Slashdot Mirror
The Joys Of Losing Your Cooling Device
nitecreep wrote to us about
Tom's latest article: What happens to procs when the heatsinks fall off?. Having just had my brand new fan stop working on my computer, I can sympathize. I've found that it takes my 1.2 Ghz Athlon to reach 80 degrees Celsius in about 6 minutes, from time of starting machine. The results of running without a heatsink at all are....interesting.
gives a new meaning to firewire...
yea. i had a quantum fireball burst into flames before...
lost a mere 20 gigs of data.
--donabal
Safety First Day?
I remember I could tell if the 386 was on or not by how often the Centralized Air conditioning came on. My Celeron 433 runs so cool now I keep it under a blanket for noise protection. Ah....Peltier devices are so cool. (no pun intended). My question is: When are they going to come up with a heat sinking device that runs like the engine block on a car (I.E. the water/freon/liquid nitrogen/liquid helium/butane actually flows in channels built for it within a heatsink block)
JoeLinux
are Marshmallows, Grahm Crackers, and Chocolate. Be sure to invite all the slashcrew to share your s'mores!
<This .sig left intentionally blank>
I can sympathize too. I cracked a corner off my Athlon 1.2 (266) over a hundred bucks ago (back when they were $250). When it didn't boot I decided to see how hot it would get without the heatsink. I turned it off as soon as it started to smoke, but like any idiot I just had to see what temperature silicon starts to smoke at. Luckilly I had a calibrated measurement device, my finger. That T-Bird burned the heck out of me, and you could almost make out the "A" branded into my finger!
I've also just cracked the core on my current 1.33 T-Bird, and I've just picked up a 1.4 at lunch today. Is this some sort of marketing scheme by AMD?!!? I figure they're sticking it to me hard enough with the way their stock is plummeting, every point making my Tahitian retirement much more distant.
Cap'n BryWell there was a gap alright. The heat sink had fallen off and was lying on the card beneath it.
After applying heat transfer goop and bolting it back on, things have been running well.
I've got kind of the opposite situtation - a laptop that runs really hot. I'd like to slow down the CPU (300 mhz would be plenty) to allow it to run cooler, which might hopefully also make the battery last longer.
Does anyone know of any utilities? I don't think the BIOS will allow me to set the CPU speed and multiplier. It does support SpeedStep - is there a way to force speedstep on always?
I had just put together a new box, and was attempting to compile a brand new kernel on it (2.0.36 at the time). About 5 minutes into the compilation, I started getting reams of segfaults and I could not, for the life of me, find out why. Later, I discovered that when putting everything together, I had forgotten to plug in the power cord for my CPU fan. Nasty shiat, that is.
Not a good way of reporting the problem :)
My GeForce fan stopped working. It destroyed the GeForce card, worked intermittently while I troubleshot things, and then took the motherboard out with it.
I woke up one night at 3am thinking my pager had beeped and gone off...when it hadn't I was confused...10 minutes later...I think the same thing..but it's in the living room.
turns out that was a heat warning on my motherboard. The fan for the cpu had gotten wore out and was dying...I killed the box and bought a new fan the next day...but I could have fried my processor if the box hadn't started beeping and woken me up.
----------
ah honey, we're all resplendent - Bill Mallonee
This isn't a particularly good time to be putting that phrase in your communications...
Best Slashdot Co
While working as a network admin once, I had a processor burn out without its heatsink. Smoke started rising from the open case, so one of my comrads in arms decided to 'put out the fire' with his bottle of Coke.
I've never seen a machine burn so brightly. We were lucky (or maybe not) that the building's sprinklers were on the blink.
And I live to tell the tale....
Beware the Whyte Wolf.
With a gun barrel between your teeth, you speak only in vowels...
There are freeware programs available for Windows (probably Linux as well) that monitor the temperatures reported by your motherboard. Some like Motherboard Monitor will actually shut down your computer if one of your sensors report a temperature greater than a threshhold that you can set.
I'd link to them, but I believe that linking from Slashdot to websites hosting small free projects like this is cruel and inhumane. Go do a search and download it from one of the mirrors.
I recently bought an Agilent Arcticooler for 1.4 GHz Athlon. It is small, light, and quiet. It also cools the CPU better than the Taisol heatsink AMD ships in their retail package. I paid $44.
"I've found that it takes my 1.2 Ghz Athlon to reach 80 degrees Celsius in about 6 minutes, from time of starting machine. The results of running without a heatsink at all are....interesting."
