According to early rumors, NVIDIA planned to release three versions of the GeForce RTX 3070: a regular one based on the GA104-300 GPU, a Ti or Super version (PG141 SKU 0) based on a GA104-400 GPU with an increased number of functional blocks, and a variant on the PG141 SKU board. 5 with doubled GDDR6 video memory, but on the same GA104-300 processor as the regular model. The company ultimately abandoned plans to release the last two. Thus, on October 29, only one version of the GeForce RTX 3070 graphics accelerator with 8 GB of GDDR6 video memory will go on sale.

Nevertheless, a few days ago, information about another modification of the GeForce RTX 3070 began to appear on the Web. According to the latest data, it will use the PG132 SKU 35 board and will be built on the basis of the GPU GA102-150 with 7424 CUDA cores (the regular version has 5888 ). According to the informant, this version of the GeForce RTX 3070 can get 10 GB of GDDR6X video memory instead of 8 GB of GDDR6 memory. Recall that the older GeForce RTX 3080 is equipped with the same memory of the same volume.

We can assume that in this way NVIDIA implements defective GPU crystals that are not suitable for older video cards but are quite capable of forming the basis for less advanced solutions when some functional blocks are disabled.

The source reported not only the model of the board, the amount of memory, but also the level of power consumption of the card based on the GA102-150 processor. It will supposedly be the same 320 watts that are announced for the older GeForce RTX 3080 model.

The GA102-150-based GeForce RTX 3070 is supposed to be a response to AMD’s Radeon RX 6800 graphics card, which another network source @KittyYYuko points out will use a minor Navi 21 GPU modification – Navi 21 XL.

Recall that the release of the  GeForce RTX 3070 will take place on October 29. A day earlier, AMD’s new Radeon RX 6000 series video cards will be officially presented.