GeForce RTX 4070 Specs, Pricing And Power Draw Seemingly Confirmed
A leaked NVIDIA slide has apparently confirmed the general specifications and price of NVIDIA’s upcoming GeForce RTX 4070. The slide appears to support rumors, confirming that the GPU will come with a 200W TDP, 12GB of GDDR6X memory, and a price of $599. Best of all, the slide also shows average gaming power consumption of less than 200W TDP, with a rating of just 186W.
The rest of the GPU specs listed on the slide include a performance target of 29 TFLOPs, the use of NVIDIA's AD104 GPU die, 36MB of L2 cache, and 504GB/s of memory bandwidth. However, the slide does not give us the rest of the RTX 4070’s specs, like core count or clock speeds. But based on previous rumors, the rumored 5888 CUDA core count and 2.475GHz boost frequency are probably true.
Either way, the RTX 4070 is shaping up to be the best RTX 40 series GPU to date in terms of value and efficiency. With an average gaming wattage of under 200W, the GPU is vastly more efficient than its RTX 30 series counterparts and sucks down significantly less juice compared to its bigger 4070 Ti, 4080, and 4090 brethren. Just so you know how good 200W is for this class of GPU, the RTX 4070 basically has the same power consumption as an RTX 2070 from 2018 with the same raw compute performance as an RTX 3080.
This should make the RTX 4070 a great fit for customers who want to buy into this new $600 “mid-range” GPU price class. The low power consumption means AIB partner cards should be exceptionally small and fit into most computer chassis on the market with plenty of room to spare. Plus, the low power rating should also put less demand on the power supply and reduce the likelihood of PSU upgrades.
Interestingly, the leaked slide directly compares the RTX 4070 to the RTX 3070 Ti and RTX 3070. This suggests the GPU is considered an upgrade to both RTX 30 series counterparts, rather than targeting the RTX 3070 specifically. This would make sense, considering the 4070’s 29TFLOPs of compute performance is significantly faster than both RTX 3070 Ti’s 21.75 TFLOPs, and the RTX 3070’s 20.3 TFLOPs. Not to mention both the 3070 Ti and 3070 are incredibly similar to each other in real-world performance, with the exception of memory bandwidth.
The RTX 3070 Ti has been widely considered the “misfit” of the RTX 30 series family. When it first debuted in 2021, it was heavily criticized for not having a big enough performance gap to the RTX 3070 to justify its $100 price hike. We reflected similar thoughts in our review, noting how the 3070 Ti was a lot closer to RTX 3070 performance compared to the RTX 3080.
Thankfully the RTX 4070 replacement should be substantially quicker, and substantially more power efficient based on the leaked and rumored specifications. Even though the GPU has the same CUDA core count as the RTX 3070 with just 5888 of them, the RTX 4070 makes up for that tremendously by having vastly superior clock speeds, a colossal upgrade in L2 cache capacity — from 4MB to 36MB — and further improvements which have been made on the newer Ada Lovelace GPU microarchitecture.
The RTX 4070 is unofficially expected to debut in under two weeks.