One time my stopped working heatsink and the to go first thing was spell/grammar check my.
324006
I've noticed that any stock fan will fail anywhere between 2 and 12 months. I'm listening right now to a socket-370 fan making all sorts of noise. I have about 15 similarly-dead fans lying around, from CPU to power supply to case fans...
The only fan I've had for more than a year that still functions is the Antec PIII dual-fan cooler in my desktop system. Unfortunately I'm too lazy to buy quality fans for the other boxen.
I wish higher quality fans were included in things like power supplies (which are a pain to replace, not to mention dangerous), and especially ones bundled with CPUs. I also don't find many fans actually connected to the motherboard (for monitoring etc); usually they connect inline with a drive power lead...
I once overheated a K6-233 because a drive cable was resting on the CPU fan. Worked for weeks until I was compiling a kernel...
Unfortunately, PCs these days are made as cheaply as possible, with cheap fans, flimsy cases, etc.
NGWave - Fast Sound Editor for Windows
I remember when Patrick Norton who co-host the "The ScreenSavers" on TechTV forgot to install the heatsink for UGAM 3.0.
Article link below
The Dish: The UGM Incident
"Windows - A thirty-two bit extension and GUI shell to a sixteen bit patch
to an eight bit operating system originally coded for a four bit microprocessor
and sold by a two-bit company that can't stand one bit of competition." (Anonymous USEnet post)
Apple is like a strange drug that you just cant quite get enough of they shouldnt call it Mac. They should call it crack
I had an older SMP box (dual MMX-233's). The procs I found shipped with super cheap-o fans. They both went dead in a few weeks after ordering (only the CPU's themselves were covered by the warrenty...). I hadn't noticed at first. I started wondering why KDE (really, the OS in general) was starting to run VERY slow.
I hit every newsboard I could find looking for suggestions. Every suggestion was a flop. One day, I cracked open the case (I forget why exactly) and had a found esentially a microwave oven inside the box. WTF?!? I thought. I quick scan of everything showed me the 2 fans just sitting there jittering (not spinning). DOH!
(No, I don't don't work for these guys)
3d-cool.com has a great selection of cooling things for just about anything. I've since ordered a ton of stuff from them. Fast and reliable, they are. I ordered a couple of the super-duty fans for the older slot-CPUs and the thing ran great! A bit loud but...
The SMP box is now collecting dust (but I know it's 100% ready for mnore when I need it)since I found a Super-Worth box for real damn cheap at an EggHead Auction.
I'm not a prophet or a stone-age man,
I'm just a mortal with potential of a super man.
that should say 6 seconds, not 6 minutes. I believe AMD's spec is that a K7 will die in six seconds without any cooling.
On the two AMD DDR boards that I've played with, there is a new BIOS setting that allows the motherboard to power down the machine after a specific temperature. I have mine set to kill when the motherboard's thermal probe measures over 65c.
This is a great feature, and thankfully I've never had the pleasure of testing it out. Hopefully it responds better than the on-die thermal diode of the Palominos.
--------
It's OK to be social, just don't tell anyone about it.
Well, anyhow, last week I mounted a cheap heatsink to my brand new AMD 1.3G CPU and it burned up before I really got to the post screen. Ish. A few days later my new copper heat sink showed up, but I was more afraid of chipping the CPU than having it burn up... done that too.. They (www.hardocp.com) have long since commented about the few seconds it takes to make an AMD processor keychain by running it without a heatsink, but man... don't even mess with something that may be marginal.
+++ UGUCAUCGUAUUUCU
I want that Fairlane! :)
It doesn't mean much now, it's built for the future.
60 degrees Fahrenheit? That's below room temperature, which seems a little excessive for a PC. 60 degrees Celsius would make sense however. My Mobo will automatically shut itself off if the temperature exceeds a certain threshold (I use 40 degrees C) to avoid damaging the components, but I'm pretty sure the Bios is just polling every few seconds or so for that data, in the example in the article it wouldn't be fast enough.
I read the internet for the articles.
I had the fan fail several times on a 1.2 GHz Athlon, but didn't suffer any damage. Apparently AMD design engieers considere the heatsink, but not the fan, as an integrated part of the CPU.
So the guy's pissed, wondering how he's gonna get the offending non-dividing Pentium out. Then he realizes, hey it's a Pentium. He took off the heat sink, ran his comp for a while. It did it's impression of an Easy-Bake oven and generated enough heat to melt the glue. He then removed it and got his replacement.
