The Fermi Thread - Part 3

[ElSel10]

Look at the thermi:

480
IDLE
[*IMG]http://www.geeks3d.com/public/jegx/201004/gtx480_temperature_idle.jpg[/IMG]

LOAD
[*IMG]http://www.geeks3d.com/public/jegx/201004/gtx480_infrared_under_load.jpg[/IMG]


480 SLI
IDLE
[*IMG]http://img707.imageshack.us/img707/1218/gtx480sliidle.jpg[/IMG]

LOAD
[*IMG]http://img97.imageshack.us/img97/8459/gtx480sliload.png[/IMG]



HD5870
IDLE
[*IMG]http://www.geeks3d.com/public/jegx/201004/radeon_hd_5870_infrared_idle.jpg[/IMG]

LOAD
[*IMG]http://www.geeks3d.com/public/jegx/201004/radeon_hd_5870_infrared_load.jpg[/IMG]


HD5870 CF
IDLE
[*IMG]http://www.hardware.fr/medias/photos_news/00/27/IMG0027445.png[/IMG]

LOAD
[*IMG]http://img52.imageshack.us/img52/1130/img0027448.png[/IMG]

http://www.hardware.fr/articles/787-...x-480-470.html

Let's not forget a 4870 under load:


4870 CF under load:


And before you try and say it's irrelevant because it's from last generation, no it isn't. If you were fine with HD4000 cards then the only reason you wouldn't be fine with GF100 would be unreasonable bias (avoiding the f-word).

[Manicdan]

Let's not forget a 4870 under load:

4870 CF under load:

And before you try and say it's irrelevant because it's from last generation, no it isn't. If you were fine with HD4000 cards then the only reason you wouldn't be fine with GF100 would be unreasonable bias (avoiding the f-word).

that might be irrelevant depending on how they loaded it. if that was with furmark, then yes it a bad comparison. but if that was tested under crysis, its acceptable.

[Mk]

i want one of thought heat cameras....... how much do you think they cost?

[Chickenfeed]

Let's not forget a 4870 under load:


And before you try and say it's irrelevant because it's from last generation, no it isn't. If you were fine with HD4000 cards then the only reason you wouldn't be fine with GF100 would be unreasonable bias (avoiding the f-word).

In fairness the 4870 thermals were due to ATIs choice of fan profile. They were quite tamable cards with the reference cooling but they chose higher operating temps over higher noise levels ( and *had* that choice ; ie the fans / heatsink have plenty of headroom - case in point look at the 4890, very loud but I'm more than sure that same thermal picture would be in the 4890s favor ) Nvidia on the other hand didn't have much of a choice ( you can't ignore a high TDP ). As for the cards long term durability, I'm sure they'll be fine under normal usage conditions and the 4870s only help support this fact, which if I am not mistaken was your point (I'd be suspect of tri sli configs on air though ).

[Manicdan]

basically TDP is really the only thing that matters, if its going to have more heat, the card will be hot or loud. or something i hope for is they go with a 3 slot cooler eventually, cause not everyone wants to fill in all 7 PCI slots, and not everyone enjoys super loud gpus, but still likes to OC a little and see safe temps. seriously, add in 10$ more of copper/aluminum, charge 50$ more over reference, and be happy.

[flippin_waffles]

480 load


Geez, to me that really looks like a fire hazard. Apparantley those power cables get so hot they bend like a wet noodle.

[ElSel10]

that might be irrelevant depending on how they loaded it. if that was with furmark, then yes it a bad comparison. but if that was tested under crysis, its acceptable.

The thermal images of the 480 were loaded in the same way as the HD4000 cards. Also several of the reviews have noted GF100 temperatures are in the high 80s when fan speed is set to 70%+.

[trinibwoy]

SemiAccurate ponders GF104. http://www.semiaccurate.com/2010/04/...ode-names-pop/

However, this little tidbit was most entertaining

GF100 is horribly weak in pixel ops. It has only 64 ROPs (Render Output Units), and that lack of pixel power shows up at higher resolutions.

