D-14 On GPU?

[Face2Face]

What you posted was speculation. You didn't post any results of actual measured heat output.

Not all the power used is turned into heat. 

How is he plain wrong? Results from just about everyone including the Kraken G10's results on the website show a gtx 780 at 40c at load. You won't see the same result for a cpu. 

I think you are too focused on temperatures rather than heat generated.

 

TDP which is the maximum amount of heat generated by the CPU or GPU, which the cooling system in a computer is required to dissapate. Per Wikipedia http://en.wikipedia.org/wiki/CPU_power_dissipation

 

TDP of 3570K  - 77w  ----  TDP of GTX 780 - 250w

 

The surface area argument alone is a good explanation on why you are seeing cooler temps on the GPU vs. CPU, even if the wattage consumed is nearly 3X the amount.

 

Keep in mind we are comparing a 22nm  chip  with 1.4 billion transistors  to a 28nm with 7.1 billion transistor chip. I don't really feel you are getting it, and maybe I am not doing a good job of trying to explain it?  Does some research on your own and post your finding, I would be interested to see what you find.

[WoodenMarker]

I think you are too focused on temperatures rather than heat generated.

 

TDP which is the maximum amount of heat generated by the CPU or GPU, which the cooling system in a computer is required to dissapate. Per Wikipedia http://en.wikipedia.org/wiki/CPU_power_dissipation

TDP of 3570K  - 77w  ----  TDP of GTX 780 - 250w

The surface area argument alone is a good explanation on why you are seeing cooler temps on the GPU vs. CPU, even if the wattage consumed is nearly 3X the amount.

 

Keep in mind we are comparing a 22nm  chip  with 1.4 million transistors  to a 28nm with 7.1 billion transistor chip. I don't really feel you are getting it, and maybe I am not doing a good job of trying to explain it?  Does some research on your own and post your finding, I would be interested to see what you find.

I know what TDP is. Even though the TDP spec of the gtx 780 is higher, it doesn't mean the amount of heat generated is the same.

Heat and temperature scales the same when using the same cooler provided the surface area of contact is the same. Although it isn't the same in this case, I don't think there is a big enough surface area difference that would suggest that the gpu outputs more heat. 

 

We're both speculating here with lack an actual benchmark measurement of heat for both scenarios. the TDP spec is the theoretical max amount of heat that could possibly be generated. Actual results aren't nearly as high. 

[applebook]

Face2Face is onto something.

 

I still doubt the "benchmark" claims posted by NZXT, but if it's accurate, the same cooler is making significantly more contact with the much bigger 780 die size compared to your average i5/i7 processor. 

 

Look at recent AMD chips. They are significantly bigger and can be kept at much lower temperatures under load even when consuming (allegedly) similar wattage as an Intel chip. AMD chips make more contact with CPU blocks and are drawing less power per mm^2.

[Face2Face]

Face2Face is onto something.

 

I still doubt the "benchmark" claims posted by NZXT, but if it's accurate, the same cooler is making significantly more contact with the much bigger 780 die size compared to your average i5/i7 processor. 

 

Look at recent AMD chips. They are significantly bigger and can be kept at much lower temperatures under load even when consuming (allegedly) similar wattage as an Intel chip. AMD chips make more contact with CPU blocks and are drawing less power per mm^2.

 

The temps claims from NZXT are true. I have seen this setup on a Lighting like my own and pushing 1.35v @ 1400Mhz it tops out @ 53c. Many other users on OCN are showing a big decrease in temps as well. Even Linus in his own testing showed a decrease in temps... oh wait... no he didn't -- He didn't even mention the temps until someone pinged him on youtube. I watch LTT videos and like most, but his review on the NZXT G10 was extremely rushed and embarrassing on his part.

 

Like you said the cooler is making more & better contact with the GPU die. WoodenMarker claims the difference is not that big, I guess 400% is not that big? 

[applebook]

The temps claims from NZXT are true. I have seen this setup on a Lighting like my own and pushing 1.35v @ 1400Mhz it tops out @ 53c. Many other users on OCN are showing a big decrease in temps as well. Even Linus in his own testing showed a decrease in temps... oh wait... no he didn't -- He didn't even mention the temps until someone pinged him on youtube. I watch LTT videos and like most, but his review on the NZXT G10 was extremely rushed and embarrassing on his part.

