!

Wonderful, so AMD has finally caught up to the Pentium III.

The Pentium IV's clock throttling is the real gem because if the cooling fails in a mission critical scenario (eg a server) the system will still keep going.

 

[RSMemphis]

Originally posted by: PH0ENIX
Uhm,
I thought the situation was that the thermal management could deal with STEADY increases in temperature - such as a low-quality heatsink or a failing/failed fan?
Whereas sudden phenomenal increases in temperature such as complete HSF removal, where the temp would go from say ~50c to ~300c in about 1 second, the system simply cant react fast enough to shut itself down and save the core...

When the CPU is not under load (like POSTING), the shutoff will be fast enough even if it is just through the mobo. It is important to have an internal diode reading though, because an external temp reading would react to slow. An internal reading is fast enough.
If you remove the heatsink under full load, it could totally happen that you fry your CPU, but then a baseball bat should be taken to your head.

A heatsink will come of in the rarest instances, mostly during a move. But not when the computer is running.

 

[PH0ENIX]

When the CPU is not under load (like POSTING), the shutoff will be fast enough even if it is just through the mobo. It is important to have an internal diode reading though, because an external temp reading would react to slow. An internal reading is fast enough.
If you remove the heatsink under full load, it could totally happen that you fry your CPU, but then a baseball bat should be taken to your head.

A heatsink will come of in the rarest instances, mostly during a move. But not when the computer is running.

Firstly, I wasn't implying that you ought to remove the heatsink, full load or not.
So, i'm a firm believer in the baseball bat theory also

Now although it's almost a given that a HSF wont come off while the system is running, I do still have to refer to the shipping side of things - where if a user receives a NEW system where the HSF assembly has come unseated (and they don't notice), the potential is there to kill new equipment.
AMD obviously wont cover any warranty on the CPU in that case - and I bet getting the supplier to cover it wont be easy if they decide to be difficult about it...
Being that it's an unrecoverable loss for them - as it's a fault of the shipping company - who wouldn't have warranted the system's intact arrival - as that costs more and people hate paying expensive freight.

So i'm still wary - it'd be interesting to see what happens if a system was STARTED w/o a HSF assembly - fresh outta the box.
If it can cope with that, i'd have more confidence.

But, who's gonna be stupid enough to try and prove it?
(or rich enough, as the case may be - if I could afford it, i'd love to try it )

 

[McCarthy]

If the sink comes off a current AMD chip it's dead. Period. Whether playing Quake when it happens or at a cold post. Dead, bang, little silicon snap and a cloud of smoke.

By the time the external sensors report 90 or 80 or 120 - whatever, the chip is gone. Takes what, 3-8 seconds depending on core/speed/voltage/luck for an Athlon core to burn? Get an old dead modem or something with about the same thinkness as the packaging on an Athlon. Put your finger on one side. Put a butane lighter to the other - you won't get burnt in the time it'd take a core to go, won't even get that warm.

Fan failure? Yeah, slow enough progression. That's pretty much the scenerio here. But sink comes off? Better have a P4.

--Mc

 

[0roo0roo]

Takes what, 3-8 seconds depending on core/speed/voltage/luck for an Athlon core to burn?

from what i've read i get the impression its more on the order of half a second

i've had readings of 70c from my tualatin, it didn't kick the bucket. its not that hot anymore, more like 57c now but still