No need to bet what might be happening when you can just read the temps with thermometer from ya tool box or utility on your computer.
Let's try approaching the question from this perspective. Again, how is it possible for a water cooling system manage to keep a GPU cooler than say an i7 CPU ? The die sizes are not that different . On our test system we use 6 temp sensors, infrared thermometers and a fog machine to do thermal and air flow testing. Here's what we have observed. Measurements made with infrared thermometer and HWiNFO 64
a) The GFX cards produces much more heat (295 watts each) / the CPU about 130.... advantage => CPU by 2 to 1
b) The GFX cards receive half the water flow of the CPU as the 2 cards are piped in parallel.... advantage CPU by 4 to 1.
So how do the GFX cards overcome that 8 to 1 disadvantage ? How is it that 1.25 gpm acting on 130 watts can't keep up with 0.62 gpm acting on 590 watts
Despite those disadvantages, the GPU manages to remain at 39C, the VRMs are at 52C and yet the CPU is way up in the low - mid 70s ? Any thermal heat transfer must eventually get "to air" and the large metal backplate provides a greater surface area for the metal to air heat exchange.
If, as you say, efficiently heat transfers from the chip to the PCB, then what ? Since the chip is exposed to air with no backplate any additional cooling is limited to the minute surface area of those VRMs where they are exposed on the back of the board.... the backplate quickly expands that VRM cooling surface area to 45 sq in.
The same thing applies to a CPU air cooler... what transfer co-efficient is higher ?
- Internal CPU to heat to IHS
- IHS to copper heat plate on cooler
- Copper plate to heat pipes
- Heat pipes to aluminum fins
- Alum fins to air
With the 1st 4, it doesn't matter which is higher and which is lower because as a chain is only as strong as it's weakest link, it doesn't matter what is the 2nd or 3rd weakest ... performance is limited by the final least efficient heat transfer, that being from aluminum fins to air. Decrease the surface area of the fins and performance will decrease regardless of those other coefficients. Same thing with the VRMs
Also... if the thermal transfer coefficient of the PCB material was better than metal, gotta ask the obvious question. Why don't we see air coolers and water blocks being made of PCB materials ? Better yet, why aren't PCB mounted heat sinks made of PCB material instead of highly conductive metals.
Now if you were to ask "Does the overall temperature of the PCB (say averaging 16 points around the board) increase or decrease with a backplate ? Can't say; I have never bothered to measure it since the results have 0 significance, but common sense says slightly warmer. However, I don't really care about the overall average temperature of the PCB as it will have no impact on card performance... the articles on the failing cards certainly don't list this as an issue.
Looking at this pic, why did they add a heat sink / thermal pad to the VRMs and not the entire PCB ? Cause the VRMs need to be cooled, the PCB is of no concern
So why not put a heat sink on instead of a backplate ? Certainly a heatsink w/ fins would be more effective, especially with a fan.
a) On many MoBos, it would eliminate SLI as an option as the taller fins would hit the top card
b) Another purpose of backplates is to keep dust off the PCB surface, dust and humidity can be a bad thing for contacts that are close to one another.
c) Water cooled systems benefit from backpates, as small leaks found during installation, testing and even later usage oft get "saved' by the backpate
But getting back to the articles, I don't see as any of them covering this issue point to the temperature of the PCB as being **the** problem,. No, the only problem is the temperature of the VRMs. So, if we focus on **the problem**, it's both obvious and inarguable that a heat sink does a better job cooling VRMs (or any chip) than air does. Use a heat sink when it's needed ... and don't use a heat sink when it's not needed ... unless ofc, you just want cooler temps. And the bigger the heat sink, all other things being equal, the cooler the chips will be.
Certainly, I agree, EVGA created this problem by using a backplate that was not thermally coupled to the VRM. That is certainly worse the no backplate at all which is why the non-ACX models w/ no backplate are not on the list.
Yes you could improve things @ the VRM by blowing air across the PCB but you can improve **VRM temps** more by blowing air across the heat sink which has that 45 sq in of surface area. The VRMs are exposed on both sides of the board. So if you have VRM temp problems and you've done what ya can on the top of the PCB, best / easiest way to attack the problem is from the back side of the it. The relative question therefore is ...
What will cool the VRMs better ?
a) exposing the VRMs (whose exposed surface area on the back is probably 1 - 2 sq in) to air
b) affixing a heat sink to the VRMs (w/ 45 sq in of surface area) and exposing all that to air
If it's not b) then MoBo manufacturers and other electronics vendors have it all wrong.
I haven't done detailed thermal testing on Nvidia cards since the 7xx series. I'd like to do so on the next 10xx series build but would prefer to test on a SLI build (one w/ backplate on and one off). But w/ currennt CPU and driver limitations, only recommending SLI builds for 4K and can't recommend 4K until we break 120 Hz w/ Display Port 1.4 monitors so don't expect an opportunity anytime soon. If one comes up will come back and post the results.
Facebook
Twitter
Instagram