 

Like you said the cooler is making more & better contact with the GPU die. WoodenMarker claims the difference is not that big, I guess 400% is not that big? 

 

Now I'm seriously considering a G10 with a really cheap and quiet 120mm just for the silence and decent overclocking. NCIX regularly has entry Antec and Zalman AIO units for very cheap (with rebate). $30 for the G10 + $40 for an AIO are worth it if I can reuse this set-up for a few more GPU upgrades down the line. Maybe grab a Titan in 2 years when they go for pittance?  

 

I think at the end of the day, both sides are correct. GPUs can generate more heat, but because CPUs cannot be as efficiently cooled due to their design, they are much more difficult to cool even with a custom loop. 

[BallaTheFeared]

What you posted was speculation. You didn't post any results of actual measured heat output.

Not all the power used is turned into heat. 

How is he plain wrong? Results from just about everyone including the Kraken G10's results on the website show a gtx 780 at 40c at load. You won't see the same result for a cpu.

There is no speculation here.

Power consumption = heat

Watts = Heat

All power used turns into heat. Electronics do not consume power, it is simply passed through it. What goes in comes out in the form of heat, how much is based on how much it used which is measured in watts.

[WoodenMarker]

Face2Face is onto something.

I still doubt the "benchmark" claims posted by NZXT, but if it's accurate, the same cooler is making significantly more contact with the much bigger 780 die size compared to your average i5/i7 processor. 

Look at recent AMD chips. They are significantly bigger and can be kept at much lower temperatures under load even when consuming (allegedly) similar wattage as an Intel chip. AMD chips make more contact with CPU blocks and are drawing less power per mm^2.

Like you said the cooler is making more & better contact with the GPU die. WoodenMarker claims the difference is not that big, I guess 400% is not that big? 

I'm not sure because we don't have any benchmark from the heat--only theoretical specs.

I think at the end of the day, both sides are correct. GPUs can generate more heat, but because CPUs cannot be as efficiently cooled due to their design, they are much more difficult to cool even with a custom loop. 

...That makes quite a bit more sense than what I've said.

There is no speculation here.

Power consumption = heat

Watts = Heat

All power used turns into heat. Electronics do not consume power, it is simply passed through it. What goes in comes out in the form of heat, how much is based on how much it used which is measured in watts.

I don't disagree with your idea but I just think it still counts as speculation if we're making 'calculations' with no actual test numbers but instead relying on how things 'should' work.

[Sandzsteedt]

Only good way you can accurately compare the AIO liquid cooler solution on GPU vs CPU is if you took the IHS off the CPU and put the cooler on the bare die like you do with most of the GPU's.

There has been tests done with directly mounting the CPU cooler on the die. If I remember right; EKWB did even make an adaptor kit for their CPU waterblocks for direct die mounting at some point.

Checked the webs, and here's a link:

http://www.tomshardware.com/news/Water-Cooling-IHS-Intel-Ivy,21744.html

 

The surface area difference alone makes a difference, but I'm willing to bet that the bigger difference comes from the double layer of TIM and one extra metal layer called IHS in between the die and the block.

TIM is never perfect for heat transfer, the new liquid metals and wonder materials do help a bit over the regular ones, but generally TIM is used because going without would be worse. If TIM would be the greatest thing ever in moving heat we would have loops of it running to cool stuff.

 

To reverse the experiment you could take an old IHS from a CPU and cut off the edges to make a flat plate of it, then put TIM on the GPU die and place the IHS on it. Then put TIM on the IHS and mount the cooler and now you would almost have a similar setup as the CPU has. Keeping in mind that the die size is still different.

 

 

To sum up. The AIO coolers generally seem to perform better on GPU's but most of that is because the setup of the cooler in relation to the die itself is completely different.

 

EDIT. Searching a few minutes for Sandy/Ivy Bridge direct die cooling I found that people are reporting lower temps by 9C-20C range in comparison to having the IHS and double layer of TIM.

Here's one source: http://www.overclock.net/t/1269943/flat-waterblock-for-direct-to-die-cooling-on-ivy-bridge