How hot is too hot for CPUs?

[AsphyxiA]

Although overclocking is not something new to me, i still have never asked or figured out what is too hot for a CPU! Usually motherboards now-a-days can shut down the system if it reaches above 60C. Is it at least moderatley safe to go beyond this temp? Will i get micro- fracters on the chip if I go beyond this? I have seen reviews of people going well above 60C but I'm still weery.

[W1zzard]

100°c i would say

[DR.Death]

ya a 100°C but how menny fans do u have in your case?

[Zurb]

Depends on the cpu used.
At least a Newcastle socket 754 3000+ survived 118 C :D
But I should never let it run normally over 65 c

IT totally depends on the cpu used

[AsphyxiA]

ok well then what is the max temp for my X2 4200+ then if you would know. Would this be similar to other A64 cpus?

[wtf8269]

Your X2 will run hotter since it has two cores I wouldn't really worry about anything unless it is staying above 60º-65º for long periods of time. I would try to have your load in the 50º-55º range. However, the lower the temperature the better.

[Steven B]

keep your temperatures as low as possible, higher the temp the more unhealthy the operating conditions become.

[giorgos th.]

for both cpus (intel-amd) a temp lower than 50c is very good for their lifetime.

[trog100]

kinda funny really but the folks who worry how hot their cpu is are probably also the kind of folks who upgrade em long before they wear out anyways.. the kind of dudes who just run them as they come (the vast majority) in poorly ventilated cases with stock coolers covered in crud probably keep em for years and have no problems..

all i know is that these things will run a lot hotter without problems than a lot of what u read in places like this suggest.. the same applies to grfx card chips..

trog

[W1zzard]

kinda funny really but the folks who worry how hot their cpu is are probably also the kind of folks who upgrade em long before they wear out anyways.. the kind of dudes who just run them as they come (the vast majority) in poorly ventilated cases with stock coolers covered in crud probably keep em for years and have no problems..

all i know is that these things will run a lot hotter without problems than a lot of what u read in places like this suggest.. the same applies to grfx card chips..

trog

actually you are right .. but if you tell people temps of 120° or even a lot higher they tend to not believe you.. so say 100°c and everything is good :)

[trog100]

they wont believe your 100c either wizzard.. he he he.. i also wonder just how many folks have all the fans going flat out and roaring away just to keep the temps at idle down a degree or two.. to me the only temps that really matter are the under load ones.. whether your cpu temps at idle are 30c 35c or 45c dosnt matter in the least.. how quiet it is should be the order of the day here not how cool.. course the more air u have blowing thru your case means the quicker it fills up with crud as well.. i would love to be able to switch of all or most of my fans at idle or low load.. no fan control software seems to let u do that thow.. this is an anti-crud matter more than anything else to me.. keeping the crud out means keeping the important under load temps down.. he he he..

my computer spends 90% of its time doing bugger all much.. it would be quite happy with most of its fans off.. 90% less crud..????

trog

[AsphyxiA]

ok well this gives me a little more confidence. still alittle worried but what the hell, i bought this hardware because i could tweak it. plus i got a new toy for christmas that should keep things even cooler! thanks guys!

[G.T]

Simple answer = get a decent CPU cooler (Zalman are a great investment) and if you are worried about "crud" in you system dragged in by fans, they do have these things called "filters" you can use you know. ;)

End of the day heat is the biggest killer of systems so trying to keep the temps inside low is always preferable.

[trog100]

i already have a zalman 9500 and an artic cooler for my grfx card.. and yes i am still worried about crud.. and filters dont go well with pretty modded cases with glowing case fans.. plus decent filters really do block up in no time flat and they really do massively restrict airfow when its needed at full load.. to get the same airflow even thru a clean filter neeads a fan thats 3 x as powerfull and 3 x as noisy.. which kinda defeats the objective if quite cooling is waht u are after..

the answer might seem easy for u dude but it dont seem that easy for me.. sorry..

my approach at present is woking fairly well.. my case is less than 2c above room ambient.. the whole lot runs near silent and pretty cool and the fans are all temperrature controled and mosly running at minimum speed.. that just leaves me the one problem.. the build up of crud.. he he he.. even zalmans dont work quite as well covered in a thin film of dust.. in fact zalmans cos of their massive fin area plus close proximity of fins and slow airflow suffer more from dust build up than the old fashioned type of noisy fast airflow cooler..

