GTX480 hits 87c and higher on load?

[sciwizam]

Charlie: "In the end, the card is too hot, too slow, and unmanufacturable. We told you so. Pop goes the ego."
--

I think his ego is the one in need of popping... This guy is trying FAR too hard to convince us that he's been right all along, and glossing over little issues like the fact that there will be more manufactured parts at launch than there were 58X0s.

Using Theo's words to discredit Charlie's :awe:

 

[Allio]

Unless it sounds like a leafblower at that temperature, who cares?

The main point of the article, though, is delicious. Almost every single benchmark to come out has been dodgy or obviously cherry-picked in some way. Very specific parts of Unigine, Far Cry 2 over and over and over, and now D3D9 Dirt vs D3D11 Dirt. Hardly fills you with confidence with the card's performance in the wild.

 

[Daedalus685]

Well, to be fair to Charlie he really only mentions the temperature at the end.. While he does point tit out in a powerfully lame way, is the point of that write up not to bring attention to the dx9 vs dx11 he claims nvidia was doing?

He is being his usual sensational self.. but if there is truth to nvidia releasing numbers comparing dirt2 on nvidia to ati but using dx11 for ATI and dx9 for their own stuff, well that is just a bit dirty. Mind you it seems totally unsubstantiated as to whether that screen he has shown was actually the run used to compute any numbers..

 

[ScorcherDarkly]

I think it's somewhat significant considering the stock cooler has 4 heat pipes cooling the chip, but not a huge deal. It would also be significant to overclocking headroom. I would think 100 C would be as hot as you would reasonably want to run a chip, and there's not a ton of room there. We'll know better once we get the side by side with ATi, I suppose. So glad that its almost here and the rumor mongering can stop at last.

 

[Daedalus685]

Well sorry then guys that I asked the question. I guess since I have never owned anything like a 4870X2 that it makes me ignorant.

No need to apologize good sir. It is a good value to know. Besides, I was pretty shocked the first time I got my 4870 and 4890s that high given my previous temperature experience was with relatively cool CPUs.

Besides, this is an article about fermi, which is defacto ruled by rage and feelings before logic and civility.

 

[Dark4ng3l]

No need to apologize good sir. It is a good value to know. Besides, I was pretty shocked the first time I got my 4870 and 4890s that high given my previous temperature experience was with relatively cool CPUs.

Besides, this is an article about fermi, which is defacto ruled by rage and feelings before logic and civility.

Oh yes I need to apologize as I just posted a Charlie article that is about nvidia. Some people around here have recently accused me of being an ATI marketing drone and even Charlie himself but we all know who the real stealth marketer around here are. So I must apologize for not paying more attention to card temps in reviews and having never owned one of those hot as hell high end boards.

And about the DX9 VS DX11 Dirt2 thing my understanding is that for some reason the demo of the game runs in DX9 mode on Fermi even if you set it to DX11 (some kind of bug) and the interns that they probably had do the hundreds of benchmarks to find what they could use the best for marketing failed to notice the discrepancy. I seriously doubt that nvidia would really treat us like idiots by purposefully comparing DX9 to DX11 numbers they would have to really have gone off the deep end.

 

[tviceman]

but we all know who the real stealth marketer around here are.

Who is participating in "stealth" marketing on this board?

 

[Rubycon]

Why worry about this? Semiconductors can run much hotter than this. The engineers know what they are doing. If your card gets too hot from sucking in horse hair and hits 125C it will throttle back. Don't worry about it. (but clean your room!)

 

[zerocool84]

Why worry about this? Semiconductors can run much hotter than this. The engineers know what they are doing. If your card gets too hot from sucking in horse hair and hits 125C it will throttle back. Don't worry about it. (but clean your room!)

but this is Anandtech and since only women clean and we don't have any, all of our rooms stay dirty

 

[zagood]

yea there's really nothing to see here. Someone's trying to pick a fight.

what?!?!?!!

 

[MagickMan]

FYI, only reason these higher temps are OK is because the voltage on GPUs is much lower than the stock voltage for CPUs.

Huh?

 

[Apocalypse23]

2 more days people! The world will know how these suckers perform!

