GPU diode is a bad means for controlling fan RPM, at this point; it’s not an indicator of total board performance by any stretch of use. GPUs have become efficient enough that GPU-governed PWM for fans means lower RPMs, which means less noise – a good thing – but also worsened performance on the still-hot VRMs. We have been talking about this for a while now, most recently in our in-depth EVGA VRM analysis during the Great Thermal Pad Fracas of 2016. That analysis showed that the thermals were largely a non-issue, but not totally inexcusable. EVGA’s subsequent VBIOS update and thermal pad mods were sufficient to resolve any concern that lingered, though if you’re curious to learn more about that, it’s really worth just checking out the original post.
VBIOS updates and thermal pad mods were not EVGA’s only response to this. Internally, the company set forth to design a new PCB+cooler combination that would better detect high heat operation on non-GPU components, and would further protect said components with a 10A fuse.
In our testing today, we’ll be fully analyzing the efficacy of EVGA’s new “ICX” cooler design, to coexist with the long-standing ACX cooler. In our thermal analysis and review of the EVGA GTX 1080 FTW2 (~$630) & SC2 ICX cards (~$590), we’ll compare ACX vs. ICX coolers on the same card, MOSFET & VRAM temperatures with thermocouples and NTC thermistors, and individual cooler component performance. This includes analysis down to the impact the new backplate makes, among other tests.
Of note: There will be no FPS benchmarks for this review. All ICX cards with SC2 and FTW2 suffixes ship at the exact same base/boost clock-rates as their preceding SC & FTW counterparts. This means that FPS will only be governed by GPU Boost 3.0; that is to say, any FPS difference seen between an EVGA GTX 1080 FTW & EVGA GTX 1080 FTW2 will be entirely resultant of uncontrollable (in test) manufacturing differences at the GPU-level. Such differences will be within a percentage point or two, and are, again, not a result of the ICX cooler. Our efforts are therefore better spent on the only thing that matters with this redesign: Cooling performance and noise. Gaming performance remains the same, barring any thermal throttle scenarios – and those aren’t a concern here, as you’ll see.
There was one change to the VRM, but everything else remains largely the same. Let’s talk through that now.
EVGA FTW2 & SC2 PCB, VRM, & Power Analysis
Above: FTW2 PCB analysis
Below: FTW (1) PCB analysis
EVGA’s PCB and SMDs for the FTW2 & SC2 cards are more or less identical to the preceding FTW ACX and SC ACX cards, with just a few changes to the VRM. Originally, as Buildzoid detailed in our 1080 FTW VRM analysis, the EVGA GTX 1080 FTW used NCP81382 MOSFETs, which integrate the high-side, low-side, and driver IC into one chip. In the original design, this VRM allowed for about 350A continuous current (35A average current, 70A peak current for the FETs). The old VRM also relied on doublers to create its 10-phase, since no true 10-phase voltage controller exists on computer hardware. As Buildzoid said in the VRM analysis, most controllers stop at 8 phases.
And so EVGA has used doublers and reinforced its marketing efforts regarding the VRM design, now in production since the Kepler video cards. The doubler used is an NCP81162, somewhat of a “braindead” doubler. EVGA’s original FTW also used a controller with a PWM of 400kHz going into the doubler, with the phases individually seeing 200kHz (again, doubled). Overall power quality isn’t any better than a normal 5-phase, seeing as it’s just split attention between two phases, but “10-phase” VRM sounds a lot better. There is one upshot to this approach: it distributes the heat across greater surface area, so that is a legitimate reason to opt for the bigger, doubled-up phase design.
Regarding ICX and the changes received on the power management side, EVGA has swapped out the 35A-rated NCP81382 FET for a Fairchild FDMF3035 FET, rated for 50A. Everything else remains exactly the same, including the 4C805N OnSemi high- & low-side memory VRM FETs. As a refresher, the 4C805N is rated at 30A continuous at 125C, which is more than enough for VRAM, since you won’t see more than 20A through it anyway.
EVGA ICX Thermistors & Placement
Above is an image of the EVGA front-side and back-side of the new GTX 1080 FTW2 PCB, to be outfitted with the ICX cooler. For point of clarity: When we refer to the “front” of the PCB, we are talking about the side with the GPU.
The biggest part of the ICX cooler is its implementation of thermistors for real-time temperature sensing on a second-to-second basis, useful for diagnosing VRM and VRAM temperatures. As discussed above, this is a critical step forward in video card design, as the GPU dies are now efficient enough that nearly any reasonable aftermarket cooler will keep them well within limitations. This introduces an interesting challenge: With GPU temperatures lower, fans controlled by traditional PWM based on the GPU will spin at lower speeds than might be comfortable for a VRM.
