ATI GFX Boards, temperature myths uncovered

Zardon

We try to work as well as we can with any company behind the scenes on Driverheaven because it benefits the end user -- you guys, you pose your concerns on our forum and we attempt to get the answers when we can. I have been noticing alot of discussion on our forums lately from members concerned about seeing 70-80c temperature readings from the ATI overdrive panel on their shiny new X800 boards. I posed the temperature question to one of ATIs top engineers Eric "sireric" Demers, this is what he had to say.

"At over 100C or so at the die temperature, thermal issues and possibilities of permanent damage start being important (I would strive to keep it below 100C). Certainly by 125C, permanent damage will occur. Below, that, life expectancy is shortened with higher temperature, but you should be fine for a few years.

We test our boards up to 55C ambient temperature, and at that temperature, all of them work fine. Most computers have much lower ambient temperature, and the associated die temperature should be fine, except if there is a fan failure."

Ev!L-aLphA

intresting.. my 9800 gets pretty hot sometimes

__________________

R.I.P
8/8/04
4/21/08

jeffro29

my 9800xt runs at 40c idle and 45c under load

__________________
A good stable OC is a good steal in speed.

germanjulian

dam thats just hot. fry and egg on your video card! somebody do it. the cummunity will love it

__________________
/|\ Asus P5W DH Deluxe, Intel C2D E6600, 2GB Corsair XMS2-6400C4 DDR2, E-VGA GeForce 7800 GT, Creative X-Fi Extreme Music, 500GB Seagate 7200.10 SATA, Lian Li PC-V1100 Aluminum Case Black, etc. http://germanjulian.com /|\

Tipstaff

Most people bitch about changing the fan/heatsink on their brand new video card as it will void their warranty, but sometimes it just has to be done. As long as you KEEP the stock fan/heatsink you shouldn't have a problem with the warranty. Replacing the stock ATI fan/heatsink with Artic Cooling's VGA Silencer can do wonders. Top it off with some good Artic Silver or OCZ Ultra on their and you've got one chilled mofo. Even just changing the thermal paste can lower temps by about 2-5 degress (the longer Artic Silver is on there, the better it gets.. God I love that goop!), and for the most part won't void your warranty. Since ATI uses copper heatsinks on their XTs the goop is more than enough in most cases. 'Course, just don't tell them you did it

mikeycooly

Thanks for getting this info. So Under 100c is good but not recommended.
I wouldn't eat that egg
Also I think that yoke in the resisters would be bad !

mike2h

so these things are almost as hot as prescott? what is the 'myth' uncovered here. that engineer completley sidestepped the whole issue imo.

__________________

Zardon

what are you talking about?

People have been asking me if the 70c-80c they are seeing reported is dangerous on their X800 - people are concerned!, so I asked an ATI engineer what are "safe" temperatures on the boards - and asked if he would give us figures in which possible damage could occur. He replied with the statement above saying anything under 100c should be fine and then reported figures that damage could occur. I cant quite see your logic in stating how he is "side stepping" the whole issue when he gave figures and details !!!

Honestly, people have been asking this alot on Driverheaven and even though ive responded telling people their cards are ok, I figured an offical statement would relax peoples concerns about seeing 60c-80c on their readouts. Ill give you an example one guy in email was even going to return his board when it was reporting 75c - after being used to his AMD CPU running at 45c-50c. I have a feeling he might be somewhat relieved after reading this.

The "myth" uncovered is the fact that people seem to think CPUs and GFX boards should run at the same temperature otherwise their GFX board will be damaged. That is incorrect and now offically stated. Its a little public education on the subject.

Judas

cpu comapred to GFX? it's like apples to organges.... or tomatoes maybe? GFX (GPU/VPU's) are getting WAY more complex then any CPU will really get.... it's to expected that the number of transitors...... the amount of complexity... would relate to the power requirement and heat dispated..... and would explain why with a lower frequency..... that they still cost x more then a cpu....

