This is why GTX590s blow up. Infrared thermography: GTX590 vs HD6990

Lorien

Supreme [H]ardness
Joined
Aug 19, 2004
Messages
5,197
So it appears that the 6990 is a cooler card than the GTX590 after all even in AUSUM 450W mode. Pictures speak for themselves:
GTX 590 idle vs load:
gtx590img0031534.jpg
gtx590img0031535.jpg


HD 6990 idle vs load@375W vs load@450W:
hd6990img0031264.jpg
hd6990img0031265.jpg
hd6990img0031266.jpg


Source
 
damn. thats really hot, for either but the gtx is insane you could boil water on it but with a good WC setup it could be a good card.
 
NV was fighting a losing battle, so they dropped the fan noise to advertise a "more silent solution". Looks like it bit them in the ass, so to say.
 
Again from a french site. Why does it seem that all Pro AMD stuff comes from French websites?
 
Those thermal images make since. I can see how 75W of extra power makes a 16C increase in temperature. Thats an immense amount of heat on the GTX590. 112C is pushing the limits of the materials that the card is made of. 96C on the 6990 is alot but it wont be enough to destroy the board. Reducing fan speed to lower the noise threshold is not a good idea on a monster dual card like that. Thats asking for destruction of hardware. I was thinking something else was going on with the 590's that burned than bad drivers and high voltage. Sure you card will get bricked when you push it hard, but literally melting is an engineering defect.

Water cooling should be mandatory on the GTX 590 with thermals like that.
 
Last edited:
I wonder what the settings were when the pics were taken. If that's a stock 590 at load, with stock fan profile, yeah, thats Nvidias fault.

Kinda hard to know what those pics are saying without test methodology.
 
Holy hell, who needs a space heater with something like that keeping your room warm.
 
I hope that is not accurate. If it is, sure explains why they are so underclocked. Not to keep them quiet but to keep them from frying. Sucks.
 
I wonder what the settings were when the pics were taken. If that's a stock 590 at load, with stock fan profile, yeah, thats Nvidias fault.

Kinda hard to know what those pics are saying without test methodology.

That is true, I really want to know more. If those are stock settings, then this may be the hottest card ever made. Its got a full cooler too, not even single slot. I am really interested as an engineer about modes of failure.
 
125 additional watts only raised temp by 6*c? Hrmmmmm doesn't seem right
It does if the fan rpms went up, which they probably did, and more than a little to say the least to cool that card. Both of those cards really push the limits of what is practical in a single slot, but the nV card has clearly gone over the edge there. I wouldn't be surprised if the RMA rates on the 590GTX will be or are through the roof.
 
Since when do you get a card on the shelves that say nvidia Gtx 590.
Always thought those cool coolers on the things are made by the manufacturers like Msi, EVGA, Gigabyte etc etc. You people overlooked 1 thing. The current limiter. Wonder who disabled it in those reviews if it had any. Sure nvidia wouldn't add it to the 590 as well
 
where was quality control in the company? he should be fired for this shit. should of tested it further to see if it can suvive in the real wolrd
 
Aren't you guys missing some points here?
The cards that blew up were overclocked, was overvolted, was using a driver that put power limiter out of action.
 
where was quality control in the company? he should be fired for this shit. should of tested it further to see if it can suvive in the real wolrd

the same idiot who designed the HD6900 series cooler was fired and immediately hired by nVidia. Remeber, all launch HD6950/70 cards had a corner of the PEG connector shaved off, due to clearance issues :D Delayed the damn thing for a whole month :p
 
French to english translation

Always placed in the same box, we found the GPU temperature reported by the internal probe:

GPU temperatures rise relatively high charge, which suggests to us that Nvidia has made ​​noise a priority in the calibration of its cooling system.

What that gives the image using infrared thermography:

GTX 590 at rest
GTX 590 in charge

These photographs show that the whole of the GTX 590 which reaches high temperatures peaking in charge at the feed stage.

Note that these pictures using a scale suited to the comparison of a wide range of graphics cards but smooth results beyond 80 ° C. Here's the same shot but with a different scale to better represent the differences between the high temperatures:

The GeForce GTX 590 expelling half of the hot air inside the case, the emperatures of surrounding components increases, this is particularly true for hard disks that will be severely tested as you can see in our dossier on the thermal evolution of graphics cards has been updated with complete results of the GeForce GTX 590.


