Introduction

ASUS and Noctua are easily two of the most respectable brands in the PC components industry. The two came up with a wacky and unlikely area of collaboration—ASUS's expertise at making graphics cards and Noctua's mastery with air cooling solutions of all shapes and sizes. The result is the ASUS GeForce RTX 3070 Noctua OC. This graphics card combines an overclocked ASUS TUF Gaming GeForce RTX 3070 PCB with a custom air-based thermal solution that's cooled by two Noctua NF-A12x25 120 mm fans.



Why ASUS and Noctua chose to collaborate on the GeForce RTX 3070 as opposed to the newer RTX 3070 Ti, more powerful RTX 3080 or something from the AMD camp, such as the super-efficient RX 6800, has something to do with supply. The RTX 3070 remains a successful middle-of-the-market SKU by NVIDIA. This is the newer LHR (lite hash-rate) stepping, which comes with half the cryptocurrency mining performance of the original RTX 3070.

The RTX 3070 retains its performance outlook when it comes to gaming—it beats the previous-generation flagship, the RTX 2080 Ti, which means maxed out gaming with RTX-on at 1440p, or even 4K UHD with fairly high settings. It is based on the 8 nm "GA104" silicon and endowed with 5,888 out of 6,144 CUDA cores on the silicon. It also gets 184 third-generation Tensor cores and 46 second-generation RT cores. The memory configuration is unchanged from the previous-gen RTX 2070—8 GB of 14 Gbps GDDR6 across a 256-bit wide memory interface.

The GeForce RTX 3070 is based on NVIDIA's "Ampere" graphics architecture, which heralds the second generation of the company's path-breaking RTX real-time ray tracing technology. It combines 2nd Gen RT cores that double the ray intersection performance over the previous generation, adding hardware acceleration for more effects, such as ray traced motion blur, and 3rd Gen Tensor cores, which leverage the sparsity phenomenon in AI deep-learning neural networks to increase inference performance by an order of magnitude.

The ASUS RTX 3070 Noctua OC, as we mentioned earlier, features a custom-design cooling solution designed in collaboration with the Austrian company. It features an extremely thick compound aluminium fin-stack heatsink that's ventilated by a pair of large 120 mm Noctua fans that have been extensively tuned by the company to prioritize low noise above all. The card is a whopping four slots thick. This shouldn't be too much of a problem given multi-GPU is dead. Only those with SFF or Mini-ITX builds may not find this card suitable. The card is positioned at an MSRP of $830, but given current market conditions, we expect the card will retail for $950 at least, possibly more than $1000.

Packaging

The Card

Even at a first glance, this card screams "Noctua." Both companies were wise to keep the brown color theme of the Noctua fans, though I'm sure a lot of people will hate on the design. For a first product, you need to grab the viewer's attention, and they've succeeded at that. On the back, you'll find a metal backplate with a cutout to let airflow through; here, the color theme is black with yellow/gold highlights.


Dimensions of the card are 31.0 x 14.5 cm, and it weighs 1578 g.


Installation requires four slots in your system. This is a really thick card due to the full-width fans. SLI is dead, so I'm not seeing any issues here space-wise.


Display connectivity options include three standard DisplayPort 1.4a and two HDMI 2.1. The DisplayPort 1.4a outputs support Display Stream Compression (DSC) 1.2a, which lets you connect 4K displays at 120 Hz and 8K displays at 60 Hz. Ampere can drive two 8K displays at 60 Hz with just one cable per display.

Ampere is the first GPU to support HDMI 2.1, which increases bandwidth to 48 Gbps to support higher resolutions, like 4K144 and 8K30, with a single cable. With DSC, this goes up to 4K240 and 8K120. NVIDIA's new NVENC/NVDEC video engine is optimized to handle video tasks with minimal CPU load. The highlight here is added support for AV1 decode. Just like on Turing, you may also decode MPEG-2, VC1, VP8, VP9, H.264, and H.265 natively, at up to 8K@12-bit.

The encoder is identical to Turing. It supports H.264, H.265 and lossless at up to 8K@10-bit.


ASUS includes a dual-BIOS feature with the Noctua card. The default BIOS is "quiet," and the "performance" BIOS will run a more aggressive fan curve that achieves lower temperatures at the cost of higher noise.


The card has two 8-pin power inputs. This configuration is rated for up to 375 W of power draw.


The GeForce RTX 3070 does not support SLI. Only the RTX 3090 does, and it has very limited SLI support.

Teardown

The first component that comes off is the cooler shroud, which really contains standard full-size NF-A12x25 120 mm fans. This makes it super easy to swap out the fans in the future, for different models or to replace a broken fan.


The heatsink uses five heatpipes to transport heat away from the GPU. The thermal pads on the memory are 2.0 mm thick.


Removing the cooler shroud wasn't as easy as I expected, reason being that it is glued to the heatsink, so it'll take additional force.


ASUS engineered the heatsink specifically for this Noctua card. It's a huge design that works very well. The pads for the VRM are 1.8 and 2.0 mm thick.


Once the main heatsink is removed, a small VRM heatsink becomes visible. The thermal pads here are 2.0 mm and 1.8 mm thick.


The backplate is made out of metal and protects the card against damage during handling and installation.

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

GPU voltage is 10-phase and managed by an OnSemi NCP81610 controller.


The GPU VRM uses Vishay SIC654A DrMOS components rated for 50 A.


Memory voltage uses a two-phase design and is generated by a UPI uP9529Q controller.


For memory, 50 A Vishay SIC654A DrMOS chips are used, too.


The GDDR6 memory chips are made by Samsung and carry the model number K4Z80325BC-HC14. They are specified to run at 1750 MHz (14 Gbps GDDR6 effective).


NVIDIA's GA104 graphics processor is the company's third Ampere architecture chip, the second one targeted at GeForce gamers. It is produced on a 8 nanometer process at Samsung and has a transistor count of 17.4 billion with a die size of 392 mm².