Nvidia GeForce RTX 3060 Ti - Ampere becomes even smaller and still shows its claws | Page 2

Board analysis and power supply

The card assembled as usual at Foxconn has the same multi-layer board without the more complex backdrill procedure as already the GeForce RTX 3070 FE and thus also remains quite short. The kind of power supply is again similar to Turing and there is no need to split the main power supply in two parts. MSVDD is omitted and only NVVDD (blue) remains and stands for the conventional core voltage, which is what we like to call GPU voltage. Here there are a total of 8 individual phases (a maximum of 10 would be possible, on the RTX 3070 there are 9), which are provided by a uP9512R from UPI on the back (blue). This is a digital PWM controller with VID interface, compatible with NVIDIA’s Open VReg specification.

These 9 phases are each equipped with the somewhat cheaper 50-A powerstages AOZ5311NQI from Alpha & Omega, which do not allow real MOSFET DCR, but offer a temperature protection circuit and a readout output. The voltage range of NVVDD is between 0.7 and a maximum of 1.2 volts, whereby the maximum value can never be reached by the end user without special firmware and software.

For the more frugal GDDR6 memory (FBVDDQ), NVIDIA uses a simple step-down converter and discrete assembly (rear and front), again using simpler Dual N-Channel MOSFETs instead of PowerStages (front). So you find two very cheap SM7342EKKP from Sinopower on the board. But they’ll do.

For input smoothing, a coil with 1 µF behind which a shunt is located is used both for the 12-pin connection, which is only routed internally as a rail, and for the PEG motherboard slot (on the back). The flowing currents are measured via the voltage drop of the shunt. A uS5650Q from UPI is used for this purpose (back). To the left you can also see the 1 GB GDDR6 memory module with the 16 Gbps from Samsung. That’s all the excitement you’re gonna get.

The following two pictures show an AOZ5311NQI from Alpha & Omega representing all PowerStages, as well as

On the left side of the picture you can see the already mentioned 60A N-Channel MOSFET from Sinopower for the memory and on the right side exemplarily one of the many coils for GPU and memory, whose inductance is unfortunately not known.

Cooler and disassembly

Taking the radiator apart is once again easy, a feat. It is virtually identical in construction to the RTX 3070 FE. Loosen six cute Torx 8 screws and the backplate is gone, four more and the slot bracket falls too. The rest is then quickly done, only the fixing with the clamping circle is no longer square, but rectangular. Torx 6 and ready. The body is mainly made of light metal with a quite clever surface and has a valuable feel. The actual cooler is multi-part and consists of an efficient full copper heatsink. Three flattened heatpipes are soldered to the back of the heat sink, two of which extend longitudinally to the end of the card and one to the slot bracket.

Above the heatsink and the open structure, there are a total of two PWM-controlled 8 cm fans with a very interesting rotor blade geometry, which is almost reminiscent of the design of radial fans. Well, almost. But also the maximum speed of the two separately controlled fans of up to 3500 rpm lets one suspect evil at first, but it can also remain quiet. Depending on what is really called up. Let’s hope that it will remain much quieter than the card could if it had to.

The backplate in turn is made of light metal with a matt black coating on the outside, and the manufacturer thermally connects a limited area under the coils of the NVVDD phases to the backplate.

There has already been a lot of speculation about the principle with the different suction and blowing directions, so I won’t repeat myself but only try the foil again. We will see the test in the closed housing anyway.