__________________

The_Neon_Cowboy

IMO: Any one with "decent" airflow shouldn’t have a problem and it’s not like their anywhere as hot as any "6800" flame thrower, super room heater with turbine engine blower. Poor people, must make all the dogs in the neighbor hood bark when they boot their pc that is right before the black out …

Those with out any "decent" airflow should be shot…Or at lest made to take a class or something



EDIT:

You’d get hot too you were trapped in a small confined space with little to no air flow too…Get in a closet put a hair dryier inside, plug it up and turn it on and set there for a short while, it’ll start getting hot. Well thats going on side a lot of cases…

A heat being generated but it has no where to go unless you have air flow to draw it away from the parts!

No offence to any one! I’m not going to name names! I’m mentioning something a few of these cases might be from on here I mean no offense!

Some people on here and some people locally having heating problems and will have a ton of fans and their all blowing out or all blowing in or 1 out and 5 in stuff like that is very common.

People don’t under stand in a STD case you need 55% airflow in 45% Air flow out. (Unless you’ve sealed your case air tight, then it would be 50/50, but no one does that). Some is lost to resistance some escapes on it own. In a “vented” case you’d want 45% airflow in 55% Air flow out to create a minor vacuum and suck the cooler air in trough the vents.

Another thing is room temperature turn on your air conditioners when you need it. If your not comfortable you pc isn’t either!

__________________

MythicaL

Just get a GPU cooler

__________________


Always knew I didn't belong in this world. I wasn't made for this. But I'll never forget those who betrayed me, and those who never failed my trust.

mike2h

i am not suggesting anything neg. about you trying to get info on this matter. sorry if i came across that way. your actions on behalf of people in this forum & in general are way above anything i have seen on any #(a LOT) of sites.

the 'shoulds' are what i do not like. and his temp guidline of sub 100c in itself is non- informative has he really does not give a baseline temp. this could be a problem as most people do not have thermistors on their gpu cores.
i beleive it would probaby be hard to get to 100c myself but you have said you have had reports of 70 -80c.
i would be really curious as to whether these are actual core temps or temps taken from the heatsink.
the main point being is if ati is offering a built in overclocking feature that substantially degrades the life of the card then i see this as a problem.

__________________

mike2h

you are right of course. i was thinking more along the lines of ambient temp increase in the case. sorry i did not make that clear.


NEON bty i have not seen anything like those temps reported for the 6800 series.

__________________

Zardon

from what eric said - 100c or above is the "danger zone" even when overclocked ive NEVER heard a report of these temperatures.

This statement isnt an "indepth analysis" of temperatures on ATI boards its a statement made to reassure people who have concerns seeing 60-80c (which is the general consensus of temperatures ive had from people who are worrying about their boards dying).

60-80c going on what eric has said will cause no issues to this hardware, as he has stated that at 125c damage will occur and over 100c temperatures should be classed as uncomfortable.

of course if I start hearing reports from people getting 100c+ when overclocking via ATIs own tool then ill start asking more questions, but has anyone? thats my main point.

CJintheUK

My Sapphire 9700 pro toasted itself a couple of months ago, I don't know what temp it was running at, but it was never overclocked, so at the moment I'm stuck with a crappy GFarce 5600 256mb....

Zardon

probably just a board failure, did the fan die? did you RMA it?

WaltC

You seem to want to deliberately misunderstand what Zardon is telling you...The "100C" figure is not a "temperature guideline," it is the temp at which the "danger zone" is reached and the chip might fail. AMD and Intel provide the same information for their cpus, and the "danger zone" for their cpus begins at around 85C-95C, depending on the particular cpu you are looking at.

The temps referenced by ATi are of course the temps people can measure with their x800 software (assuming it works like the 9800XT), as ATi knows where the diode or thermastore is, and understands the variables well enough to be considered a bona fide "authority" on the subject of R420 and its temperature readouts. I see no reason to second-guess them, as after all they are the ones who warranty their products, and so have absolutely no reason to prevaricate on this subject. Your observation about the "built in overclocking" degrading the life of the card is exactly backwards from the reality--which is that overdrive is specifically designed into the system in such a way so as not to degrade the life of the card. That's the reason the utility generally is good for just a few additional MHz--generally 10-20MHz in most cases. It is the only case of *factory warrantied* overclocking I'm aware of for a gpu. Also, for ATi to include overclocking which would "degrade the life of the card" would be fairly idiotic, wouldn't it?--Since ATi is the party which will have to replace these cards should they fail.

