Regularly topping the Steam hardware survey as the most popular graphics cards, Nvidia’s 60-class products are often the ‘go to’ product for the mainstream gamer, with the green team usually dominating, even when rival products can be just as good or even better. With 60-class, the power of the GeForce brand comes into its own and to be fair, Nvidia has delivered some excellent GPUs. The GTX 1060 is still supremely popular, while last-gen’s RTX 3060 is currently the cheapest GPU you can buy with 12GB of VRAM – an increasingly important spec point. You can think of the RTX 4060 as broadly on par with the old RTX 2070 Super – a $499 product back in 2019 – so it’s generally faster than the 3060, but it does ship with just 8GB of RAM.
That’s clearly a challenge for the new card, even though its suggested retail price is actually lower than its predecessor, the issue being that discounting means RTX 3060 can be a lot cheaper – and capable of superior price vs performance ratios in some scenarios while stocks last. The fact is that VRAM is becoming increasingly more important and Nvidia cutting it by a third gen-on-gen is problematic. That’s a shame because the inherent advantages of the Ada Lovelace architecture are all present and correct on the RTX 4060 and, in the case of DLSS 3 in particular, we’re seeing gaming feats on this £289/$299 product that nothing else can get close to.
So, to illustrate, in the video below, you’ll see that Cyberpunk 2077’s RT Overdrive mode is fully playable at 1080p on the RTX 4060 at 1080p resolution with frame-rates in the 50-70fps range with a 58fps average. Take a minor hit to quality with the RT Overdrive optimisation mode and that increases frame-rate by 36 percent, keeping you north of 60fps at all times. If this is the future of graphics – as Nvidia hopes it is – the RTX 4060 democratises that experience.
| RTX 4080 16GB | RTX 4070 Ti 12GB | RTX 4070 12GB | RTX 4060 Ti 8GB | RTX 4060 8GB | |
|---|---|---|---|---|---|
| Processor | AD103 | AD104 | AD104 | AD106 | AD107 |
| Transistors | 45.9B | 35.8B | 35.8B | 22.9B | 18.9B |
| Die Size | 379mm² | 295mm² | 295mm² | 190mm² | 146mm² |
| CUDA Cores | 9728 | 7680 | 5888 | 4352 | 3072 |
| Boost Clock | 2.51GHz | 2.61GHz | 2.48GHz | 2.54GHz | 2.46GHz |
| Memory Interface | 256-bit | 192-bit | 192-bit | 128-bit | 128-bit |
| Memory Bandwidth | 717GB/s | 504GB/s | 504GB/s | 288GB/s | 272GB/s |
| TGP | 320W | 285W | 200W | 160W | 115W |
| PSU Recommendation | 750W | 700W | 650W | 550W | 550W |
| PSU Cables | 3x 8-pin | 2x 8-pin | 2x 8-pin | 2x 8-pin | 1x 8-pin |
| Base Price | $1199/£1199 | $799/£799 | $599/£589 | $399/£389 | $299/£289 |
| Release date | November 16th, 2022 | January 5th, 2023 | April 13th, 2023 | May 23rd, 2023 | June 29th, 2023 |
Elsewhere, the fact that you’re getting performance broadly in line – or better than, in many cases – the RTX 2070 Super isn’t likely to be a fact picked up by many other outlets, but it’s important to Digital Foundry because that’s our current reference point for a console-equivalent GPU in the PC graphics space. Factor out the maxed settings benchmarks, bring in optimised settings and DLSS and what Nvidia reckons is a class-leading 1080p GPU actually transforms into a console-like ‘4K’ performer.