Board & Cooling
Our eyes are immediately drawn to this board's excessive number of coils arranged one below the other, like pearls on a chain. Marketing sells this as an array of 16 power phases for the GPU, which of course isn't exactly true since there are, in fact, just eight real phases. Doubling is what gives each phase two separate converter circuits.
Thus, we end up with 16 converter circuits for eight phases. Three phases dedicated to the memory are moved to the left, behind the GPU voltage converters' coils. This turns out to be an unfortunate choice, as the resulting hot-spots would be a challenge for any cooling solution to handle.
The GPU's eight real phases of course need an appropriate PWM controller as well. For this, the card does not rely on a uPI Semiconductor uP9511 as so many do, but an International Rectifier IR3595A instead.
One International Rectifier PowIRStage IR3555 per phase controls two parallel circuits, each of which gets two MOSFETs for high-side and low-side, as well as a Schottky diode. Again, this means that there are a total of 16 individual voltage converter circuits, pairs of which are always working in parallel. That is not, dear marketing department, 16 individual phases.
The encapsulated ferrite core chokes are a common sight. Their design helps to elegantly solve the problem of laying out 16 coils and their respective converter circuits on one PCB.
A total of 11 Micron MT58K256M321JA-110 GDDR5X ICs are organized around the GP102 processor. They operate at 11 Gb/s data rates, which helps compensate for the missing 32-bit memory controller compared to Titan Xp. We asked Micron to speculate why Nvidia didn't use the 12 Gb/s MT58K256M321JA-120 modules advertised in its datasheet, and the company mentioned they aren't widely available yet, despite appearing in its catalog. Because Nvidia sells its GPU and the memory in a bundle, Galax has very little room to innovate in this regard.
As mentioned, the memory's power supply is positioned to the left of the GPU's. Here, Galax goes with a uPI Semiconductor uP9509, a buck controller able to support all three phases.
For each phase, you'll find a MagnaChip MDU1514 on the high side and a MDU1511 on the low side. The coils are the same as those used for the GPU's voltage converters. And the attached mounting frame features corresponding openings for these tall modules.
A familiar Texas Instruments INA3221 monitors current and voltage levels. However, on this card the manufacturer decided to install two chips rather than one. Three shunts in the input area take the current flow to be monitored. And with three LC filters behind the three eight-pin auxiliary power connectors, the card even has some filtering for spikes.
Around back, we get a look at the linearly-arranged components. This positioning makes it easier to provide cooling via the backplate.
The backplate, which is painted black on the inside, serves more than an aesthetic purpose. It also facilitates additional passive cooling.
We already mentioned that the main sink has no built-in surface to draw heat away from the voltage regulation circuitry. Similar to MSI's and EVGA's cards, this duty is instead addressed with a frame between the thermal solution and PCB. But given the more extreme thermal load Galax allows, that approach quickly reaches and exceeds its limit.
The cooler itself is a massive piece of work. Its copper sink feeds waste heat from the GPU to a total of four 8mm pipes and a fifth 6mm one. Cooling for the memory modules comes from a circumferential copper plate attached to the brushed heat sink.
The card's three 90mm fans are equipped with nine rotor blades each. A rather steep angle suggests that they are optimized for lots of airflow, rather than constant static pressure.
MORE: Best Graphics Cards
MORE: All Graphics Content