KitGuru KitGuru.net – Tech News | Hardware News | Hardware Reviews | IOS | Mobile | Gaming | Graphics Cards
Announced back at CES 2021, today is the official launch for the RTX 3060. Nvidia's latest Ampere GPU, the RTX 3060 is meant to be the cheapest, with prices starting at £299… but by now I think we all know that is unlikely to be the case any time soon. Still, RTX 3060 sees the introduction of the new GA106 GPU which Nvidia hopes will unlock a new level of performance in this mid-range segment. We put this card through its paces and find out what it has to offer.
With Ampere products so far, both RTX 3090 and RTX 3080 have used the GA102 GPU, while RTX 3060 Ti and RTX 3070 are based on GA104. Now, with the introduction of RTX 3060, Nvidia has unveiled its smallest Ampere-based chip yet – GA106, measuring 276 mm².
RTX 3060 also marks the point where Nvidia will no longer manufacturer a Founders Edition for this GPU. Last generation, RTX 2060 got a Founders card, but not so for the RTX 3060. For this launch review, we have tested Gigabyte's RTX 3060 Gaming OC 12G.
| GPU | RTX 3090 | RTX 3080 | RTX 3070 | RTX 3060 Ti | RTX 3060 |
| SMs | 82 | 68 | 46 | 38 | 28 |
| CUDA Cores | 10496 | 8704 | 5888 | 4864 | 3584 |
| Tensor Cores | 328 | 272 | 184 | 152 | 112 |
| RT Cores | 82 | 68 | 46 | 38 | 28 |
| Texture Units | 328 | 272 | 184 | 152 | 112 |
| ROPs | 112 | 96 | 96 | 80 | 48 |
| GPU Boost Clock | 1695 MHz | 1710 MHz | 1725 MHz | 1665 MHz | 1777 MHz |
| Memory Data Rate | 19.5 Gbps | 19 Gbps | 14 Gbps | 14 Gbps | 15 Gbps |
| Total Video Memory | 24GB GDDR6X | 10GB GDDR6X | 8GB GDDR6 | 8GB GDDR6 | 12GB GDDR6 |
| Memory Interface | 384-bit | 320-bit | 256-bit | 256-bit | 192-bit |
| Memory Bandwidth | 936 GB/Sec | 760 GB/Sec | 448 GB/Sec | 448 GB/Sec | 360 GB/sec |
| TGP | 350W | 320W | 220W | 200W | 170W |
Despite utilising new GA106 silicon, RTX 3060 isn't actually a full implementation of the GPU, as it has 28 streaming multiprocessors (SMs) instead of 30. Currently only RTX 3060 for laptops uses the full GPU. For the desktop chip, 28 SMs means a total of 3584 CUDA Cores, as thanks to Ampere’s new SM structure with its two FP32 datapaths, each SM houses 128 CUDA cores.
Ampere also places one RT core, and four Tensor cores, in each SM, giving a total of 28 RT cores and 112 Tensor cores. This is accompanied by 112 texture units, and 48 ROPs which are now housed directly within each graphics processing cluster (GPC), with 16 ROPs per GPC, and 3 GPCs in total for RTX 3060.
Interestingly, rated clock speed is the highest it has been for an Ampere GPU, coming in at 1777MHz for the reference spec. This Gigabyte Gaming OC model however, has increased this to 1837MHz, and we test real-world clock speed behaviour later in this review.
There's been a lot of talk about Nvidia's decision to use 12GB GDDR6 memory for the RTX 3060. This is more than the RTX 3060 Ti, RTX 3070 and even RTX 3080, but crucially the memory interface is much narrower at 192-bit. With memory clocked at 15Gbps, total memory bandwidth comes in at 360GB/s, about 20% lower than the RTX 3060 Ti.
Lastly, for total graphics power, Nvidia rates the RTX 3060 for 170W, which is 30W less than the RTX 3060 Ti. Despite being factory overclocked, the Gigabyte Gaming OC also shares this TGP rating.