I'm not a specialist in cooling and overclocking and things like that, but the question I have is : if modern processors can't survive themselves "as is" (i.e. without thermal protection), why don't processor manufacturer sell them with an integrated heat sink and a fan bolted on forever, as an integral part of the product ? Even better, the heat sink itself could be bolted through the sides of the processor, and adventurous overclockers could still replace it with whatever piece of heat removal wizardry they want. If nothing else, it would force motherboard manufacturers to invent stronger CPU socket to hold the CPU/heat sink combo.
"A door is what a dog is perpetually on the wrong side of" - Ogden Nash
I couldn't follow the link, so I navigated the site a little... Not sure if tomshardware is slashdotted or if the link was wrong... in any case here's the mirror I got to, page came up immediately:
x .html
http://www6.tomshardware.com/cpu/01q3/010917/inde
I tried the same link both in the www6 and in the root subdomain, and both gave me a 404... try this link if thats still the case...
I am disrespectful to dirt! Can you see that I am serious?!
At least there's one guy who's banner ads are paying off...
:)
"well computer news aren't really running these days, let's fry a few processors"
Either that or the AMD parts are REALLY cheap.
Oh... they are
--- Metamoderating abusive downgraders since my 300th post.
Hmm. All of my servers scream via SNMP should the fans drop in RPM, increase in RPM, fail, stutter, and almost anything else that can go wrong with the hardware. Hell, if you open the case, I get a 'chassis intrustion detected' trap. BlackBerries, by the way, are wonderful for recieveing these things. :-)
Vintage computer games and RPG books available. Email me if you're interested.
Which makes me wonder why only now are heatsink companies installing dual fans? Yes the Socket A setup means smaller or boxier heatsinks, but still, it seems a no brainer the way small fans die all the time to have a backup. Heck - I have a 4 drive RAID 5 tower on one of my servers and I touch the drive trays constantly to see if they are warm meaning the small fan on the tray is probably dead - now I shell out the $$ for a drive tray with dal fans and a fan monitor circuit to alarm if one dies. I mean given the speed that these suckers died - you have to wonder if you were lucky enough to HAVE a BIOS that would shutdown on a fan failure coudl do it fast enough even if the heatsink was still attached.
I love AMD processors and use Athlons in all my machines - but what was AMD thinking when they left out the thermal diode or an overheat circuit?
Top Most Bizarre/Disturbing Error Messages
Only computer I ever fried was a Spectrum, but that was quite funny. I bought it at a car boot sale for £4 and was cycling home with it in a carrier bag over the handlebars. It went between the spokes of the front wheel, causing the wheel to saw through the plastic case and a large chunk of metal that looked quite useless.
Anyways, when I got it home I plugged it in, and it booted up with a fabulous display of white acrid smoke. Needless to say, Bubble Bobble didn't work.
This is related and I don't feel like asking Slashdot or doing a Google search so:
I have a Athlon 800 with a Giga-byte mobo (VIA KT133). As soon as I got the mobo/proc I overclocked it via the bus speed. I set it to 112, (highest speed that would boot), which got me 896 MHz. During the extremely hot summer it started locking up fairly frequently so I knocked it back down to what it was suppose to run at. I just installed a GeForce2 MX200 and 256 MB PC133 RAM and now I can't clock it higher than 104 (834 MHz). It just won't boot higher than that. Anybody have any insight as to whether or not the GeForce or the RAM could be preventing the overclocking or if I've screwed something up from running it hot in the past. I'm too lazy to swap all my components out to try and see what is actually preventing it.
> When exactly does it become too dangerous?
You can download docs from AMD's Web site.
I have one several months old that says the maximum die temperature is 90 C for Athlons < 1.1G, 95 C for faster. (I downloaded it before the 1.33 came out, so you might want to find a fresh version if you have a recent processor.)
However, the sensors are not part of the die, so it's probably hotter than what your sensors report.
FWIW, I have a 1.2G Athlon that runs about 48 C when semi-idle and rises to 56 C after several minutes of continuous number crunching, and have never had any problem. That's a pretty big margin of error for sensor vs. die temperature.
Also FWIW I started with a screamin' 7800 RPM FOP-38 fan, but I got tired of listening to it and replaced it with a 4800RPM FOP-32, and didn't notice any difference in the temperatures. (I have heard that the heat sink is much more important than the fan itself. The two named fans both come with identical heatsinks.)
