Nvidia and partners are purportedly concocting at least three more GeForce RTX 30-series graphics cards, according to Twitter leakster Zed Wang, aka MEGAsizedGPU (opens in new tab). In the pipeline, according to this latest green team rumor, are an RTX 3060 8GB, an RTX 3060 Ti with GDDR6X, and an RTX 3070 Ti based on the GA102 GPU. However, we think adding these cards shortly would signal that there are still several months to wait for mid-range GeForce RTX 40-series (Ada Lovelace) cards to hit the market.
If true, introducing a new GeForce RTX 3060 with 8GB of RAM might make the biggest splash, as the X060 is often the most popular SKU of any generation. Of course, this new model will have less VRAM than the original (12GB cut to 8GB), but there is no light shone on any other spec changes. In the worst-case scenario, Nvidia might cut the bus from 192-bit to 128-bit for 8GB, reducing both memory quota and bandwidth. Without rebalancing in some other way (CUDA cores, clocks, memory speeds), it will simply be inferior to the currently marketed RTX 3060 and be a change that would draw heavy fire from enthusiasts.
A more positive change might be coming to the GeForce RTX 3060 Ti refresh if Zed Wang’s info is correct. This model is said to be getting a significant memory speed upgrade. It will move from 8GB of 14 Gbps GDDR6 to the same quota, using 19 Gbps GDDR6X. Given this info, we should expect about a 35% boost to memory bandwidth.
Last but not least, the source reckons there will be a GeForce RTX 3070 Ti leveraging the GA102 GPU. This GPU powers Nvidia’s flagship cards which it seems keen to shift recently. Perhaps there are bins full of GA102 GPUs with just enough working cores for this new RTX 3070 Ti. Such economics are often the reason behind production re-jigs like this.
| RTX 3060 | RTX 3060 8GB* | RTX 3060 Ti GDDR6X* | RTX 3060 Ti | RTX 3070 Ti* | RTX 3070 Ti | |
|---|---|---|---|---|---|---|
| GPU | GA106 | GA106 | GA104 | GA104 | GA102 | GA104 |
| Die Size | 276 mm² | 276 mm² | 392 mm² | 392 mm² | 628 mm² | 392 mm² |
| Memory Bus | 192-bit | 128-bit | 256-bit | 256-bit | 256-bit | 256-bit |
| Memory Size | 12GB GDDR6 | 8GB GDDR6 | 8GB GDDR6X | 8GB GDDR6 | 8GB GDDR6X | 8GB GDDR6X |
| Memory Clock | 15 Gbps | 15 Gbps | 19 Gbps | 14 Gbps | 19 Gbps | 19 Gbps |
*Specifications are unconfirmed.
It is almost certain that Nvidia will launch the first Ada Lovelace architecture GPUs during a GeForce Beyond special broadcast a week from Tuesday. However, it is hotly tipped and customary for Nvidia to start at the premium consumer end of things and then work its way down over the following months. With Ampere, things were highly protracted, with the journey from GeForce RTX 3090 to RTX 3060 launches taking five months, and it’s best not to mention the RTX 3050 (add a further 11 months).
So, at GeForce Beyond on September 20, we expect the RTX 4090 launch and probably some details about pricing and availability. If we are lucky, we have similar information about the GeForce RTX 4080 and RTX 4070.
Tom’s Hardware’s graphics editor has recently warned not to buy into the high-end Ampere series. Meanwhile, Nvidia and partners push the RTX 3080 and above harder with price drops and the new Spider-Man Remastered promotion.
