Sobach :
It doesn't work that way... you can't directly compare the TDP figure between a CPU and GPU like that. To begin with, the CPU TDP consist of the CPU only, while the GPU's number includes the entire card, including the VRM and VRAM etc., and it's before you even start taking into account real world application issues between the two like die sizes/IHS etc.
This is why my 4790k w/ a CPU-380i block with a MCR320 (120mmx3) radiator hits high 50s/low 60s under load, while my EVGA 980 Ti Hybrid OC'd @ 1400mhz / 7600mhz hits only high 40s under max load.
No. First off, inclusion of the rest of the system is off topic, I only mentioned the "radiator". Yes, the block, pump, materials, etc have an impact but that's not part of the subject at hand.
Thermodynamics is thermodynamics, heat is heat..... the **radiator** performs no differently whether it is connected to a CPU, a GPU or an aquarium header which is what is generally used by test sites to test radiators. If the radiator is receiving 40C water, it doesn't perform any differently if it's coming from a CPU or GPU. Consult any site focusing on custom loop design and this will be readily obvious. Why do ya think they use a heat block to test radiators instead of testing them on a CPU / GPU ?
https://martinsliquidlab.wordpress.com/2012/04/15/alphacool-nexxxos-xt45-360-radiator/4/
Yes, the total wattage of the GFX card includes the GPU, VRM and VRM. The CPU's VRM is on the MoBo. You can add a MoBo block to cool this, custom loopers routinely do. But ignoring this for the moment and let's take a look just why that argument just doesn't have any hope of standing up. A MoBo with VRM, N/S bridges, power control, and all the on board LAN, sound and everything else might pull 30 watts, 40 if it's got PLX. 2 x 8GB of RAM about 6-8 watts.
Now a GFX card doesn't have all that load; it doesn't have the on board sound, LAN, ports to drive and everything else. DDR5 is far more efficient than DDR3 and there's a lot less of it, there's 150 watts from the MoBo going into just 2 GFX cards thru the PCIE slots.... so let's be generous and call it the cards VRAM / VRM on the card as 20 watts. A 980 Ti, over clocked easily breaks 300 watts.... taking off that 20, heck take off 40 and you are still left with TWICE the output of your 4790k.
And while a full cover GPU water block does include the card's GPU, VRAM and VRM, the hybrid CLC does not.... If you have done many builds on 7xx and 9xx series, using air and water cooling, one observation becomes immediately obvious. Water cooling oft provides no performance improvement whatsoever, and when it does it's marginal and likely comes from havng a cooler VRM. The CLC hybrid coolers provide none whatsoever because with the 7xx / 9xx series, we hit the voltage wall before than temperature wall (assuming you have a non-reference card and adequate case cooling).
Yes, you are correct... the VRAM and VRM are not cooled by the hybrid and this is its primary weakness. The performance of the non-reference 9xx will, if anything, be limited by the VRM which hybrid coolers in fact worsen unless you add yet another cooler as the EVGA hybrid does. Not that it is any more effective than the air cooled variety. For many CLC type GPU coolers, they cool the GPU, but the air cooling which previously cooled the VRAM and VRM is now gone.
As shown above, the heat generated by the GPU far exceeds that generated by the VRM and VRAM. On a full cover water block, the sheer thermal mass of the block, provides a large portion of the cooling..... when you design a custom water loop, you look for the radiator to take care of only 60% of the heat load because a significant portion of the heat is not transferred to the coolant but instead transferred to air inside the case from the large surface areas of the backplate / block. The CLC / Hybrid gets no such benefit. So no, taking off the minor loads from the VRM / VRAM does not even come close to offsetting the additional cooling provided by the thermal mass and large surface area of the full cover block and backplate on a full cover water block
For the heat addressed by the radiator, thermal performance is dependent primarily on 5 things:
-The wattage it has to handle
-The surface area of the radiator
-The speed of the fans .
-The material of the radiator ... CLC's have reduced performance due to aluminum
-The flow from the pump .... CLC's have low flow pumps
While the material and pump contribution is significant, the two largest contributors are the size of the radiator and the speed of the fans. The way you make up for an inadequately sized radiator is to provide high speed fans. And that's exactly what every CLC does. The high speed fans needed to offset the small surface area, is eliminates one of the primary advantages of water cooling and that is having a machine that is dead silent. And here, we don't only have to deal with fan noise but the pump noise commonly associated with CLCs and very pronounced on the EVGA 980 Ti hybrid
http://www.overclock.net/t/1558890/youtube-evga-gtx-980-ti-hybrid-review-updated/10#post_24099547
http://forums.evga.com/EVGA-Hybrid-980980TiTITANX-Loudness-level-m2354643.aspx
http://forums.evga.com/980-Hybrid-being-loud-is-there-a-fix-m2347905.aspx
The 980 Ti starts to throttle at 85C... so what is the significance / advantage of you running in the low 60s ? It has no impact on performance. Note also that my old 780s (pulling 294 watts each) are running at 39C with fans spinning far, far less than the hybrid is. The water cooled VRM is about 18C higher. Your VRM is not water cooled and if anything is limiting your OC, that would be it.
Of the 980 Ti's available, the EVGA SC is the one that puts the smallest load on it's cooling system as it's the poorest overclocker. In gaming, the G1 pulls 30 more watts than the EVGA SC
Model / fps / % overclock / Max Core / Max Memory OC
Gigabyte G1 134.8 131.4% 1512 2100
http://www.techpowerup.com/reviews/Gigabyte/GTX_980_Ti_G1_Gaming/33.html
Palit Jetstream 133.1 129.7% 1515 2100
http://www.techpowerup.com/reviews/Palit/GeForce_GTX_980_Ti_Super_JetStream/33.html
Asus Strix 131.7 128.4% 1472 2070
https://www.techpowerup.com/reviews/ASUS/GTX_980_Ti_STRIX_Gaming/33.html
MSI Gaming 130.5 127.2% 1507 2040
http://www.techpowerup.com/reviews/MSI/GTX_980_Ti_Gaming/33.html
Zotac Amp 130.4 127.1% 1465 1990
http://www.techpowerup.com/reviews/Zotac/GeForce_GTX_980_Ti_Amp_Edition/33.html
EVGA SC 126.8 123.6% 1491 1900
http://www.techpowerup.com/reviews/EVGA/GTX_980_Ti_SC_Plus/33.html