my system has only been together about two weeks and i can see the dust building up on that pretty zalman cooler already.. shame really.. he he..

trog

[Poisonsnak]

I think there is a lot of error between motherboard sensors (I think 15°C would not be uncommon) but if you want it straight from the horse's mouth it appears to be 65°C:

http://www.amdcompare.com/us-en/desktop/details.aspx?opn=ADA4200DAA5BV

[trog100]

AMD Athlon 64 X2 Dual-core.. max temp 65c..

seems very low that one.. too low in fact.. if i had one of those particular cpus with its stock cooler average case and didnt live in iceland i would be worried.. in fact if that really is a safe max temp i certainly wouldnt buy one..

trog

[Poisonsnak]

Hehe yeah 65 is a tad low. I have a 4400+ and at 1.45V / 2.5GHz I hit about 60 or maybe 62 at full load, I think AMD posts those kind of numbers because they know how inaccurate those temp sensors can be.

[trog100]

let me try a theory poisonak.. AMD an Intel need to keep selling us new CPUs.. for a long time they have been doing this by ramping up the speed of single core chips.. die-shrinks and lower core voltages have helped em do this.. the last prescot die-shrink for Intel was a disaster.. it didnt work.. the new smaller die if anything ran hotter than the old one.. end of line for selling us faster single core cpus.. mainly cos they are running well too hot.. 80 or 90c for these things probaby isnt unusual..

sooo.. along come multicore cpus.. at present by slowing core speeds down a little they have seemingly managed to have two cpu chips that dont generate much more heat than one.. we all think gee-whiz.. twp of these things must go better than one so they sell us a few more new cpus..

my logic tells me that if one cpu core on a chip is causing heat problems sticking two similar cpu cores on the same chip is gonna cause even bigger heat probelms..

so how they getting away with it.. simple really.. most of the time one of those chips sits there doing exactly bugger all.. and is likley to for the next year at least.. the new xbox has three cpu cores.. two sit there doing bugger all.. the new xbox is known to run too hot.. ????

sooo.. as soon as someone comes up with a game that actually uses all three of the xbox cpus and puts em all under load.. the xbox is gonna melt.. he he he..

i recon microsoft are just gambleing that by the time someone does write such software IBM will have come up with some new cooler running chips and the newer xboxes wont melt.. time will tell..

either way i recon both amd and intel are relying on the fact there aint no software out there at present that will make their dual core chips work hard or do anything much at all.. if someone did actually make two cores do twice as much work as one core the heat problem would rear its head in a pretty dramatic way.. by stipulating a very low 65c max temp for the dual core cpus they are in essence covering their tracks.. course quite how u are supposed to keep those two heat generating cores down to a max of 65c when software really does start making em both actually do something they aint letting on.. he he he

trog

[Solaris17]

^ u have a good point.

[Poisonsnak]

Yeah that is definitely a good point. Plus motherboard sensors seem to vary so much from board to board that a reading of 65 may actually be 80 or something.

I sure noticed going from the single 1.8/512k 3000+ to the dual 2.2/1M 4400+ the idle temps are maybe 5 C higher or so but man the load temps rocketed up from 45 to about 60 and I ran both of them (the 3000+ and the 4400+) at 2.5GHz when I had them.

[trog100]

bit more on the "how hot is too hot" thing..

"During all this, we kept heading back to the Intel BIOS to check the processor temperatures, and we were surprised to see that it really hadn't changed that much. The Pentium EE 955 runs hot to begin with, and other than a sharp increase to the heatsink-fan, we didn't really notice a sharp jump on any of the indicators. The stock temperature at idle registered at 70-75 degrees C for the processor and 42-47 degrees C for the internal temperatures. When clocked at 4.27 GHz, this increased to 82-87 degrees C for the processor and 50-55 internal. Some of these results are high to say the least, but through all of this, the Pentium EE 955 was incredibly stable and seems to be able to take a high thermal operating temperature."

full review.. the yet to be released 65nm intel top end dual cpu..

http://www.sharkyextreme.com/hardware/cpu/article.php/3261_3573786__11

trog

[fullback43]

before i had my nice liquid cool i had all fans and someone push my computer bacl against a wall the cpu temp was 118c and the everything else was 115c. my computer ran fine

[DR.Death]

it does but it shorten the life time very significantly

[orionpc]

my cpu is a 2400+ runing at 1.7gb but i can overclock it and run at 2.4gb and run at 47c is it good ?