[570091D]

^HAHAHA^

well charlie hasn't changed his ways (not that i was expecting him to), he states the obvious about gf100's ROPs, then claims that there will be no more cards after this batch... please. and finally, the gf104 will be a huge flop because gf100 has an ROP shortage at high resolutions....

i have to say, my 9800GX2's would hit 107-110 C under heavy load with the factory TIM. replacing that brought my load temps down 10 C, and none of these cards are close to those temps... everyone is calling this a potential RMA liability but i just don't see it.

[trinibwoy]

How is it possible to blog about GPUs for a living and not know the difference between a ROP and a TMU? And even if he were to get that right, texture cache efficiency increases with resolution, not the other way around. I wonder how he feels about Cypress' 32 ROPs

/boggle

[570091D]

How is it possible to blog about GPUs for a living and not know the difference between a ROP and a TMU? And even if he were to get that right, texture cache efficiency increases with resolution, not the other way around. I wonder how he feels about Cypress' 32 ROPs

/boggle

they are the world's most efficient ROP's, the sun shines directly out of them, and they will cure cancer! -CD

[Dami3n]

Look at the thermi:

480
IDLE


LOAD



480 SLI
IDLE


LOAD




HD5870
IDLE


LOAD



HD5870 CF
IDLE


LOAD


http://www.hardware.fr/articles/787-...x-480-470.html

Sorry, but use a thermal camera with a card with full cover backplate -> FAIL

[Chickenfeed]

Sorry, but use a thermal camera with a card with full cover backplate -> FAIL

I seriously doubt the inclusion of a backplate would change things that much. They are still metal and will heatup just like the pcb. Yes there is more surface area for the heat to spread to but to go as far as believing the backplate is a game changer... Assuming that those load pictures were taken after an extended heavy load ( an hour + ) that gives the back plate plenty of time to absorb heat and thus reflect it in the pictures.

Anyways just look at the side of the 480s heatsinks ( which are still plastic casings similar to the 5870s ) It is clear that the shroud gets very very hot ( ie do not touch hot ) They do have vents yes but the 5870 sink also has some vents in the same area ( just fewer of them and they are smaller )

Again none of this should be a problem in larger cases with decent airflow and assuming you dont let anything touch the cards for extended peroids, all should be well ( ie lose / messy cables )

[highoctane]

You can tell the shrouding is masking/hiding the heat by looking at the color where there are openings. Just like exhaust manifolds on cars use a small piece of tin/steel barely spaced off the exhaust manifold as heat shielding. If the shrouding was off and a picture of the pcb was taken it would would look very different than the one we are seeing.

[Chickenfeed]

Regardless of how it is looked at the surrounding case temps tell the story. 480s bleed crap loads of heat into a case. Anyone considering SLI better have insane airflow or plan to go the watercooling route otherwise their overall system temps are surely going to suffer substantially. These appear to be at least as bad as the 4870x2s ( mine made a good 10-15C difference on my northbridge on my x48 board and a good 5C increase on other components at full load )

[highoctane]

As long as stability is in check/maintained I'm not sure I would over emphasize added heat dump. Sure its going to be hotter but whether or not that creates other issues remains to be seen.

[Chickenfeed]

As long as stability is in check/maintained I'm not sure I would over emphasize added heat dump. Sure its going to be hotter but whether or not that creates other issues remains to be seen.

