TECH NEWS – Since the cards are currently sold with 8 GB of memory, with a 16 GB Ti model also available, this would be a change… but it would absolutely come with a downside.
Recently, rumors have been circulating that Nvidia is working on a 9 GB version of the GeForce RTX 5050, but it now looks as though more higher-memory models may be on the way as well. According to China’s Board Channel Forums, Nvidia is preparing to expand the GeForce RTX 5000 family with two additional variants: the GeForce RTX 5060 Ti and the GeForce RTX 5060. Those cards are already available in stores, but the new versions would ship with more VRAM.
It appears that, as Samsung and Micron ramp up production of 3 GB GDDR7 modules, Nvidia’s AIB partners have been instructed to refresh the memory configuration of the models mentioned above. In practice, that would mean the GeForce RTX 5060 Ti and RTX 5060 moving from their current 8 GB baseline to 9 GB of memory. The RTX 5060 Ti does of course also have a 16 GB model already, but that variant costs $50 more than the 8 GB one.
The three biggest memory manufacturers have already established production capacity for 3 GB GDDR7 modules. Nvidia is reportedly planning a multi-supplier strategy to ease the pricing pressure caused by VRAM shortages. On top of that, 3 GB modules make it possible to increase memory capacity without widening the memory bus, which sounds like an optimal balance between cost and performance. Compared with traditional 2 GB modules, a single 3 GB GDDR7 chip can boost VRAM capacity by 50% at the same bus width, helping relieve VRAM bottlenecks in high-resolution gaming and AI inference tasks. As Samsung and Micron continue expanding output, these modules are expected to debut on the market in the near future. Nvidia has reportedly already folded this strategy into its product roadmap. Sources within the supply chain say the 3 GB GDDR7 modules are currently in the product design phase, with products using them expected to appear officially in late May or early June.
Reports claim that the laptop RTX 5090 series has already started using 3 GB VRAM modules. This time, the shift is said to extend directly to the rest of the RTX 5000 lineup – specifically mainstream consumer-level parts like the RTX 5060 Ti 8 GB and RTX 5060 8 GB, which could switch straight to 3 GB modules. These models would be configured with three 3 GB modules each, giving them the planned 9 GB total VRAM capacity. This approach allows Nvidia to increase memory capacity while keeping the GPU core and the rest of the specifications unchanged, making it look like a convenient solution for easing the severe VRAM-related price hikes and supply shortages affecting the market right now.
However, while 9 GB would mean a 1 GB bump in VRAM, it would also cut total available bandwidth. The current 8 GB RTX 5060 Ti and RTX 5060 both use a 128-bit interface, whereas a 9 GB configuration would only need three 3 GB modules tied to a 96-bit bus. If Nvidia keeps the same 28 Gbps memory speed as the earlier models, that would result in total bandwidth of 336 GB/s. The existing models, by comparison, sit at 448 GB/s thanks to the 128-bit interface. That is a 33% gap, which is pretty substantial when these cards would already be dealing with serious bandwidth limitations. Nvidia could try raising speed to 30 Gbps, but that would still only mean 360 GB/s, which remains well below the current models. In other words, both cards would be giving up a big chunk of bandwidth in exchange for one extra gigabyte of VRAM.
That could save Nvidia some money, since it would only have to supply three GDDR7 modules to partners instead of four, but whether any of that would translate into lower prices for buyers is another question entirely. The rest of the cards’ technical specs are expected to stay the same, so hopefully pricing will remain unchanged too. According to current plans, Nvidia’s AIC partners are expected to introduce both the GeForce RTX 5060 Ti 9 GB and RTX 5060 9 GB around May or June, which means we will probably hear more about them at this year’s Computex.
Leave a Reply