The Report of My Thermal Death Have Been...

A reader writes "Not too long ago, Tom's Hardware posted a video of the grisly events that take place when the heatsink is removed on an AMD Athlon MP 1.2GHz in an attempt to show that the chip has inadequate thermal protection unlike the Pentium 4. Apparently, this is not the case. This new video, which looks like was done by AMD, shows the system continuing to work when the heatsink is removed. Even 9 minutes of Quake3 without the fan operating wasn't enough to destroy the processor. So who is right? It's in AMD's best interest to show that their product doesn't disintigrate under extreme conditions. " Update: 10/30 14:11 GMT by H : Note that it was Terry 'quad3d' Wang that actually did the video - not AMDZone.

Comments

The article ...

[Forager]

Here's the original Tom's article.

Here's the text of the new article direct from the source:

Thanks to the millions of people who e-mailed me about this.

Do we title this: We TOLD You So!
or do we title it: Maybe we were right about him?
or do we title it: AMD Won't Burn a hole in your wallet, or your motherboard?
or do we title it: AMD Slaps Around Little Tom-Tom? as Van Smith put it?

Well whatever you want to call it, there is a new video out on the internet showing what REALLY happens to an AthlonXP (or MP) when your heat sink falls off, or when your fans fails. Now I'm not saying that Tom Pabst is completely wrong, but let's just say this makes his testing methods look a little 'suspect' at best. Here's a quote from Van Smith:

A video has appeared on the Internet countering a huge dose of FUD my former employer inexplicably dumped on his readership a couple of months back. The new video, with AMD credits all over it, is entitled "How an Athlon(tm) MP 1.2GHz Really Copes with Heat Emergencies." The piece demonstrates the AMD Palomino Athlon subjected to brutal circumstances such as heat sink removal while playing Quake III and boot up attempts when a CPU cooler is not attached. In all scenarios, the Palomino comes out unscathed. A similar though much less thorough test came out with unsurprisingly different results at Tom's Hardware. Ouch! Looks like dispensing bad medicine can result in a mouthful of looser teeth. Good job Ben & Joe, perhaps you can also give THG a crash course in analyzing computer technology.

The video also presents the much more realistic situation when the CPU fan fails. In that case the Palomino continued to play Quake III for several minutes before shutting down. Again the chip was undamaged.

This video may look like it's from AMD, but I'm pretty sure it's not, even if it makes a great case. The original download site appears to be down, so I mirrored it onto AMDZone right here.

Other mirrors: Mirror.

Update: One of our readers e-mailed me to give me his first hand experience with the thermal control capabilities of the AthlonXP:

I installed my amp1800 (sic) with the heat sink rotated 180 deg. and after 3 hours of trying, incessantly, and not being able to boot I found the problem, I rotated the heat sink and all is fine. I must say, I was sweating bullets when I found the problem.

No fried chip, no smell of burned silicon. Looks like the thermal diode is working.

(end article)

~Aaron.

Re:The article ...

[ChrisMul]

The video demonstrates an AthlonXP processor. Tom was testing the "pre-XP" version. This leads to a VERY unfair comparison...Tom directly mentions that the XP will use the same core that the dual-capable Athlons do...this core does have built in temperature management. The Pre-XP processors DID NOT and WOULD FRY if left unhelped. This doesn't demonstrate that Tom's test was wrong or his results invalid...this shows that AMD knows they screwed up but were at least smart enough to fix the problem in the next significant rev of the chip (though that doesn't fix the situation for those of us with Pre-XP chips)

Re:The article ...

[slamb]

HardOCP.com ran an article the other day that said they were installing an 1800+ XP processor, and reached across the desk to grab something, bumping a key on the keyboard, which booted the machine with no heatsink and destroyed the processor.

You're saying that they worked on a powered machine and blamed AMD when they fucked it up.

ATX power supplies, unlike AT power supplies, run power through the motherboard whether the system is on or not. Some newer motherboards have a giant green LED in the middle of the board to alert you of this fact. (It even glows for like five seconds after you pull the power, since the capaciters take some time to discharge.) This is what makes possible the keyboard power-on feature that you described, as well as scheduled power-on and such.

You never work on a machine that is receiving power. Since they made such a basic mistake, I wouldn't trust their diagnostic skills. Don't believe them when they say heat fried the processor.