Also, some say that the silver-based thermal grease is mere snake oil, but I replaced the thermal tape on the FOP-* with the s-b.t.g., and saw a drop of several degrees C.
Last but not least, make sure your case has good airflow and your room is reasonably cool. Heat flows from the die to the heatsink to the air in the case to the air in the room -- your room air is your ultimate "sink" for the CPU's heat. A big heatsink with proper sealing helps the first step, any fan on the heat sink at all seems to help the second step, the case fan(s) help the third step.
Notice that (unless you're cooking your CPU) the whole system comes into equilibrium, and you want to minimize the equilibrium temperature for each mass in the chain.
AMD actually recommends using a power supply with its air intake on the bottom, so that it will suck hot air straight off the heatsink and blow it out the back of the box, but I have never found one like that on the shelves locally. Failing that, you might want an auxiliary fan sucking out from right behind the CPU. (My tower case won't allow that, so I put an 80mm fan blowing in at the bottom of the front and two 80mm fans blowing out at the top of the back. These fans are much quiter than the power supply fan, so I don't find them annoying.)
Sheesh, evil *and* a jerk. -- Jade
I remember the day I spent trying to figure out why my box would bluescreen and reboot at random times after 1 to 15 minutes after boot. I tried everything including finally reinstalling the whole OS. Still had the same problem. I was about to start switching out memory and had opened the case to get a better look inside when my knuckle wandered into the CPU fan, I pulled back expecting a sharp rap and to my surprise got nothing, Turns out the fan was dead. Now they say a Athlon 1Ghz will fry in a matter of seconds without a fan, but mine managed to live through a whole day of reboots and installs with no damage... If anyone thinks Silver paste and copper heatsinks are nothing but a fad I say I'm sold. But more importantly It shows what weird errors you can get with a dead CPU fan, I never would expect a BSOD. I figure it had to be the CPU giving off bad data or inverted bits because of high heat.
iRepairIT - iPhone, Mac, & PC Repair
They say it can be frelly run with no fun attached. See low heat dissipation thanks to 0.13 and 0.15 micron processes
When I installed my AMD Athlon 900MHz several months ago my fan fell of for a few seconds. I went to grab my fan and my finger rubbed up against my processor for half a second as I reached to turn off my computer. My finger welted up in seconds leaving a large blister for several weeks. Infact I can STILL see the scar on my left index finger if I look hard enough.
.. but not enough to melt my skin .. OUCH ..
To say the least, I will not be putting my fingers near an atlon processor without a fan on it ever again. I was lucky I didn't fry my processor along with my finger! I figured the thing would be HOT
If you're getting 80 degrees C on your T-bird 1.4, something's deficient with your heatsink/fan. Even a lowly Volcano II will do 55. That's under full load, mind you.
I'll tell you: bad things. This is from personal experience incidentally. I had a shiny new K6-3 450MHz and the world was my oyster. Unfortunately I had also recently heard about using to peltiers to overclock processors, and so it seemed like the perfect opportunity to give it a whirl. I got a 120W peltier and by enlongating the distance between the heatsink and the CPU I was able to wedge it in there. I only fed it 5V, but that was enough. It didn't have any problems at first, but it also didn't O/C worth spit so I gave up on that part and left it on. Little did I know the damage had already begun. About a week later my system started randomly crashing, which was odd because I ran Linux exclusively. It got worse and finally got to a point where it would barely boot. I finally pulled up the heatsink and processor, and to my undying horror saw loads of brownish-green corrosion on the underside of the processor, grouped around several pins. Even worse, touching one caused it to bend easily, whereupon it fell off. We had an evaporative cooler at the time, and the combination of condensation, dust, and electricity ate through the pins like acid. I then learned that it is very difficult to solder a CPU pin back on. It would work for a while and then get flakey again and I'd re-solder the pin on, and then another fell off and I had to fight with it too. It's currently sitting up in my closet in an antistatic bag now.... Moral of the story: never use peltiers unless you know what you're doing.
"Click here to find lowest price for AthlonMP 1.2 GHz."
I'll take two!
On this subject, does anyone know of any Linux software that will monitor/report CPU temperatures? I've searched before, but with no luck, and I don't feel like disassembling one of the Windows ones...
I'm sure a lot of people here have cpu overheat stories. Mine fits with Tom's test quite well, but with a slight twist.
;)
My somewhat weird setup is a P3 in a slocket in an old slot1 BX board. This means that the system cannot monitor the CPU temp and shut it down safely.