[DR.Death]

ya that is good i am runing mine stock and i am around 41°C (p4 3.0 ghz)

[trog100]

a little connundrum for u DR death.. take a san diego 3700+ which runs at 2200 and san diego 4000+ which runs at 2400 and a san diego FX57 which runs at 2.800..

now assuming the FX57 is gonna run hotter than the 3200 or 4000 do u recon its gonna die a premature death.. or which one of the above chips is gonna live the longest do u recon.. ???

and how long do u recon your chip at 41c is gonna live.. ???

and what about the new intel 65nm dual core chip which appears to run 72c at idle.. how long do u recon that one is gonna live..??

trog

[lascar]

lifetime of the processorcore :30 years
u can overcloked it up to 45%...indeed the lifetime of the core will be shorten (3years(o/c55%)<------->10years(o/c 25%)..but anyway u change ure hardware every 2 years... even every year if ure a g33k!... Life goes On! and i can't help burning my hardware...

[infrared]

I flashed my bios about 3 weeks ago only to discover the true temperature of my cpu to my horror! It was running @ 3.7ghz... 73*c only 2c of thermal throttling...

I went to a nearby computer shop and bought the gigabyte 3D rocket cooler! (kicks ass!)

Anyway, temperatures are now 65c @ 3.85 after an 8 hour torture test with 2 instances of prime 95, which showed it to be stable.

I'm now testing my cpu @ 4ghz, and temps after 30 mins are at 67c with 2 instances of prime 95... looking good so far!

I'm really not bothered by how long my cpu lasts, if it dies in only 6 months or so, that would be fine for me... just means i'll have to buy a quicker one and overclock the balls off that!

[trog100]

"lifetime of the processorcore :30 years
u can overcloked it up to 45%...indeed the lifetime of the core will be shorten (3years(o/c55%)<------->10years(o/c 25%)..but anyway u change ure hardware every 2 years... even every year if ure a g33k!... Life goes On! and i can't help burning my hardware..."

that one just has to be a figure someone has pulled out of thin air.. ???

he he he.. okay then.. when is overclocking not overclocking..

"take a san diego 3700+ which runs at 2200 and san diego 4000+ which runs at 2400 and a san diego FX57 which runs at 2.800."

is the 3700 and underclocked FX57 and will it live for 90 years or is the FX57 an overclocked 3700 and will it have a very short lifespan..

i recon the point i am trying to make is that we really aint overclocking a lot of these chips at all.. its amd who for marketing reasons have underclocked the buggers in the first place.. all we are doing is making em go at the real speeds they are capable off..

the only chip thats for real.. in amds case is the flagship (we are faster than intel pr chip at silly prices) FX57.. all the other single core san diego chips for example the 3700 and the 4000 are pretty much the same as the FX57.. just rebadged and reboxed and sold at prices people will actually pay.. its quite clear no one actaully buys the FX57 at its silly over $1000 dollar price.. he he he.. they probably just send em out to reviewers for the pr launch.. he he he

as i say.. when is overclocking not overclocking.. ????

trog

[Steven B]

a little connundrum for u DR death.. take a san diego 3700+ which runs at 2200 and san diego 4000+ which runs at 2400 and a san diego FX57 which runs at 2.800..

now assuming the FX57 is gonna run hotter than the 3200 or 4000 do u recon its gonna die a premature death.. or which one of the above chips is gonna live the longest do u recon.. ???

and how long do u recon your chip at 41c is gonna live.. ???

and what about the new intel 65nm dual core chip which appears to run 72c at idle.. how long do u recon that one is gonna live..??

trog
First off, the FX57 will run hotter because it was designed to run hotter, if a processor is left alone to run as the manufacturer specified then everything is fine if you install the heatsink correctly. Now about the 65nm dual core intel chips, they have high temperatures because they were designed to withstand that much heat, just liek the dothan was designed to withstand operating temperatures of 100c. There is no temperature at which a processor is safe or unsafe when you are comparing all processors. Like you cant compare northwood and presscott temperatures in terms of safe because the prescott can run 10c hotter and still be safe. You cant compare differnent brands and cores in terms of their ability to safely handle heat.