(Side note: while you don't want the machine to be powered, it is a good idea to make sure the machine is grounded to avoid static damage. For this reason, I'd suggest plugging it into a power strip and turning off the power strip. Ground but no power.)

the probability that your heatsink is just going to fall off is pretty low anyway. If the lack of heat protection keeps the cost down, then I'm all for it.

Agreed. Someone made the comparison between running a processor with no heat sink and running a car with no coolant. The car isn't exactly going to like it, either. Why do we consider this such a problem for PCs? Heatsinks don't just fall off. If they did, it couldn't be good for the system anyway to have a huge conductor rolling around on your motherboard.

Toms review

[OmegaDan]

If you read the toms hardware review, it said the athlon MP/XP thermal diode were capable of detecting no more then a 1 degree/celcius per second change and no more ... they've carefully engineered this "test" so the rate of change is below that threshold... toms test was worst-case and thats what you have to plan for.

Re:Toms review

[Strepsil]

Actually, if you read through the article (finally found it), it was a Siemens engineer that claimed the thermal diode was the problem when the CPU fried on one of their boards. AMD was not consulted.

It could well be a case of passing the buck because they didn't want to blame their motherboard which may not have implemented support for the thermal diode properly. That's my conspiracy theory, anyway.

Re:Toms review

[James_G]

they've carefully engineered this "test" so the rate of change is below that threshold

I'm sorry - how did they "carefully engineer" it so the rate of change is below the threshold? It looks to me like they pulled the heatsink straight off, which is exactly what was done in the Toms hardware review. So now you've seen two basically identical tests product completely opposite results. Who to believe eh?

Personally, I couldn't care less. The odds of a heatsink accidentally dropping off are pretty slim anyway, so I'm not going to lose any sleep over it.

this has nothing to do with the CPU

[Blymie]

All this shows is a motherboard with sufficent onboard temperature monitoring, shutting down because the CPU is over heating. It has nothing to do with the CPU having any sort of built in temperature throttle like the P4 has (or even the 68040 had). Tom's Hardware and this video are showing two completely different and seperate things. One shows how a motherboard reacts in order to save an amd cpu (this current video), and the other shows what happens if the cpu is left to cope with a heat problem on its own (Tom's tests) without the motherboard stepping in to rescue it.

I wish people would at least FILTER some of these posts. Its seems that /. just doesn't care some days... or maybe its just on slow days ;).

Re:whatever

[GeekDork]

The bad thing is that there are so many ways to screw up. You could get fooled into using one of those neat spacers which might happen to be too thick, resulting in inadequate cooling, thus frying the processor. The Athlon processors can burn before the POST is through, so if your heatsink has fallen off or shifted during transport, you're up for some quick, expensive fun.

Also, you could be one of those freaks using a water cooling system (please shoot me when I start pumping water into my computer!). There's a nice article over at Dans Data about a burnt Intel Celeron(!!) after a cooling failure.

A third problem is the limited use of most monitoring programs or a bad configuration. Motherboard protection can be configured so that all you get out of a failed fan is a nice beeping noise. I tell you something: no one can hear you scream in a dark cellar. I have an old Slot A Athlon 750 running nonstop, and if the fan fails, I'll most probably lose it. The only solution would be to have a monitoring program slow down and/or even shut down the system. Some BIOSs/Mainboards don't support this, so it would be pretty much impossible to prevent an incident directly after switching on, because when the speaker starts beeping it might be too late.

So far, the only solution I've seen to this is an extension to the VapoChill rigs. The system is held in reset state until the cooling system has reached its (sub-zero) working temperature. Only then, the reset bridge is opened and the system is allowed to start up. I have not seen any comparable functionality on a mainboard so far and I don't know whether it would actually help or if the processor produces enough heat even in reset state.

I don't trust Tom...

[UnAmericanPunk]

I watched this video and found it quite interesting. One difference between this video and tom's video was the motherboard and chipset that was used, that could have something to do with the different outcomes. Tom used some motherboard I've never even seen before, I had to hunt the net to find info about it. The new amd video uses a motherboard with the amd 760 chipset.