One day after some h/w modification (I seem to recall it was moving a tv card from slot to slot) I must have accidentally jammed a power cable in the CPU fan.
After about 10 freezes in a day I wonder if my graphics card is overheating. It is as cool as ever (about 40 deg C), so I just touch the P3 heatsink, and blister my finger. So I spit on a spare finger and *tssst* the spit boils away instantly. I turn the machine off, thinking, "Oh well, I wanted to upgrade anyway".
I cool the poor fevered cpu down with strips of damp tissue paper and find the trapped cable.
An hour later I nervously turn the machine back on and hey presto, the old thing jumps into life as if nothing happened.
I'm currently planning to upgrade to an MP athlon, but I will have to have a checklist for when I muck around with anything
-Yarn - Rio Karma: Excellent
I've seen MANY MANY MANY shipped computers which had the heatsink not on the processor when it arrived at the destination. Having the processor be able to destroy itself when you loose a heatsink is just bad karma...
Not to mention that the processor obviously gets hot enough to catch something on fire if it happened to be in the wrong spot in the case.
Product liability lawsuit anyone? This makes me want to reconsider my AMD is better than intel position.
you could run a P4 without a fan, although it's probably not recommended, voids all warranties, etc. etc., and slows the thing down considerably. It'd be interesting to see how much power it needs and some performance data on this, maybe compared to notebook processors. Instead of going for MHz-Rates nobody needs and which give only very small Performance-Benefits anyway, maybe the industry should consider to think more about processors needing less power and hence less cooling.
Think not only laptops, but also desktops in a work-environment, where small, silent, easy handling and even powerconsumption is much more of a concern, than computing power. Embedded devices, where a noisy and large heatsink+fan is not an option, or even computing power per m^3 when heatdissipation starts becoming a limiting factor.
"By the way if anyone here is in advertising or marketing... kill yourself." -- Bill Hicks
I have a Athlon 1.4 and a Abit MB, and I have MBM set to poll 1/sec and it does change, so I'm sure that the chipset itself polls more than once every few seconds (I'd guess actually that the chipset does the measurement whenever it is polled).
However the -5,-12V readouts don't work, Temp sensor 1 reports 23C all the time (even under heavy load) and Core 1 reports around 0.41V (it does change), while the bios reports core 1 at 3.41V. Also Fan 3 and Fan 4 don't report anything, but do give power to the fan. Anyone know what's up? It's using the standard VIA686B chipset selection. ??
If God gave us curiosity
I have a 1.4G running at 1.48 (Can't clock it any higher : ) and it idles around 49C and tops out (after several secs) at 61C when running CPUBurn.
However I just had an idea. You can get temp sensors (external) that have a small flat plate that you can put between your CPU surface and the heatsink that will measure the CPU temp. If you could plug one of these into one of the aux fan jacks, and have it "vary" the "RPM" of the "fan" to be, say, 100x the tempurature, you could get a much better readout from your CPU temp, and have it power down the system immediately if it detects a large jump in temp. This might be about 1sec faster than getting the under-the-chip sensor readout, but 1sec may be just enough time to save your processor.
If God gave us curiosity
We booted up those systems normally, started Quake III Arena, running the NV15-demo, and then removed the heat sink.
Cool, I do this all the time!!
All of my servers scream via SNMP should the fans drop in RPM, increase in RPM, fail, stutter, and almost anything else that can go wrong with the hardware. Yeah, but how would an RPM detector know whether the fan is where it belongs or lying in the bottom of the case? 8-)
Seriously, you're smart to pay extra for good servers with lots of conditions monitored, but don't count on it to cover everything. You must also have something sensing the CPU temperature, but it wouldn't save the AMD chips if the heatsink falls off -- the CPU's heated up so fast that the one with an internal temperature sensor fried before the motherboard circuit could shut it down. SNMP would be even slower -- but if your system is slowing down or crashing because the fan died, it would probably tell you that.
Servers are a different case than desktop/home machines. I think servers are always "some assembly required", and I would hope that in the process of putting in your hard-drives, network interfaces, etc., you would take a glance at the CPU area. Desktop boxes are often sold to people who wouldn't understand what they are seeing even if they did open the case, and sometimes with warranty stickers where they get torn if you do open the case.
Have you seen these things? The processors (P3 xeons - 2 megs L2 cache) are literally the size of paperback novels. The fans aren't glued onto the heatsinks; that would be bad. Too difficult to remove/replace.
Vintage computer games and RPG books available. Email me if you're interested.