The sensors talk to an MCU (microcontroller). EVGA hosts two MCUs on the new ICX coolers, both SC2 and FTW2 series cards, with one of the two dedicated solely to controlling the actually functional RGB indicator lights on the top of the card.
But back to the actual sensors: There are two primary types of sensor used in a PC hardware, with EVGA opting to use thermistors of the negative type. NTC thermistors have a resistance-temperature curve, responding to higher temperature results with lower resistance. The MCU reads the voltage off of the thermistor and then calculates the temperature of the sensed object, whose output is then to control fan RPM for the newly added VRM fan.
With a better understanding of how EVGA is sensing its temperatures, let’s now go back to the topic of placement of those thermistors:
The front of the PCB – the one hosting the GPU – positions three thermistors near the VRAM modules, with an additional four thermistors for controller and VRM (collectively “power”) components. It is not feasible to mount thermistors under the VRAM modules (which use a 170 BGA package, or sometimes 190 BGA for G5X), so EVGA must settle on the next best thing. The thermistors are therefore mounted just off to the side of the VRAM package, top-side of the PCB, in a position that EVGA determined internally to be close in temperature to the VRAM case temperature. In terms of testing and engineering, EVGA likely mounted thermocouples (as we do) or thermistors to the top of the VRAM packages, then moved around the thermistors on the PCB until the temperature matched as closely as possible.
EVGA’s memory thermistors are labeled 1-3, with “Memory 1” starting to the lower left of the GPU substrate, “Memory 2” at the top center, and “Memory 3” to the middle-right. It would be reasonable to hypothesize that Memory 2 will almost always run hottest, as it is encroached on by the capacitor banks and inductors for the VRM.
The thermistors labeled as “PWR” are number 1-5. Starting with the top of the PCB, or the “front,” the bottom-most thermistor is “Power 3” and is adjacent to MOSFET #2. Anyone who read our EVGA VRM thermal analysis content may recall that we determined MOSFET #2 (bottom-up) to be one of the two major hotspots and fault spots on the ACX PCBs. “Power 2” is placed closer to MOSFET #7, which is conveniently the next FET that we determined most worthy of a thermocouple. This FET runs hot because it’s dead-center of the FETs, and so gets a bit roasted by its neighbors. FET #2 (we originally hypothesized and later proved) could be more variable in temperature depending on the case setup, as it’s closest to the motherboard while still remaining flanked. Usually, that’s going to be more choked for air. “Power 1” is located next to FET #10, with “Mem Power 1,” sometimes called “Power 4” in Precision XOC, representing the memory VRM temperature.
The bottom of the PCB, or the “back,” contains two more sensors. One measures the back-side of the GPU temperature, the other measures the back-side of the small 1v PLL minor rail for non-compute functions within the GPU. The latter would be “Power 5,” the former is “GPU,” or sometimes “GPU Extra.”
EVGA ICX vs. ACX Changes to Heatsink & Backplate
Let’s dig into the changes to the heatsink proper.
With ICX, as we reported in our initial CES coverage, EVGA has modified its baseplate to accommodate new “pin fins” for additional surface area. The baseplate is also loaded with thermal pads for this generation of coolers, almost as if to say, “we get it! Here are your damn thermal pads!” These thermal pads bridge contact between the baseplate and the VRAM – all eight packages – the inductors, the FETs, minor power staging components, and even that insignificant 1v PLL. From here, EVGA’s baseplate then connects (again via thermal pad) to an updated fin stack. The fin stack uses L-shaped fins in a few key locations to provide more airflow to the baseplate from the fans, primarily over the FETs and inductors (where the thermal pads contact these L-shaped fins). This is preferential to a flat coldplate, where you’d lose some of the dissipation potential in favor of greater surface area. That’s another item discussed in our EVGA VRM final analysis from late last year.
The thermal pads are ~1mm for the most part, and are all rated at 1.2W/mK.
Other minor changes include holes bored through the fin stack, allowing better horizontal displacement of air over the surface of the PCB.
The biggest update, again, is individual fan speed control via MCUs hooked up to thermistors. The VRM fan can spin separately in this way, so you could run a 30% GPU fan speed that might run you close to 80C in some application, but still maintain a higher VRM fan speed. Keep in mind that the curve is different for each fan, so 45% does not = 45% on both fans. 45% on one might be ~1700RPMwhile the other approaches 2000RPM.