AMD RDNA 5 Dual Issue Brings Major GPU Architecture Changes

The graphics card market is about to enter a period in which the typical two-year upgrade cycle has been disrupted, and what you can buy today depends as much on what's unavailable as on what is. Conflicting reports of a postponed next-generation Radeon graphics card release, Nvidia's shift in graphics card marketing approach, and the unexpected revival of older graphics cards all point to the same thing: a clash between waiting and buying now.

According to reports from board partners, the next generation of gaming chips, namely RDNA 5, could be available only towards the end of 2027 or even in early 2028. This timeline would mean AMD has RDNA 4 much longer than expected, which matters for buyers waiting for a new Radeon generation and could leave them with it for over a year beyond the typical cycle.

AMD's Next Generation Delay

RX 9070 GRE is already available in more regions than originally advertised; it has 12GB of GDDR6 and is rated at 220W of board power. It's positioned above the RX 9070 but still in the high-end 1440p gaming range. The latest tests show it delivers around 120 fps at 1080p and around 87 fps at 1440p in "rasterized" games.

The output is solid on paper, but in many games, ray tracing also gives Nvidia an advantage, while in 4K gaming, frames have to be upscaled to ensure smooth play. Below that, a report suggests a budget card option with 8GB of storage, which is another reason AMD needs something to rely on until RDNA 5 arrives.

AMD might keep the current generation alive, but Nvidia is reportedly working on super versions of its RTX 50 series, and the biggest upgrade might not be clock speed; it could be memory capacity. Reports suggest a12GB version of the RTX 5060, along with 8GB and 24GB versions of the RTX 5070 Super and an RTX 5080 Super,

Nvidia's Possible VRAM Response

Those cards would add to one of the biggest gripes about recent GPU generations: high-priced cards with limited memory, which concerns gamers, creators, and local AI users. Those models could arrive between the end of 2026 and CES 2027, leaving shoppers with a tough choice if they're looking to upgrade their GPUs now.

The next-generation battle is not the only one between AMD and Nvidia. One manufacturer has gone even further by reintroducing a graphics card that supports 12GB of VRAM, DLSS, ray tracing, and CUDA. Its memory capacity can also be more valuable than other newer 8GB cards for creators or those using local AI, though it requires a 550W power supply. The problem is that the chip was originally designed 5 years ago.

The card in question is the GeForce RTX 3060 12GB. Palit has resurrected it as the RTX 3060 Infinity 2 OC, and it's now being re-released in 2026. Its basic specifications remain largely unchanged: 3584 CUDA cores, 12GB of GDDR6 memory on a 192-bit bus, a boost clock of nearly 1.8 GHz, and about 170W of power consumption.

In addition to the familiar internals, Palit's new cooler includes a compact dual-fan cooler, a zero-RPM fan mode, and a top-mounted backplate. RTX 3060 is today around 62 fps on ultra at 1080p and around 42 fps on ultra at 1440p. In those same tests, the RTX 4050 reaches about 69 and 45 fps, respectively, and the RTX 4060 reaches about 89 and 60 fps, respectively.

While the RTX 3060 is playable for 1080p gaming, the RTX 4060 offers the same amount of memory but delivers outdated performance and should be viewed as a 1080p gaming chip with an upgrade. The extra memory helps in specific ways. 12GB can prevent texture and stutter issues caused by memory limits when the 8GB card is full, and it can also handle larger video projects, 3D scenes, and local AI models.

VRAM doesn't necessarily equal raw power, and a lower GPU speed doesn't mean it's fast just because of its memory. More importantly, the RTX 3060 lacks newer features from the Blackwell era, such as multi-frame generation, and it consumes more power than the RTX 5060, all of which point back to the ultimate determinant of success: price.

RTX 3060 is an interesting option for budget-conscious creators, CUDA users, and gamers.

The comeback doesn't make much sense for gaming, since the RTX 5060 is much faster and has newer features; however, official pricing hasn't been announced, though it's expected to be around $330. Intel's Arc B580 also offers 12GB of memory, but AMD's newer cards deliver more powerful rasterized performance at similar prices.

There is some reason for optimism: AMD might wait for the next big generation, Nvidia may counter with new VRAM-super cards, and the return of the RTX 3060 is due to memory capacity still driving sales even on older graphics cards.

But the basic rule remains: VRAM isn't the best indicator of which graphics card to buy. Comparing them by price, features, power consumption, and performance will give the most accurate way to compare any build to these options.