[Poisonsnak]

I agree that you can't compare 2 different CPUs in terms of stability at given voltages or temperatures but I think what trog is worried about is long-term reliability, and the fundamental behaviour of transistors.

If you look here:
http://en.wikipedia.org/wiki/Transistors

The maximum junction temp for any silicon transistor is 150 to 200 C, I remember the 150 figure from my electronic materials class back in the day. I also remember a 10 C increase in temp will halve the lifespan of a transistor, so a 20 C increase would quarter it, etc. These are pretty fundamental properties don't you think? My quick googling didn't yield any useful information to confirm what I said about transistor lifespans though (this is me remembering stuff on an exam I studied for 3 years ago)

[trog100]

"First off, the FX57 will run hotter because it was designed to run hotter"

that one is just another assumption.. based on another assumption.. that the $1000 FX57 is actually a different chip than the very much cheaper 3700 or 4000.. ????

all the san diegi chips will come off the same production line.. they are in essence the same chip..

the offiicial story being that they are all tested to see what speeds they will run at and then clocked badged priced and boxed accordingly..

this of course is total bullsh-t.. if they get it right and have a better than expected yield the whole damn bunch of em could end up being capable of FX57 or very close to FX57 speeds..

so what do they do.. sell a whole bunch of super fast chips and expect everyone to pay silly $1000 dollar prices for em.. or as an alternative ship a whole bunch of $200 cheapo super fast chips and wreck the cleverly arrange cpu to suit everyone priceing system..

nope what gets badged and boxed its entirely governed by market demand.. and market manipulation.. which quite clearly is gonna demand a whole bunch more $200 chips than $1000 ones..

as i say the only real cpu that is running somewhere near its max speed is the top end flagship one.. try taking a FX57 san diego core much past its out of the box speed and it will bomb on u.. 3 gig is about the max for these chips on air.. and my $300 3700 san diego chip stands pretty well as much chance of getting there as the $1000 FX57 does..

we are being conned by clever market priceing.. both with CPUs and grfx cards.. dont get me wrong i aint complaining about being able to get $1500 perfornace for a $500.. but the fools who are paying top price for these things should be..

trog

[Poisonsnak]

First off, the FX57 will run hotter because it was designed to run hotter
I'd say that refers to the FX-57 being clocked higher from the factory than the 4000+. The FX-57 runs 2.8GHz and the 4000+ runs 2.6GHz therefore the FX-57 is hotter. I would fully expect both chips to generate the same amount of heat at the same clock and voltage.

[Steven B]

i dont even knwo if it runs hotter, if it did at stock compared to the others at stock then it was designed to meant to run hotter.

Thats how AMD rolls, intel doesnt take a chip and just change its ettings to make it run differnt, intel actually cuts differnt parts of silicon made for the production of differnt chips. A 3.6ghz processor has much better silicon than a 3.0ghz processor, thats y you see a higher oc on a 3.6ghz processor than a 3ghz processor. And many times they run with the same heat, as there are differnt revision of the 3.6ghz one being 89watts and the 3ghz being 89 watts.

[trog100]

i have been overclocking.. cpus and grfx cards for the last ten years.. when these things were actually what they claimed to be they use to be sold with something like a 10% safety margin speed-wise .. u could use this safety margin to get a little more performance out of a part but only a little..

the idea of taking a $200 part and with few tweaks giving it the same performance as a $1000 part was unheard of.. the $1000 part was really a different beast then the $200 part..

now i buy a $200 part and with a few tweaks get it to perform better than a $1000 part.. wise up dudes.. the only reason i can do this is cos both parts are the same parts.. just badged and sold at different prices to suit different market segments..

this dosnt apply in all cases.. sometimes the part is a more expensive part deliberately crippled and cant be uncprippled.. sometimes its just software/bios crippled and can be easily uncrippled..

sometimes it really is a cheaper part.. but even the "cheaper" part will come in several difference price range versions.. and the "cheaper" version of the cheaper part is more often than not a cripples faster version of the same part..

