Interesting TI images (behardware)

[3DVagabond]

I found it interesting looking at these thermal images of O/C cards. While we are all concentrating on GPU temps, the power stages of these cards are actually what's heating up the most. With the exceptions of the reference design and the Toxic. The XFX seems to be a well balanced design too. Even though the VRM run hotter than the GPU, it's only because the GPU is running quite a bit cooler than reference.

I wish more sites would include this info. Maybe we'd have better balanced designs if it was published more?

http://www.behardware.com/articles/821-19/roundup-9-radeon-hd-6850s.html

I won't post all of the images here. The link will take you to them.

 

[Madcatatlas]

Wow... are you telling me that these card producers dont have the "brains" to test this out for themselves before they ship the card?

What possible errors could we have here that are not highlighted, by that i mean, under what circumstances have the measurements been taken etc. Im sure some of them are listed, but probably not all.


Notice how the
HIS HD 6850 IceQ X Turbo is about 20% warmer than the Sapphire Toxic.

And im sure several people were under the impression that the IceQ cooler was good.

 

[3DVagabond]

I think they pretty much know what the reviewers are going to look at. They know they are going to measure gpu temps. They know they are going to check noise. They know they are going to use Furmark. They design the cards to do well within the parameters of the typical review. In there defense, at the clocks/volts that the cards are released at, the VRM temps are fine. O/C them though and the VRM get hotter faster than the GPU. I'll be reading behardware though to see this information. I thought that the IceQ was about the best deal going in 6850's. After reading this though, I'm very inclined for the Toxic or XFX. There is, to my knowledge, no reference design ever released for sale to the public.

 

[Idontcare]

Wow... are you telling me that these card producers dont have the "brains" to test this out for themselves before they ship the card?

C'mon dude, you seriously going to imply the producers don't have "brains" when the basis of the article is to analyze a single sample of each given card?

I'm pretty sure the producers know exactly what they are making, and have the statistical data to support their QRE versus claims made by a 3rd party based on a sample size of one.

 

[3DVagabond]

C'mon dude, you seriously going to imply the producers don't have "brains" when the basis of the article is to analyze a single sample of each given card?

I'm pretty sure the producers know exactly what they are making, and have the statistical data to support their QRE versus claims made by a 3rd party based on a sample size of one.

Well, every review is a sample of one. What I take from your post is we shouldn't even bother with what it says in reviews. It's too small of a sample size. Surely, I'm misunderstanding something?

 

[Madcatatlas]

C'mon dude, you seriously going to imply the producers don't have "brains" when the basis of the article is to analyze a single sample of each given card?

I'm pretty sure the producers know exactly what they are making, and have the statistical data to support their QRE versus claims made by a 3rd party based on a sample size of one.

Im actually saying the same thing you are saying in your first line here IDC, which is why i mention that the basis of this test must be flawed or must have deviations from some "norm" or standard.

For example, i belive the tests are intentionally showing warmer cards than what is actually the case in...any case , all for the sake of making the article look better/more interesting.

 

[Arkadrel]

Whats clear is...

[FONT=Arial, Helvetica]Sapphire HD 6850 TOXIC
[/FONT][FONT=Arial, Helvetica]XFX HD 6850 (ZDFC)

have far better cooling solutions than anyone else, when it comes to their 6850's.
[/FONT]

 

[Idontcare]

Well, every review is a sample of one. What I take from your post is we shouldn't even bother with what it says in reviews. It's too small of a sample size. Surely, I'm misunderstanding something?

Yes, absolutely. The only thing that is consistent between products is their performance when operating at the same specs.

But everything else from power-consumption to operating temps to lifetime is sample-dependent and varies from chip-to-chip and component-to-component.

There is a parent distribution that has a mean and std-dev from which one can speak to probabilities regarding the attributes of any given member of that distribution, but to be sure no single sample is going to give you the requisite information needed for you to be able to speak to the statistical characteristics of the parent distribution.

Im actually saying the same thing you are saying in your first line here IDC, which is why i mention that the basis of this test must be flawed or must have deviations from some "norm" or standard.

For example, i belive the tests are intentionally showing warmer cards than what is actually the case in...any case , all for the sake of making the article look better/more interesting.

Doh! Yeah I totally misunderstood where you were going with your post. My bad. I wish I could blame it on drinking, but alas that was actually a sober-post, so I'm left with laying blame at a lack of sleep or just a general deterioration of my mental faculty...my ego insists that I go with the lack of sleep angle here :sneaky:


edit: updated my dialogue with Madcatatlas owing to me having had a stick up my ass that required firm, but steady, hands to remove I'm told the splinters will extricate themselves in a matter of days, so, yeah, its all good

 

[BladeVenom]

I was hoping for more pictures of GTX 590s catching on fire.