My other issues against tom are the benchmarks that he uses now, compared to what he used to do. Last review I just remember seeing the bapco and quake3 tests. Which is interesting because of all the rumors that bapco and intel are the same company. Read here. Then the quake3 benchmarks... umm, wasn't that what the P4 was optimized for? I remember back when the K7's first came out and tom threw a barrage of tests at it... a good FPU one was the 3DStudio one (which I didn't see in the last test).

Lastly, I remember the massive intel ad banners on the site when Tom did the 2Ghz P4 review...
So in my eyes, Tom sold out and I don't trust his reviews anymore.

Re:Correct me if Im wrong...

[ryanr]

No, Tom's results show that when the heatsink is removed, the blue smoke gets out. The AMD video shows the machine shutting itself down. The differences is that the Intel processors simply slow way down, but keep running.

AMD's test is apparantly usings a thermal shutdown sensor, which shuts the power down when the CPU gets too hot. Not all motherboards have those, and they can be disabled in software. Intel processors apparantly have a similar function built-in.

Re:Third party verification

[aiken_d]

I've been following Tom's Hardware for years now, since the days of truly atrocious english translations. I still check Tom's once a week or so for news and updates, and I find it a valuable resource.

However, to me at least, he seems far too emotionally attached to his subject matter to deliver unbiased opinions. Over the past few years, I've seen him be zealously anti-Intel, anti-AMD, anti-Intel, and now fairly anti-AMD again (albeit less fanatically than in the past). He's decidedly anti-Rambus (as am I, but I don't write articles purporting to be unbiased).

It's a good resource site, but very prone to sensationalism and exageration when flawed test results line up with his prejudices.

When serious reporters have all of their suspicions confirmed, they intentionally calm down and redouble fact-checking to make sure they don't embarrass themselves. Tom doesn't seem to have that concern, so I always read the site with a grain of salt and awareness of what his current emotional attachments are.

Cheers
-b

A Computer Shop's point of view

[Murphy(c)]

I work for a local computer shop, and although we are no Dell or Gateway, we've seen a lot of weird hardware stuff over the years.

And the fact of the matter is that AMD CPUs can really burn up just like in Tom's Video.

We've seen more than once, a customer coming back with what he said was a defective AMD CPU, and when we check the CPU, we could see the adhesive barcode that we put on the underside of each, partially burned out !

And don't get me started on the damn fragility of those AMD CPU, we've got dozens of them broken because some guy slipped when installing a fan on them.

Murphy

Hardware p0rn

[Codeala]

I object to this type of video trash. They provide an unrealistic, degrading portrayals of CPUs that take off their heat sinks/cooling units for no apparent reason! Like that sort of things happen all the time in real life.

I don't care who is making them, it must be stopped. Will someone think of the children!

Another test.... .45 to the core

[Catbeller]

I propose a new test: strap the mobo+Athlon to a wall.

Scientifically fire a .45 into the heart of the processor. Run performance test.

Tests may show that the Athlon does not hold up under impact of a projectile. A video of this process may be necessary to prove the point to the skeptical.

Naysayers and Athlon proponents may argue that this test does not reflect real-world operating conditions, but who cares -- it's a great video.

Always looked to me like the heatsink compound...

[Tom7]

When I watched that TH video, it struck me as odd that the intel chip had no heatsink compound and the athlon did. Personally, I think it was that compound burning, not the silicon or housing...

Re:whatever

[harlows_monkeys]

Sure a fan can fail, but there are monitoring systems that tell you when this happens...

Perhaps you should read the referenced articles before posting?

1. Many people find they need a rather big (and heavy) heatsink to keep an Athlon cool enough. These heavy heatsinks can fall off.

2. The CPU thermal sensor used on Athlons cannot respond to temperature changes faster than about 1C/second. Tom's tests showed the CPU melting in just a few seconds, so your monitoring software would do nothing to help you in that case.

Motherboards can handle the problem

[Twid]

I just got a new Athlon 1800+ CPU with an ABIT KG7-RAID motherboard. It had a thermisistor for the CPU with good default settings, and it has a CPU FAN RPM detector. In the BIOS you can turn on a feature to shut the system down if the CPU fan fails, and you can also set alarms for CPU temp.

So, given that, I'd rather have the better performing and cheaper Athlon system. The risk seems minimal and by the time time you pay slightly more for the Intel CPU, the Intel mobo, and the Intel Rambus RAM, you're paying a lot more. My personal opinion.

- Twid