You’ll be waiting a while for RTX 4070 and RTX 4060 Ti if these rumoured specs are accurate

$699, $599, $499 or $399? It's anyone's guess.

GeForce RTX 4080 looming large

The dust has barely settled on RTX 4070 Ti, yet the rumour mill never stops spinning. Looking ahead at the next crop of RTX 40 Series GPUs, word on the grapevine is that RTX 4070 and RTX 4060 Ti will be entering production in the coming months, ahead of launch in the second half of 2023.

Bad news for those longing for more affordable mid-range parts, yet a prolonged rollout shouldn’t come as too much of a surprise. Nvidia clearly wants RTX 40 Series positioned as a premium product for the time being, as there are plenty of RTX 30 Series GPUs still in the channel to cater for other segments.

Club386 has previously suggested current AD104 GPU would be repurposed for other products and sources at Wccftech appear to back up that assumption. The latest gossip is GPUs that didn’t make the grade for full-fat RTX 4070 Ti will be repositioned further down the stack. Rumoured specifications at this early stage should be taken with a pinch of salt, yet it’s always fun to speculate, so here’s how the mid-range duo could stack up.

GeForceRTX 4090RTX 4080RTX 4070 TIRTX 4070*RTX 4060 Ti*
Launch dateOct 2022Nov 2022Jan 2023Q3 2023Q3 2023
CodenameAD102AD103AD104AD104AD104
ArchitectureAda LovelaceAda LovelaceAda LovelaceAda LovelaceAda Lovelace
Process (nm)44444
Transistors (bn)76.345.935.835.835.8
Die size (mm2)608.5378.6294.5628.4628.4
SMs128 of 14476 of 8060 of 6046 of 6034 of 60
CUDA cores16,3849,7287,6805,8884,352
Boost clock (MHz)2,5202,5052,610TBCTBC
Peak FP32 TFLOPS82.648.740.1TBCTBC
RT cores12876604634
Tensor cores512304240184136
Memory size (GB)24161212TBC
Memory typeGDDR6XGDDR6XGDDR6XGDDR6XGDDR6X
Memory bus (bits)384256192192TBC
Memory clock (Gbps)2122.42121TBC
Bandwidth (GB/s)1,008717504504TBC
Power (watts)450320285200200
Launch MSRP ($)1,5991,199799TBCTBC
*unconfirmed GPU and specifications

If Nvidia is indeed using AD104 as the building block, potential specifications aren’t all that difficult to guesstimate. Cleaving 14 available SMs would return 5,888 cores for RTX 4070, which is believed to be retaining a full 12GB of memory and 192-bit bus.

RTX 4060 Ti, meanwhile, is cut more savagely, with just the 34 active SMs and 4,352 cores. Shipping frequencies could well be heightened to compensate, and it remains to be seen how much memory will be attached to the first x60 Series part. GPU power is said to be reduced to 200W, though given Ada Lovelace’s impressive efficiency thus far, we suspect this may change and could drop even lower prior to launch.

Computex 2023 would be an opportune moment for Nvidia and its partners to provide concrete details, and all eyes will be on the bottom row; MSRP. Going by what has come before, $599 for RTX 4070 and $499 for RTX 4060 Ti seems a logical continuation. As a point of reference, remember RTX 3070 and RTX 3060 Ti launched a couple of years ago at $499 and $399, respectively. Either way, expect to be hearing plenty more in the coming months.