The message ATi is providing is the very same ones cpu manufacturers provide--as the ambient temperature in the case goes, so go the vpu temps (not to forget cpu temps, and the temp of every other component inside the box.) The most efficient way to cool all onboard components is to lower the ambient case temps. Lots of people run with the side cover off just for that purpose. IMO, 60-80C degrees for a vpu like R420 is fairly ideal. GPUs with 12-16 pixel pipelines, clocked at 475MHz and 520MHz respectively, are going to run generally much hotter than 2 GHz AMD cpus (for instance) with roughly the equivalent transistor count which are manufactured on the same general process size. Most cpu transistors are devoted to cache these days, whereas most vpu transisitors are devoted to active computation; and even more importantly, I think, is that the 12-16 pixel pipelines are operating *in parallel*, unlike in a cpu where the computation is more linear and serial, but clocked very fast. It's kind of like the difference between PATA and SATA IDE controllers: the first is wide and slow, the second narrow and fast; or between DDR SDRAM and Rdram, where the first is wide and clocked slower, the second is narrow but clocked faster.

Anyway, what I get out of this is that the official word from ATi is that if your gpu temps are <100C you are A-OK. And again--my advice to folks is that if they are concerned about component operating temps that they direct their efforts toward lowering the ambient temperature inside their cases--remove the side cover and take your ambient temps after a couple of minutes and you can clearly see the benefit of improving your ambient temp situation.

BWX

I don't know, Take a Thermalright SLK900 w/ ASC with a 4000rpm 80mm fan on it, put it on a CPU of your choice, (AMD, Intel, whatever) and put the same cooler on an X800P or XT- I'm thinking the temps would be similar when both chips are stressed. Don't forget video card cooling solutions are always smaller and less powerful. Ambient temps and airflow are everything though...

__________________

HecHacker1

Maybe DH should ask what are the expected temperatures while running the X800. It's clear that there is no danger to the card under 100C. But what did ATI intend the X800 to run at? Maybe a more specific range, besides just 0 - 100C (or negative for that matter). It is important, because even though my AMD chip, for example, can run up to 80C or whatever "safely" there is no way in the real world it will do that, it simply crashes after about 65C, although no permanate damage is done.

Most likely, those people with cards in the 70-80C range simply have case cooling problems. Ask them to remove their case panels.

Falstaff

I never overclock my 9800 and it runs damn good no matter what I am up too, lol...

__________________

BWX

Yeah, and then put a small slow speed house-fan next to it (just for a test of course)-

Also those people should list their room temp at the time the card's temp was taken..

Those things can have an enormous affect of the temp of all the components inside the PC.

tb77

NVIDIA set the max. threshold to 120°C .... my gf6800 gt (350/1000 running at 400/1100) get's a max of 80°C

Thomas

SeraphicSorcerer

I have a question, and I know this may be covered somewhere in some deep dark corner of the forum but it's never been answered for me, and It really just never has made sense in my mind.


Simple physics, Heat RISES why put the damn gpu UNDERNEATH the PCB?


Why isn't it on top going up towards the cpu/psu fans to dissapate the heat more efficeitnely?

HEAT RISES simple physics, doesnt heat go over the entire board much more and effect the cmponents more being on the bottom then sayu if it ws on the top? sure it'd still be a little warm but the heat would rise up off the fan and heatsink more.... and the cpu fan even tho blowing on the cpu .... it does circulate air therefore allowing air out the case (assuming you have some form of exhaust port even with no fan air will always find holes...

__________________

Rasta Monsta

I agree with one thing Mike is saying. . .ATi should be able to provide "normal" temp ranges. I suspect the reason they are unwilling to is that the cards run at a wide range of temps due to inconsistancies in manufacturing (application of thermal compound, seating of HS, whatever).