This article says nothing about this being the reason the card is frying. It simply says GTX 590 is the reason for half the hot air in the case. It's possible the card very efficient at cooling the gpu's but is unfortunately sending allot of that heat into your case and it's your responsibility to make sure you are ventilating your case sufficiently for it. It may require some serious airflow to keep case temps down. Well more than 6990 at least.
 
Ok take a psu. remove the all the ocp protections. Now start pulling a load from it till it goes poof. Do we run around saying its junk coz it went poof after we did something to over its spec?
 
the same idiot who designed the HD6900 series cooler was fired and immediately hired by nVidia. Remeber, all launch HD6950/70 cards had a corner of the PEG connector shaved off, due to clearance issues :D Delayed the damn thing for a whole month :p

LOL :D

He must have one hell of a resume and interviewing skill.
 
Again from a french site. Why does it seem that all Pro AMD stuff comes from French websites?
Not like the french are using or need their infra red equipment for warfare. Imagine all the infrared cams that they will loose when they run away
 
French to english translation of 6990 page from same site

The Radeon HD 6990 takes the same format as that of the Radeon HD 5970: PCB measuring 29cm and the housing of the cooler exceeds half an inch, bringing the total size to just over 30cm. A map does not therefore fall in all cases. Particularity of its cooling system: a turbine plant that removes half of the hot air outside the case and the other half toward the back of the card, or to the HDD rack.

The Radeon HD 6990 has two different bios, one with a limit of consumption to 375W and the other with a limit of 450W.

Infrared thermography

System at rest.
System load, 375W.
System load, 450W.
Graphics card at rest.
Graphics card support, 375W.
Graphics card support, 450W.

If the system is well cooled off, this is clearly not the case in charge and all components are hurting, especially when the mode is used 450W.

Temperature readings and noise

The Radeon HD 6990 causes extreme temperatures in both the CPU load, as for the chipset or only for the hard disk. Contrary to what happens on many systems, it is preferable to place an extraction fan at the front of the box, so as to prevent the hot air expelled by the Radeon HD 6990 rises in the housing.

Unfortunately, it is clear that this behavior does not promote questionable thermal noise to break records. Note that these vary between 52.8 and 59.4 dB mode and 375W by optimizing the cooling of the housing, it should be possible to stay in the bottom of that range.


Interesting that they mention that they change the fan at the front of their test bench for the 6990 to extract the hot air being blown out of the front of the card towards the drive bays and they don't mention that on the GTX 590... Hmm not gonna jump to conclusions but it'll be interesting to know if they took care of all the hot air coming out of the front of the GTX 590 the same way they did with the 6990. Things that make you go hmm? ;) Edit: I may be misreading that and/or it could be lost in translation. They do mention it is a recommended way to have it configured but I don't believe they say they did it or not.
 
Last edited:
Ok take a psu. remove the all the ocp protections. Now start pulling a load from it till it goes poof. Do we run around saying its junk coz it went poof after we did something to over its spec?

yes, since it will fail to deliver power past a certain point, reguardless of OCP and OPP. If it blows up... it's bad by design and component choice.
 
Those are some sweet thermal images.

I just got to play with one of those cameras at work, and it was thoroughly entertaining.
 
Ok take a psu. remove the all the ocp protections. Now start pulling a load from it till it goes poof. Do we run around saying its junk coz it went poof after we did something to over its spec?

Finally....a voice of reason emerges. I''d be very concerned if the cards failed simply by raising the clocks. Overly exceeding the voltage requirement and blowing the darn thing up, while crying foul is idiotic at best. Max voltages are posted for a reason. I don't know of one other electrical product or semi conductor that will 'live' while over current and over volting...without some form of thermal circuit breaker. Both companies are pushing the envelope where heat and power draw is concerned. While the higher operating temps are okay on new hardware, it will be interesting to see long term results as to what happens when the fans aren't so efficient (due to dust build-up or bearing in-efficiency) as both manufacturer's cards get older.
 
Ok take a psu. remove the all the ocp protections. Now start pulling a load from it till it goes poof. Do we run around saying its junk coz it went poof after we did something to over its spec?

Wait, so let me get this straight...someone drove the GTX590 past specs, it went poof, and NOW they want to cry? What the fuck?

