As rumoured, the RTX 3060 Ti has a baby brother: the plain ‘ol RTX 3060. The new graphics card ought to look identical from the outside - although it’s somehow pictured with only a single fan in Nvidia’s press materials - but it uses Nvidia’s low-end GA106 GPU rather than the mid-range GA104 we saw at the heart of the 3060 Ti and 3070.
From the charts provided by the Green Team, the 3060 should deliver roughly RTX 2070 or 2070 Super levels of performance - meaning comfortable 1080p or decent 1440p gaming in AAA titles, perhaps stretching to 4K in older games or at reduced settings.
Interestingly, the card features 12GB of GDDR6 memory, meaning it has more VRAM than both the RTX 3060 Ti and the RTX 3070, which have just 8GB. It also has faster 15Gbps memory, compared to 14Gbps on the RTX 3060 Ti and RTX 3070. However, this is somewhat balanced by its 192-bit memory interface, which is narrower than the 256-bit equivalent on the two more expensive cards.
Compute is also reduced significantly, with the 3060 having only 3584 CUDA cores compared to 4864 on the 3060 Ti and 5888 on the 3070. The smaller GPU means that the card is able to draw less power than the 3060 Ti, with a rated TDP of 170W compared to 200W on the Ti.
It’s hard to gauge how these trade-offs will affect performance, especially within the memory subsystems. Presumably the 3060 will be able to hold more textures in memory at once, perhaps making it a bit more future-proof especially for RT titles, but loading textures could take slightly longer and therefore frame-rates might be a touch lower, while the reductions to compute should result in a significantly slower card overall. We’ll have to test this for ourselves once the card arrives at DF HQ.
The card’s recommended retail price is $329, which is $70 below the RTX 3060 Ti, and it’ll be available from late February… although it’s hard to imagine that the stock scarcity issues that have plagued the RTX 30-series from the beginning will have been resolved in just a month’s time.
The RTX 3060 will also be available on laptops starting on January 26th, alongside mobile versions of the RTX 3070 and RTX 3080 - which conveniently now come with a ‘Laptop’ suffix so you can easily tell them apart from the desktop variants. Happily, this date is when AMD’s Ryzen 5000 laptops will be available too, so many models will get both a CPU and GPU update at the same time - something that doesn’t always happen when CPU and GPU releases are further apart.
While the new laptop parts use the same second-gen RT cores and third-gen Tensor cores as other RTX 30-series cards, the overall configurations here are quite different. For example, the RTX 3080 Laptop gets either 8GB or 16GB of GDDR6 and 6144 CUDA cores, while the desktop 3080 uses 10GB of faster GDDR6X and sports 8704 cores. That makes it hard to judge relative performance between the two; proper real-world testing will be the best way forward here.
Unfortunately, this is further complicated by the number of different configurations available to laptop makers, which include more than a dozen Max-P and Max-Q RTX 3080 Laptop variants alone, with varying clock speeds, memory speeds and power targets.
The comparison to previous-gen laptop parts is a little more straightforward, with Nvidia claiming that the RTX 3080 is around 50 per cent faster than the 2080 in games like Control, Minecraft RTX and Borderlands 3 at 1440p. The 3070 is described as up to 50 per cent faster than the RTX 2070 and the RTX 3060 is described as 30 per cent faster than a PS5… and no, we don’t know that this means either. In general though, we’re seeing slightly smaller gen-on-gen gains on the laptop side of things than we saw on desktop, likely thanks to the more strict power constraints in these mobile form factors.
As well as the new hardware, there are some new features to be aware of too. One of the biggest is Resizeable BAR support, which provided single to low double-digit performance gains on the RX 6000 series cards. It works by allowing all of a card’s VRAM to be accessed directly, rather than via a 256MB I/O buffer. It’ll be interesting to see if this is feature is available on all RTX 30-series laptops, and whether it’s just enabled in the background or whether it can be toggled on and off to gauge its performance multiplier.
Note that Resizeable BAR will also ship on the RTX 3060 desktop graphics card, with VBIOS updates required to enable the feature on earlier RTX 30-series cards. To make use of the feature, you’ll need to be running an AMD B550 or X570 motherboard that supports the feature; on Intel some Z490 boards and all Z590 boards should support the feature as well.
Nvidia’s Max-Q framework now includes some updated features as well, including Dynamic Boost 2.0 and WhisperMode 2.0. The idea behind Dynamic Boost 2.0 is to increase GPU or VRAM power at the expense of CPU power when a game would benefit from it, thereby increasing frame-rates in GPU-bound scenarios. This works on a frame-by-frame basis, thereby allowing for quite dynamic power allocation. WhisperMode 2.0 runs the other way, minimising heat and noise in exchange for lowered performance; this now runs at a system level rather than in GeForce Experience.
As we mentioned before, availability for these new 30-series laptops begins on January 26th, with Nvidia promising more than 70 laptop models from various OEMs - like Acer, Asus, Dell, Gigabyte, Lenovo and Razer - right out of the gate. These start at $1000 for RTX 3060 machines up to around three times that for the most well-equipped RTX 3080 models. If you’re in the market for a laptop upgrade, be sure to check the CPUs as well - some models will come with last-gen AMD or Intel chips rather than the new Intel 11th-gen or AMD Ryzen 5000 processors, and it’s always nice to get the latest components if you can.
That’s all for now, but we’ll be back soon to share some more CES 2021 highlights - including Razer’s incredibly bonkers announcements, the best HDMI 2.1 monitors to look out for and much more. In the meantime, why not check out our breakdown of AMD or Intel’s CES 2021 announcements?