 

[shangshang]

I am also confused as to what power consumption has to do with GPU temperature.
I could make an HD5450 run at 100c if I wanted, and cool a GTX295 down to 60c, tell me which is using more power. Heat has nothing much to do with power, and everything to do with the cooling system used.

hmm heat and temperature are related... but different things. Temperature is a measure of energy within 1 system (ie. internal energy). Heat is a measure of kinetic energy transferred between 2 systems.

If everything stays the same, then a system with higher power consumption will have a higher temperature.

But taking your example above, the HD5450 runs at 100C and GTX295 at 60C... you ask which system is consuming more power.. and we may (wrongly) conclude that the HD5450 consumes more power because it has a higher temperature. However, it is wrong to conclude like so because the 2 systems are different, so we cannot do a direct comparison without involving entropy into the equation.

The equation for temperature (T) for a system at equilibrium is :

T = (change in energy) / (change in entropy)

In order for the GTX295 system to run at such a relatively low temperature, there has to be a big big cooler be coupled to the system (GTX295), and it is this big big cooler that will cause an increase in entropy to the environment (the surrounding air). The GTX295 will cause a greater increase in entropy to the environment than the HD5450. If we're talking about just 1 system, and if every thing stays constant, then an increase in power consumption will cause an increase in internal energy which also necessarily means an increase in temperature (according to the equation above). However, if we ask questions about 2 different systems (GTX295 and HD5450), then we must take entropy into account in order for the questions to make sense.

There, I hope I've cleared up some of your confusion. I've only had 1 year of physics and general chemistry in college a while back. So my explanation could be missing a lot of details.. any person with a firm grasp of the 2nd law of thermodynamics please feel free to correct me.

 

[mmnno]

Seems low for 250W. Must be a beastly cooler, I wonder how much the metal sink weighs.

 

[Executor_]

I seriously doubt that nvidia would really treat us like idiots by purposefully comparing DX9 to DX11 numbers they would have to really have gone off the deep end.

These are the same people who constantly rename their old cards in order to fool the masses into believing they're releasing new cards. Nothing is too low for them.

 

[Lonyo]

hmm heat and temperature are related... but different things. Temperature is a measure of energy within 1 system (ie. internal energy). Heat is a measure of kinetic energy transferred between 2 systems.

If everything stays the same, then a system with higher power consumption will have a higher temperature.

But taking your example above, the HD5450 runs at 100C and GTX295 at 60C... you ask which system is consuming more power.. and we may (wrongly) conclude that the HD5450 consumes more power because it has a higher temperature. However, it is wrong to conclude like so because the 2 systems are different, so we cannot do a direct comparison without involving entropy into the equation.

The equation for temperature (T) for a system at equilibrium is :

T = (change in energy) / (change in entropy)

In order for the GTX295 system to run at such a relatively low temperature, there has to be a big big cooler be coupled to the system (GTX295), and it is this big big cooler that will cause an increase in entropy to the environment (the surrounding air). The GTX295 will cause a greater increase in entropy to the environment than the HD5450. If we're talking about just 1 system, and if every thing stays constant, then an increase in power consumption will cause an increase in internal energy which also necessarily means an increase in temperature (according to the equation above). However, if we ask questions about 2 different systems (GTX295 and HD5450), then we must take entropy into account in order for the questions to make sense.

There, I hope I've cleared up some of your confusion. I've only had 1 year of physics and general chemistry in college a while back. So my explanation could be missing a lot of details.. any person with a firm grasp of the 2nd law of thermodynamics please feel free to correct me.

I had no confusion, I was just pointing out that fact. Everything else doesn't stay the same, so saying "oh no the GTX480 runs at 87c" is meaningless, because there is more at work than power consumption. The temperature of the GTX480 gives no indication of power consumption, only of the effectiveness or otherwise of the supplied cooling solution.

Add in the lack of knowledge of the test setup which resulted in those temperatures and you end up with something which is entirely meaningless.

The HD5970 runs cooler than the HD5870, just to highlight the disconnect between power and heat when comparing two entirely different graphics cards where almost nothing is the same. If I really wanted, I could get an HD5870 to run at 80c under idle if I set the conditions up correctly. That doesn't mean it's suddenly using more power to end up running hotter.

 

[Zstream]

Why are people not worried about this?

If a card reaches 90C while having the GPU cooler set to 10%, then yes, nothing to worry about. If a card reaches 90C while having the GPU cooler at 80%, then YES, we have something to worry about.