Let's approach this from a non-gaming angle: Hay guys, I took my new M1A, loaded some .308 winchester with hotter loads than the M1A is designed for, and when I fired the gun, it blew up, the stock shattered and the operating rod sheared off - BUT THIS IS SPRINGFIELD ARMORY'S FAULT NOT MINE.

Same thing in the end.
 
yes, since it will fail to deliver power past a certain point, reguardless of OCP and OPP. If it blows up... it's bad by design and component choice.

all psus will go poof bad or good when you load it over its spec. Some just go poof quicker than others. But there's only so much a circuit can take. That's why if you don't know a thing or its new don't go just throw a current at it without knowing its limits. It will go poof
 
I don't think the issue is the gtx590 is going to blow up, but that extra 125watts is going to dump a lot of heat into the case and make a lot more noise through increased fan speeds. Considering the northbridge is normally right next to the pci-e slots that'll increase in temps too. That might cause lower overclock depending on which chipset/cpu you have.

If only the Evga 2Win wasn't so expensive compared to buying two 460s. Obviously not as fast as these cards, but a lot less power consumption, heat, and noise.
 
all psus will go poof bad or good when you load it over its spec. Some just go poof quicker than others. But there's only so much a circuit can take. That's why if you don't know a thing or its new don't go just throw a current at it without knowing its limits. It will go poof

If its a crap psu, a good psu will shut down. I recently plugged in a hard drive that had a short in it. Guess what happened, my psu shut down the rails the hdd was using (I could tell as my other mechanical hdd shut down also) and the computer kept running even though the rail the hdd was on was overloaded by a short.

Sorry but if it was designed good it would shut down before it damaged itself. It's not a new concept even cpu's shut down when they reach there limit (thermally, you can still kill them by overvolting). If you make the protective limits bypassable then whats the point still a flawed design.
 
Last edited:
If its a crap psu, a good psu will shut down. I recently plugged in a hard drive that had a short in it. Guess what happened, my psu shut down the rails the hdd was using (I could tell as my other mechanical hdd shut down also) and the computer kept running even though the rail the hdd was on was overloaded by a short.

Sorry but if it was designed good it would shut down before it damaged itself. It's not a new concept even cpu's shut down when they reach there limit (thermally, you can still kill them by overvolting). If you make the protective limits bypassable then whats the point still a flawed design.

OMG, can you people even read the articles? They were using OLD DRIVERS that seemed to have an error with not regulating the voltages. The people over at OCC said their cards WOULD regulate them when they went too high.

Look take you HD 6990, find its stock volts, use the first beta driver that ever came out and increase the volts by 0.25/0.3(0.95-1.2/.9-1.2). See what your HD 6990 does. :rolleyes:
 
OMG, can you people even read the articles? They were using OLD DRIVERS that seemed to have an error with not regulating the voltages. The people over at OCC said their cards WOULD regulate them when they went too high.

Look take you HD 6990, find its stock volts, use the first beta driver that ever came out and increase the volts by 0.25/0.3(0.95-1.2/.9-1.2). See what your HD 6990 does. :rolleyes:

So the takeaway is nVidia can't write drivers (not that AMD can...)?

Well, that's not news. We've known that for years :D
 
So the takeaway is nVidia can't write drivers (not that AMD can...)?

Well, that's not news. We've known that for years :D

Actually, they released new drivers for the GTX 590 before it was released and these websites still used the old one. Why eat a 2 week old pizza when you can eat a new one?
 
Actually, they released new drivers for the GTX 590 before it was released and these websites still used the old one. Why eat a 2 week old pizza when you can eat a new one?


because it takes 2 weeks to do a review and nvidia recommended the drivers the cards came with.. dur?!?!
 
because it takes 2 weeks to do a review and nvidia recommended the drivers the cards came with.. dur?!?!

This...reviewers get a pre-release set of drivers normally...so that when the product DOES hit the market, we have a review to base our decisions on immediately. Makes sense, but means that they have to tell us to take the review with a grain of salt.

But still...nVidia RTMed with drivers that caused this, so again...nVidia can't write drivers? Or test them?
 
because it takes 2 weeks to do a review and nvidia recommended the drivers the cards came with.. dur?!?!

Thats funny, how come OCC was able to get their review in the first day and they used the new drivers?
 
Still is stupid to have OCP functions available only through drivers instead of at the hardware level, so you're not dependent upon having the "right" drivers. But guess cutting corners is just the norm these days for some companies.
 
Back
Top