With the RTX 4060, Nvidia is launching a new generation of video cards for the still very popular 1080p gaming segment. At the end of last year, the manufacturer started rolling out the RTX 40 series with the fastest models, and then the lower ones were gradually upgraded. Now we come to the RTX 4060, a video card focused entirely on 1080p gaming and should be available from €329.
Small AD107 GPU for the big mainstream segment
Those familiar with the symbols on Nvidia chips know that a lower last number means a larger, more fancy chip, and a higher number means a smaller GPU. The AD107 is the smallest graphics processor based on the Ada Lovelace architecture and is used for the laptop versions of the RTX 4050 and 4060. For the desktop version of the RTX 4060, Nvidia now also uses the AD107, while the manufacturer has been using it for generations. It uses the 106-chip for x60-class video cards. Compared to GTX 1060, RTX 2060, and RTX 3060, RTX 4060 is a downscaling.
Any chip that sits below your video card’s cooling block will interest most gamers much less than the card’s price-performance ratio. The largest group of PC gamers still play at 1080p, according to numbers from the Steam Hardware Survey, and based on the popularity of previous x60-series GeForce cards, purchase price plays a very important role for this majority.
However, it is worth comparing the specifications of the RTX 4060 with those of its predecessors, which were equipped with 106 GPUs. Even the RTX 3050 was released with a GA106 GPU, albeit with a third of the computing cores disabled. The AD107 chip is smaller than the AD106 GPU of the RTX 4060 Ti, and therefore cheaper to produce.
The fact that we’re only seeing video cards in limited form with larger chips partially disabled in this generation is a sign that it’s a thing yields TSMC’s product concerns are good. This means that manufacturers like Nvidia have to make choices that differ from how they built their portfolio with previous generations. For example, the RTX 4080 actually got “its” AD103 GPU instead of the AD102 chip which wasn’t good enough for the RTX 4090. Probably because there were so few objectionable chips that were taken out of range. And now Nvidia is using the 107 chip for an x60-class video card for the first time, while in the past the codename was reserved for lower-score boarding cards.
The desktop version of the RTX 4060 gets the AD107 fully enabled and therefore has 3072 active compute cores. That’s less than the RTX 3060, but more than the RTX 2060, though Nvidia has started counting its CUDA cores differently since the RTX 30 series. The number of Tensor cores, RT, texture units, and output processors is also lower than that of its predecessors. A newer, more efficient architecture should compensate for this difference and the greatly increased amount of cache should absorb step hits to the narrower 128-bit memory bus. Bottom line, the RTX 4060 should run more efficiently, which is perfectly in line with one of the most important laptop GPU characteristics.
As mentioned, it is mainly about whether the RTX 4060 can live up to its expectations. This new GeForce card falls into the price range where most buyers can be found, and Nvidia likes to point out how popular its previous x60-series video cards were in the Steam Hardware Survey. Compared to its predecessors, the RTX 4060 also has a more favorable suggested retail price. In addition, the RTX 4060 provides support for features such as DLSS 3 and AV1 encoding thanks to the Ada Lovelace architecture. You can read about the performance in this review.
The GeForce RTX 4060 is available today with models tracking suggested retail price, with custom models from board partners tomorrow afternoon. Nvidia doesn’t release a Founders Edition for the RTX 4060. For this review, we ran our tests with a Gigabyte GeForce RTX 4060 Windforce OC.
video card | RTX 4060 Ti | RTX 4060 | RTX 3060 | RTX 2060 |
---|---|---|---|---|
GPU | m 106 | m 107 | GA106 | TU106 |
building | Ada Lovelace | Ada Lovelace | Ampere | Turing |
practical | TSMC ‘4N’ 5 nm | TSMC ‘4N’ 5 nm | Samsung 8 nm | TSMC 12 nm |
die size | 190 mm² | 156 mm² | 276 mm² | 445 mm² |
transistors | 22.9 billion | 18.9 billion | 13.25 billion | 10.8 billion |
CUDA cores | 4352 | 3072 | 3584 | 1920 |
Tensioner cores | 136 | 96 | 112 | 240 |
RT cores | 34 | 24 | 28 | 30 |
texture units | 136 | 96 | 112 | 120 |
robs | 48 | 32 | 48 | 48 |
base clock | 2310MHz | 1830MHz | 1320MHz | 1365MHz |
clock increase | 2535MHz | 2460MHz | 1777MHz | 1680MHz |
L2 cache | 32 MB | 24 MB | 3 megabytes | 3 megabytes |
Fram | 8 GB GDDR6 | 8 GB GDDR6 | 12 GB DDR6 | 6 GB GDDR6 |
memory speed | 18 Gb/s | 17 Gb/s | 15 Gb/s | 14 Gb/s |
memory bus | 128 bits | 128 bits | 192 bits | 192 bits |
bandwidth | 288 GB / s | 272 GB / s | 360 GB / sec | 336 GB / s |
Tgp | 160 watts | 115 watts | 170 watts | 160 watts |
MSRP (In the foreground) |
449 euros (8 GB version) | 329 euros | 349 euros | 369 euros |
released | May 23, 2023 | June 28, 2023 | February 25, 2021 | January 7, 2019 |