 

[Qbah]

My HD4870 was running 90C+ under gaming with stock cooler and stock settings. It's just the way it was designed. As long as it's within nVidia's acceptable parameters, it's perfectly fine.

 

[vshin]

87 degrees is actually not bad. My 4870 has been running 100-101 at full load using stock settings for over a year with no problems. This is more a function of the default driver settings since it keeps the fanspeed low and blips it high when the temp crosses 100. I can tweak the settings manually to keep the temps around 85-90 but clearly ATI has designed these cards to run boiling hot.

 

[tviceman]

2 more days people! The world will know how these suckers perform!

I don't think we're getting official reviews from all the hardware sites until Monday though.

 

[jvroig]

87°C isn't really a big deal.

According to SA, 87C figure is in a gaming/demo scenario, while Furmark is 98C. I read the Xbit article you pulled that image from, and they said they used some software including Furmark. They did not specify if the temps are from Furmark. But if it is, then that image (as contributed to this thread) is a little misleading, since it should be compared to 98C (the Furmark temp).

In the end, however, whether high 80's during gaming bother you or not is not very important, and much less so is the 98C in Furmark. As long as it is within nVidia's spec, then all is good. I assume that if you are running such a high-end card anyway, your system cooling should be able to handle it anyway.

If this is a demo of the actual units that will eventually be launched very soon, then this is the 480sp unit (since the 512sp was canned to be launched later when the quirks are fixed, right?). In that regard, Charlie was right at first, since he was talking about the 512sp parts and initial clockspeed target (700 or 725?), since the 480sp already reaches almost 100 at furmark, then the 512sp part would theoretically breach that.

Of course, it is disappointing that Charlie published this recent article about the demo and 87/98C being too hot. If it's already under spec, then it is ok. He should have been content saying "I told you so" about the 512sp part.

 

[1h4x4s3x]

We will!
If not, I'll go insane.

#above me
Yes, it's a bit unclear. But I'm pretty sure Furmark raises the temperature quite a bit more than those mentioned 75°C.

 

[Daedalus685]

Of course, it is disappointing that Charlie published this recent article about the demo and 87/98C being too hot. If it's already under spec, then it is ok. He should have been content saying "I told you so" about the 512sp part.

He surely mentions it in the article but the tone of the article really seems to be that of "I was right" because "The dirt 2 numbers floating around are rubbish." He was not using the temperature to prove his point, merely to add to it, yes, in fact it is a warm chip. Of course we know the temperature means very little if it still works as intended.

 

[McWatt]

Of course, it is disappointing that Charlie published this recent article about the demo and 87/98C being too hot. If it's already under spec, then it is ok. He should have been content saying "I told you so" about the 512sp part.

Every few days Charlie writes something new that directly contradicts his last story. In the end you can cherry pick detail #1 from February 22, detail #2 from October 11, and detail #3 from January 1 and claim that Charlie predicted everything correctly, or you can note that 90% of his details were incorrect.

This is how "psychics" scam their customers, too. People get excited about what they get right and forgive them for what they got wrong. "There was an important woman in your life whose name started with a P, L, E, or maybe M" becomes "I can sense that your grandmother was named Mary" in the retelling. These scam artists shouldn't be given a free pass. Hold them accountable for every claim they make.

I'm sure Charlie does have some sources in the industry, but because he's willing to publish anything, regardless of likelihood, and give it the editorial tone of complete fact, his published claims end up being no closer to reality than random guesses.

Little stories like this one, where he claims that a temperature equal to my stock 4870's is "too hot," make him even more despicable. I'm getting yet another ATI card (5870) for myself. I'm not an nVidia fanboy. I know, however, that Charlie's "reporting" is pure garbage and that Charlie himself is the same.

 

[jvroig]

Yes, it's a bit unclear. But I'm pretty sure Furmark raises the temperature quit a bit more than those mentioned 75°C.

Yes, most likely. It's just frustrating to not know for sure. I even went back to the AT review. I read the section on it over and over, and I couldn't determine what they did to load the GPU to get the loaded temps. It would be great if they mention the specific method/software they used to load the GPU, that way reviews could be better paired against another review if you see they used similar/identical temperature testing methodology.