Colorful GeForce RTX 3060 Bilibili 12G review - Introduction

Colorful GeForce RTX 3060 Bilibili 12G review

Join us as we review the Colorful iGame GeForce RTX 3060 E-Sports edition OC, which has the same 12GB of memory and 3584 shading processors activated as the standard model, but with a factory boost clock of 1822 MHz. This graphics card has been tweaked right out of the box for you, and our tested model has a two-fan design for cooling. With that 3584 shading cores and Ampere architecture, this 3060 series is bound to impress in the 2560x1440 (WQHD) domain. If we look back at the previous generation, the product would sit at GeForce RTX 2070 (SUPER) performance levels and, in due time, will replace that series. If stock becomes available in plentiful volumes though. The GPU is again fabricated on an 8nm node derived from Samsung. This process further develops Samsung's 10nm process; no EUV is applied in production just yet. The first wave of announcements has seen the GeForce RTX 3080 and 3090 being released first, and, as a bit of a surprise, the GeForce RTX 3060 Ti and 3070. It's now late February 2021 and NVIDIA is set to release its more 'regular' 3060 prices 329 USD.

As you will have noticed, the 3060 GPU cores count is about 26 percent lower than with the RTX 3060 Ti, which has a GA104 chip with 4864 shading cores (shader/stream/cuda cores = all the same thing with a different name). NVIDIA is launching the 3060 series with the 12GB model, which's remarkably enough is 2GB more than the GeForce RTX 3080 (!). Later on, they'll likely silently slip in a 6 GB version, though that has not been confirmed. NVIDIA advertises the series with 13 'shader teraflops' and 25 'RT-ops', the latter giving an indication of the ray-tracing performance. Notable is that a change is in effect, the memory runs ar 15 Gbps as opposed to the usual 14 Gbps, likely to compensate for the perf hit of going 256-bit towards 192-bit on the memory bus due to that memory configuration. It's the same for the shader core cluster, it's clocked higher in the boost frequency compared to the Ti model, also compensating a bit for the lower number of shader cores. 

The Ampere lineup nearly doubles ray-tracing performance with Gen2 ray-tracing cores and 3rd iteration Tensor cores. These cards will all be PCIe 4.0 interface compatible and offer HDMI 2.1 and DisplayPort 1.4a, but most importantly is that exorbitant shader processor count (referred to as CUDA cores by NVIDIA). With just over a third of the shader processor count seen from the flagship product, we now meet the NVIDIA GA106 GPU. And despite being a lower segmented card, it still holds a sizable GPU die. In this round, NVIDIA is not seeding Founder edition cards, aka FE GeForce RTX 3060. But of course, they do present the reference specification; a boost clock of 1780 MHz and a base clock of 1320 MHz. 

Colorful GeForce RTX 3060 Bilibili 12G

So when we received the Colorful GeForce RTX 3060 Bilibili 12G admittedly I did get a smile on my face. Dual slot, indeed a colorful design, two fans, little RGB and even an OC button. I mean what's not to like. The card we are testing thus is a $389,- model. The differences between the two: 1. Higher Clockspeeds 2. Special edition design. Although the card's design is distinctive in that it borrows the general layout from the RTX 3080/3090, it has a more traditional short PCB, which makes it stand out from the crowd. The card is factory overclocked, with a boost clock of 1822 MHz and a slightly higher TDP of 187 W, and it is connected through a single 8-pin socket only.

The card, which is based on the Chinese social media network Bilibili and from what we understood is only be accessible for purchase on the platform. Some Chinese visitors of the Bilibili website have been able to get their hands on the limited-edition card through a lottery contest. As you can see the dual-fan spinner design makes the card really compact. Based on an advertised 187W TGP  the card is of course paired with 12GB GDDR6 graphics memory at 192-bit running at 15 Gbps. It is a more simplified-looking product that comes with a shy factory tweak, making it perform a few percent above reference/baseline performance. The card has a single (6+2) pin power header. And yeah, that factory tweak is 1822 MHz on the  Turbo (1780 MHz = reference), which is tangible alright. Let's have a look at how well the product performs.