Introduction

Gigabyte today released the GeForce RTX 3050 Gaming OC, its top custom-design graphics card powered by the RTX 3050 Ampere GPU, which NVIDIA is debuting today. The RTX 3050 is an entry-mainstream product for 1080p gaming with fairly high details, or with moderate details and ray tracing enabled, with DLSS to keep frame-rates high. For most others, the appeal is in it being the most affordable RTX 30-series card, meant for 1080p e-sports gaming. Gigabyte is giving the chip an elaborate cooler and a factory-overclock.

The GeForce RTX 3050 is based on the same 8 nm GA106 silicon as the RTX 3060, although heavily cut down. While the RTX 3060 features 28 out of 30 streaming multiprocessors (SM) present on the chip, the RTX 3050 only has 20 out of 30. This results in 2,560 CUDA cores, 20 RT cores, 80 Tensor cores, 80 TMUs, and 32 ROPs. The card features 8 GB of GDDR6 memory across a 128-bit wide memory bus, compared to 12 GB across 192-bit for the RTX 3060. The memory clock is lowered, too, at 14 Gbps vs. 15 Gbps, resulting in a memory bandwidth of 224 GB/s, as opposed to 360 GB/s on the RTX 3060.



A key piece of specification that sets the RTX 3050 apart from the RTX 3060 is the PCI-Express bus width of just 8 lanes, that's PCI-Express 4.0 x8, compared to x16 on the RTX 3060 even though the GA106 is perfectly capable of x16, and all the custom-design boards we're testing today reuse PCB designs from RTX 3060 products with PCB traces for all 16 lanes in place. NVIDIA offers this explanation: "Dropping to 8 PCIe lanes improves supply. It allows us to source a wider variety of chips for the life of the product." In other words, NVIDIA is currently harvesting GA106 chips that didn't make the cut for RTX 3060, but could in the future switch to the smaller GA107 silicon, which physically has 24 SMs (3,072 CUDA cores) and a 128-bit GDDR6 memory bus, along with a PCI-Express 4.0 x8 bus. This switch would come with no difference in performance.

The Gigabyte RTX 3050 Gaming OC, which we are reviewing today, comes with a strictly 2-slot design, but one that's over 28 cm long, featuring an elaborate aluminium fin-stack heatsink with three 80 mm fans ventilating it. The PCB is only two-thirds the length of the card, so all of the airflow from the third fan vents through the heatsink and out of a large cutout in the backplate. The cooler also features some RGB illumination. On offer is a factory-overclock of 1822 MHz, compared to 1777 MHz reference, while the memory is left untouched. The card draws power from a single 8-pin PCIe power connector. Display outputs include two each of DisplayPort 1.4a and HDMI 2.1 connectors, which is an interesting mix. Gigabyte is pricing this card at $379 MSRP, a fairly big premium over the $249 base-pricing fantasy for the RTX 3050.

Packaging

The Card

The Gigabyte RTX 3050 Gaming OC is instantly recognizable as a recent Gigabyte card, sharing the design theme of their latest AMD and NVIDIA cards. A high quality metal backplate is installed, too.


Dimensions of the card are 28.0 x 12.0 cm, and it weighs 707 g.


Installation requires two slots in your system.


Display connectivity options include two HDMI 2.1 and two DisplayPort 1.4a ports.


The card has one 8-pin power input. This configuration is rated for up to 225 W of power draw.


The GeForce RTX 3050 does not support SLI.

Teardown

Gigabyte's thermal solution uses two heatpipes that make direct contact with the GPU core. The main cooling assembly also provides cooling for the memory chips and VRM circuitry.


Once the main cooler is removed, you can clearly see how the PCB is shorter than the actual cooler.

High-resolution PCB Pictures

These pictures are for the convenience of volt modders and people who would like to see all the finer details on the PCB. Feel free to link back to us and use these in your articles, videos or forum posts.



High-res versions are also available (front, back).

Circuit Board (PCB) Analysis

The GPU VRM is a five phase-design controlled by an OnSemi NCP81610 controller.


OnSemi NCP302155 DrMOS chips are used for GPU voltage; they are rated for up to 55 A output current.


The memory VRM is single-phase and managed by a uPI uP1666Q controller.


For memory, an Alpha Omega AON6994 dual-MOSFET is used. It's specified for currents up to 82 A.


The GDDR6 memory chips are made by Micron and carry the model number D9ZPM, which decodes to MT61K512M32KPA-14:C. They are specified to run at 1750 MHz (14 Gbps GDDR6 effective).


NVIDIA's GA106 graphics processor is made using Samsung's 8 nanometer node and has a transistor count of 12 billion with a die size of 276 mm².