NVIDIA RTX 4070 Ti is finally official
Today NVIDIA introduced its third and thus far the cheapest RTX 40 desktop GPU.
The RTX 4070 Ti is a high-end GPU to offer up to 40 TFLOPs of compute performance. This card is based on AD104 GPU with 7680 CUDA cores, which is a full configuration of this graphics processor. The card is equipped with 12GB of GDDR6X memory clocked at 21 Gbps. Since the GPU is attached to a 192-bit memory bus, the maximum bandwidth will be limited to 504 GB/s, which is actually 104 GB/s slower than its predecessor, the RTX 3070 Ti.
NVIDIA 4070 Ti is to launch as a partner card only, which means no Founder Edition, although the render above may suggest otherwise. Almost every RTX 4070 Ti card that has been leaked or unveiled so far has three fans and 2 to 3.5-slot cooling design. This should be more than enough for the 285W TDP, though.
This is a third desktop card based on Ada Lovelace architecture and third Ada GPU to be announced, following AD102 (RTX 4090) and AD103 (RTX 4080). They are all based on the same process TSMC 4N custom process node. The AD104 GPU will have around 10B transistors fewer than AD103 (35.8B in total), with the die size around 295 mm².
Reviewers are to publish their test results as early as tomorrow, while the card is to become available for sale on January 5th. The MSRP of $799 will apply to the cheapest models at launch, mainly the cards that do not offer factory overclocking.
|NVIDIA GeForce RTX 40 Series Specs|
|VideoCardz.com||GeForce RTX 4090||GeForce RTX 4080||GeForce RTX 4070 Ti|
|Architecture||Ada (TSMC 4N)||Ada (TSMC 4N)||Ada (TSMC 4N)|
|GPU Die Size|
|Release date||October 12th, 2022||November 16h, 2022||January 5th, 2023|