 

[Piano Man]

I'm glad there is a review site doing these thermal images so we can really see how well a card is getting cooled.

 

[notty22]

A particulars vendors bios, with its corresponding custom fan curve could effect the thermal pictures and make the coolest chip, with the best components and design appear hot.
Also the vendors choice of core voltage.
I'm not saying its not interesting to look at the pictures/read the review. But its not all telling.

 

[3DVagabond]

Yes, absolutely. The only thing that is consistent between products is their performance when operating at the same specs.

But everything else from power-consumption to operating temps to lifetime is sample-dependent and varies from chip-to-chip and component-to-component.

There is a parent distribution that has a mean and std-dev from which one can speak to probabilities regarding the attributes of any given member of that distribution, but to be sure no single sample is going to give you the requisite information needed for you to be able to speak to the statistical characteristics of the parent distribution.

A particulars vendors bios, with its corresponding custom fan curve could effect the thermal pictures and make the coolest chip, with the best components and design appear hot.
Also the vendors choice of core voltage.
I'm not saying its not interesting to look at the pictures/read the review. But its not all telling.

So, do you guys not think there is pertinent information offered by these images to influence a buying decision? I found it very revealing that many (most) of these supposedly better custom designed coolers were not cooling the whole card as well, never mind better than, the reference design.

 

[Idontcare]

So, do you guys not think there is pertinent information offered by these images to influence a buying decision? I found it very revealing that many (most) of these supposedly better custom designed coolers were not cooling the whole card as well, never mind better than, the reference design.

Of course there is pertinent info offered by the images. I think it is a great way to evaluate GPU cooling solutions, reference and third-party.

I just wish they'd grab three cards of each to give us even the roughest of indications as to just how variable the absolute values are. If it is a tight distribution then we can proceed with confidence that single sample sizes are adequate. If we see a large std dev then it would behoove us to bear that in mind going forward as we see more and more of these kinds of TI reviews.

But I have to admit, all those images inspire me to want to do is ghetto strap some 120mm fans hanging off the side of the cards to cool them down.

 

[Dark Shroud]

This just reminded me why I don't buy HIS cards anymore.

And that my next card will be a Sapphire Toxic or ASUS DirectCU.

 

[3DVagabond]

Of course there is pertinent info offered by the images. I think it is a great way to evaluate GPU cooling solutions, reference and third-party.

I just wish they'd grab three cards of each to give us even the roughest of indications as to just how variable the absolute values are. If it is a tight distribution then we can proceed with confidence that single sample sizes are adequate. If we see a large std dev then it would behoove us to bear that in mind going forward as we see more and more of these kinds of TI reviews.

But I have to admit, all those images inspire me to want to do is ghetto strap some 120mm fans hanging off the side of the cards to cool them down.

Would be nice to have larger review samples, for sure. Not gonna happen though. The next best solution would be for more sites to do this type of measurement. You would loose some of the scientific control, as not everyone would measure the same, but avg. out would be useful none the less.

 

[3DVagabond]

Here's a pretty complete selection of TI's for those who consider this while choosing cards. I really do find this revealing of heat issues that aren't typically revealed by standard reviews.

http://www.behardware.com/articles/747-1/report-graphics-cards-and-thermal-characteristics.html

The original date of the article is April 2010, but it's obvious by the inclusion of more recent cards that it's been updated.

 

[Idontcare]

Here's a pretty complete selection of TI's for those who consider this while choosing cards. I really do find this revealing of heat issues that aren't typically revealed by standard reviews.

http://www.behardware.com/articles/747-1/report-graphics-cards-and-thermal-characteristics.html

The original date of the article is April 2010, but it's obvious by the inclusion of more recent cards that it's been updated.

Dang, that IS some rather extensive testing they have done there.

http://www.behardware.com/articles/747-33/report-graphics-cards-and-thermal-characteristics.html

I wish they had a way to generate the TI for the backside of the card that is sandwiched by the shroud of the other one.

I also think it would be neat to see if water-cooled cards fair any better when it comes to this metric of assessing the PCB backside thermals.

Have you come across any TI's for watercooled cards?

 

[3DVagabond]

Dang, that IS some rather extensive testing they have done there.

http://www.behardware.com/articles/747-33/report-graphics-cards-and-thermal-characteristics.html

I wish they had a way to generate the TI for the backside of the card that is sandwiched by the shroud of the other one.

I also think it would be neat to see if water-cooled cards fair any better when it comes to this metric of assessing the PCB backside thermals.

Have you come across any TI's for watercooled cards?

Nope, but if I do I'll be sure to post them.