overclocking nowadays is about knowing what part to buy.. the general rule being never buy the over-priced clocked to within %10 of its real performance abilities part.. cos if u do u are being ripped off and just chucking your money away..

these kinda parts should be just reserved for the likes of Alienware $5000 boxes for dudes with money to burn..

when u read that the 12 pipe massively underclocked gto2 grfx card for example is a "failed" 16 pipe X850 card dont believe it.. its a deliberately "bios" crippled real X850 xt pe top of the range card being sold of as a lesser cheaper part..

why would they sell a more expensive part disguised as cheaper part for less money.. at first glance it dosnt make sense but at a second glance it does..

they want folks to forget about yesterdays "king of the hill" and just concentrate on "todays" new range.. in short they want them off the market as quickly as possible.. they could just chuck em in the bin.. but its more economical to sell em off "crippled" disguised as a cheaper part to the mass market segment..

course its getting so bloody obviouse what they are doing that its gonna cause problems.. folks who pay "top" prices for "top" performance aint gonna be overly happy when when they find out they could have similar perfromance for a fraction of the money..

AMD tried to stop this tweaking or cripple undoing a few years back by locking cpu multiplyers.. when a fsb speed was for real it worked.. now cos its possible to crank up fsb speeds to silly figures it dosnt work.. it seems the mobo makers and the cpu makers arnt exactly playing the same game.. which for us "uncripplers" is good news.. he he he

"uncrippling" dont quite have the same ring to it as "overclocking" thow does it.. he he

i spose i have answered my own question.. when is overclocking not overclocking.. ie.. when its "uncrippling"..

trog

[lascar]

;) :laugh: :toast: :banghead: :D :o :p Trog100 :"these kinda parts should be just reserved for the likes of Alienware $5000 boxes for dudes with money to burn..

when u read that the 12 pipe massively underclocked gto2 grfx card for example is a "failed" 16 pipe X850 card dont believe it.. its a deliberately "bios" crippled real X850 xt pe top of the range card being sold of as a lesser cheaper part.. "

true im agree with u , "sold at different prices to suit different market segments.."(: D Marketing).....boyS the only thing they care isn't performance & reliability:nutkick: but ure MONEY:pimp: :ohwell: :eek: ---) save ure money..stop wasting it ... o/c ure hardware smartly!

[SyK1]

hey i have an p4 prescott 3ghz im running it at 3.3ghz its 53-55 celsius att load but it idles at aprox 45 degrees is this ok, or should i change the cooler?

[infrared]

yeah, that seems reasonable. The idle temps on the stock cooler are always fairly high due to it's crappy design.

I've managed to ger my prescott 3.2ghz up to 4.2ghz with a diferrent cooler (gigabyte 3d rocket) and the highest temps i've seen are 53c and it idles at 35c! :cool: So you will be able to get a lot more out of your cpu safely. <=== key word.

I wouldn't try going for a 1ghz overclock though, it is a tad excessive. I run this thing 24/7 at 3.85ghz, and just shove the fsb up when i need it. :rockout:

[trog100]

gonna stick my neck out here and say that.. all the latest intel chips are "too hot".. their last two die-shrinks have failed abysmally to get the heat down.. in short they have major problems..

when u read reviewers having to remove the stock intel heatsinks and fit after-market zalmans to get the things to run stably enough to benchmark it tells its own story..

intel seem to have major problems which aint been fully picked up on yet.. over the coming few weeks i recon they will be thow..

also as their chips run hotter.. they seem to be stipulating lower max operating temps for em.. keeping the "hotter" chips at these "cooler" temps aint gonna be easy thow.. he he he

AMD seem to be cruising.. intel seem to be sinking fast.. and the reviewers are being kind to em.. not sure how long this will last thow..

so the answer to "when is a cpu too hot".. could simply be "when its an intel" and trying to "leap ahead".. he he

jeeesh i am beginning to sound like an amd fanboy in the making.. ????

trog

[Steven B]

intel has also disabled the avanced halt state (C1E) as well as EIST, so there is no way to take down the multi to 14 on htese new 900 series. THey run extremely hot from what ive heard from people that have the retial version.