The underlying message is, "over a hundred is a bit too hot, under a hundred is anybody's guess."

That is an uncomfortable answer for such expensive hardware. . .at 80 your $400 card will last a couple of years, but at 60 it'll last longer? Imagine buying a car that mildly overheated every day, never breaking down but shortening its useful life. . .

Ouch!

rasta

P.S. Seraphic. . . . . .physics may very well be more complex than you think!

__________________

SeraphicSorcerer

I've taken
Physics/Gravitaional/Quantum and Space Time Quantum Mechanics...
Physics can be annoying/confusing//amazing/awe inspiring/ and just plain weird.
All that is irrelevant, the point being this is physics 101....
I've probly forgotton more then I should being an electrical Engineer, which is why I ask why this question
Does anyone have any idea on an answer?


I mean dang man. Simple physics this is just "heat rises" flip the board over. the heat will rise off at a higher rate also, the quicker the heat rises the "faster" cold/cooler air has to replace that therefore making an airflow... which provides even better cooling my point is, why has no other company ever thought, hey let's uh, flip the board The heat'll uh, rise quicker haha. I mean it's just something I was thinking of while doing these temps, it won't affect circuity, and im sure the outputs wont mind.

I'm just trying to ask a friendly question

__________________

BWX

I have thought the same damn thing for years.. But if you look at most mobo's today, the northbridge is very close to where the video card heatsink would be if you reversed the card.. But still I think they could and should do it, the thing would also be held onto the card by gravity instead of getting pulled off by gravity.. I think many 9700pro's died because those chintzy plastic fasteners melted and the heatsink was able to disconnect with the VPU, mainly because it was a poor design in the first place..

But yeah, someone should do that, I've never heard a reason why they couldn't.. Plus there is usually a case fan right near there- it would suck the heat right off the thing..

__________________

Tipstaff

I can see 2 problems with mounting the GPU on the rear side.

The case fan would pull out air that the GPU fan would need to cool itself down (there would be no to little useable air since both fans would be fighting for the same air/space). It's one of the reasons that you don't put a slot fan that has it's fan facing upwards right near the fan of the GPU.

That aside though, the other problem is that you would be pulling warm air into the GPU fan that just came off the Northbridge chip and from that dang hot CPU.

I'm not sure exactly what the cost difference is, but I think it's more a matter that production wise they can straight run the boards off the assembly lines since everything is on one side of the card (more cards per hour = less money spent). They're looking at it being more cost effective doing it this way versus the 2-4 degrees benifit of mounting the GPU on the reverse side.

BWX

Yeah but it's just wasted space right now, and the space you want (PCI slot) is being taken up. I'm looking inside my rig right now and there is a bunch of space where the VPU cooler could go- and it wouldn't run out of air- there's plenty of air right there- especially with fans sucking the heat off both of them, and fans on the side pushing cold air in.... It just seems like the logical thing to do- especially now that video cards need so much cooling. It's just like putting five people in the front seat of the car and having no one in the back seat- doens't make sense.. I bet anything BTX will have the same dumb design carried over from 1982...

I say flip the damn things over and work out the small problems there might be later on.

__________________

sireric

We do not quote a "normal" range, since we don't have one. Anything below 100C is "normal, safe" operating range. Variations in the physics of the die will lead some to run hot, some to run cooler. As well, every degree variation of the system temp will lead to at least the same variation on the gfx chip. We rate our boards up to 55C on the system side, which will lead to asic measured temperaturs of 100 (maybe as high as 105c). Anything below that is "fine" and will exceed all normal life expectancy. I personally would start wondering if my system temperature wasn't too hot if I saw my reported temperature hit close to 100c. Otherwise, 90c, 80c, 70c, 60c, 50c, as some people have reported, are all fine. The biggest contributing factor is the system temperature. If people have system fans and use a computer around 20~25c ambient temperature, then I doubt system temp would be above ~35c. That will lead to fine asic temps.

mike2h

thx for the response! anyway you have an awnser for the gpu above the pcb instead of below it? never could figure that one out myself.

__________________