NVIDIA RTX 4070: (almost) RTX 3080 performance with better power efficiency
Today NVIDIA released its fourth RTX 40 GPU to the market.
NVIDIA Ada Lovelace desktop family has grown and welcomes a new member. The RTX 4070 is a new card based on a cut-down AD104 GPU. There are fewer CUDA cores on this card compared to RTX 4070 Ti, but the memory subsystem has remained the same. More importantly, this card will retail at $599, which is $200 less than the Ti variant.
The GPU maker brought back its Founders Edition design for this launch. This means that NVIDIA will compete with board partners by offering their model at MSRP. At least this time, the company worked closely with all AIBs to launch as many “MSRP” cards as possible. For this reason, yesterday’s review launch was exclusively dedicated to cheaper cards.
Today NVIDIA lifts the embargo on premium RTX 4070 models. At the same time, the RTX 4070 officially hits the shelves today, so if you were planning to get one of those higher-end cards, there won’t be much time to read the reviews.
NVIDIA claims that their RTX 4070 GPU is running up to 1.7x faster than RTX 3070 TI thanks to DLSS3. This upscaling technology is exclusively supported by RTX 40 “Ada” GPUs. Without “Frame Generation”, which is the backbone of DLSS3, the GPU should be 20% faster. At the same time, NVIDIA claims that RTX 4070 should be just as fast as RTX 3080 in gaming, which was more or less confirmed by the press reviews yesterday.
The RTX 4070 is equipped with an AD104-250 GPU with 5888 CUDA Cores (23% of GPU is disabled). This $599 card features 12 GB of GDDR6X memory clocked at 21 Gbps. Since this SKU has a 192-bit memory bus (not 256-bit like RTX 3070), the actual bandwidth will be limited to 504 GB/s. This is still higher bandwidth though because RTX 3070 only offered 14 Gbps memory.
You can now order your RTX 4070 GPU from popular retailers:
|NVIDIA GeForce RTX 4070 vs. 3070 Series|
|VideoCardz.com||RTX 4070 Ti||RTX 4070||RTX 3070 Ti||RTX 3070|
|Architecture||Ada (TSMC 4N)||Ada (TSMC 4N)||Ampere (SAMSUNG 8N)||Ampere (SAMSUNG 8N)|
|Max FP32 Compute|
12 GB G6X
12 GB G6X
8 GB G6X
8 GB G6
|Interface||PCIe Gen4 x16||PCIe Gen4 x16||PCIe Gen4 x16||PCIe Gen4 x16|
|Release Date||January 5, 2023||April 13, 2023||June 10, 2021||May 8, 2021|