Of course but it is something that must be accounted for none the less. As far as single gpu boards go this is the hottest one ever so people will have to rethink some things ( cooling demands after power consumption ) I'd hope most people who put these cards in their machines are well aware of this but there will be the vergence in the force with the odd few running their systems at the brink as is and it is an easy catalyst to send things overboard but thats nothing new ( ie average joe puts a 8800gtx in his hp oem pc which has a 250 watt sparkel psu... BooM -grilled pc sandwich :0 )

[ElSel10]

Regardless of how it is looked at the surrounding case temps tell the story. 480s bleed crap loads of heat into a case. Anyone considering SLI better have insane airflow or plan to go the watercooling route otherwise their overall system temps are surely going to suffer substantially. These appear to be at least as bad as the 4870x2s ( mine made a good 10-15C difference on my northbridge on my x48 board and a good 5C increase on other components at full load )

If what we are ultimately looking at how much heat is given off, regardless of the reasons, then why does nobody have a problem with the HD4000 cards?

And if we just want to see how how it actually gets for longevity sake, then yes it does make a difference on the HD5000 cards as to whether or not the backplate is on.

[Chickenfeed]

If what we are ultimately looking at how much heat is given off, regardless of the reasons, then why does nobody have a problem with the HD4000 cards?

And if we just want to see how how it actually gets for longevity sake, then yes it does make a difference on the HD5000 cards as to whether or not the backplate is on.

I wasn't arguing potential longevity, just that the backplate won't do much for general pcb temps. It is there strictly for the memory as it is double sided unlike the 480/470 ( which has all its memory on the gpu side )

[ElSel10]

I wasn't arguing potential longevity, just that the backplate won't do much for general pcb temps. It is there strictly for the memory as it is double sided unlike the 480/470 ( which has all its memory on the gpu side )

That isn't true though, as seen in post #2939

[Chickenfeed]

You totally missed my point... the only reason the backplate is there is due to memory, it isnt there to benifit the pcb directly (otherwise we'd probaley see one on the GTX4xx cards...) it might have a very minor side effect but its there for the benifit of the ram. I don't recall cards using backplates that didn't have memory on the rear side of the pcb.

[STEvil]

I wonder if any of these pictures are EMI calibrated?

[flippin_waffles]

Well this is interesting:

Moreover, I noticed the image to get somewhat blurry on the GeForce when running the test demo and checked this out with screenshots (on another level as I could not make identical screenshots on the Chaser map). You can see the difference yourself (the screenshots were captured at 1920x1080).

It is easy to see that the Radeon produces a better-looking picture. All the textures are sharp, without fuzziness or anything. To remind you, I selected the same graphics quality settings for both graphics cards: DirectX 11, Very High, 16x AF, AAA, Advanced DOF and Tessellation. Besides, I selected the High Quality texture filtering mode in the Catalyst and GeForce/ION driver (the Quality mode is selected by default).

From the xbit 480 review in the Metro test.

http://www.xbitlabs.com/articles/vid...tx-480_11.html

[YukonTrooper]

Well this is interesting:





From the xbit 480 review in the Metro test.

http://www.xbitlabs.com/articles/vid...tx-480_11.html

What's even more interesting is the part that you seemed to look over:

Perhaps this reduction of quality is the trick the GeForce GTX 480 resorts to in order to deliver a higher speed in Metro 2033? The game developer answered to our question promptly. Here is what Oles Shishkovtsov, 4A Games Chief Technical Officer said:

No, the observed difference in quality is not due to the performance of the graphics cards. Indeed, graphics cards from Nvidia and ATI render the scene differently in some antialiasing modes. We are trying to find a way to correct this issue. Again, this has nothing to do with speed.

Hopefully, people from 4A Games will find a solution.

Sorry if you wanted to leave that part out, Flip.

[ElSel10]

SemiAccurate ponders GF104. http://www.semiaccurate.com/2010/04/...ode-names-pop/

However, this little tidbit was most entertaining

Fail. If the 480 clearly outperforms the 5870 despite is's 60 TMUs and 48 ROPs, then what will be the issue with GF104? Further the specs of the GTS430, 440 and 450 (all using GF104) have supposedly leaked here:
http://www.dvhardware.net/article41879.html

All but the 430 have 64 TMUs (more than the GTX480) and 32 ROPs (same as the GTX285). Major fail on Charlie's part.