NVIDIA is reportedly planning to delay the launch of its next-gen RTX 40 series graphics cards yet again. The GeForce RTX 4090 (and perhaps the RTX 4080) was originally set to land in August. Courtesy of falling prices, limited demand, and flooding of the channel inventory, we’re now looking at a November retail release for the higher-end Lovelace SKUs.
Furthermore, NVIDIA is also (allegedly) prepping its next BFGPU, aka the RTX 4090 Ti for an earlier-than-expected launch. While a 2022 roll-out is a big no, we might get an early to mid-2023 retail release with a possible CES 2023 sneak peek.
|GPU||GA102||AD102||RTX 4090||AD103||RTX 4080||RTX 4070 Ti (AD104)||RTX 4070|
|Arch||Ampere||Ada Lovelace||Ada Lovelace||Ada Lovelace|
|Process||Sam 8nm LPP||TSMC 5nm||TSMC 5nm||TSMC 5nm|
|TP||37.6||~100 TFLOPs?||83 TFLOPs||~50 TFLOPs||47 TFLOPs?||~35 TFLOPs||35 TFLOPs?|
|Memory||24GB GDDR6X||48GB GDDR6X||24GB GDDR6X||16GB GDDR6X||12GB GDDR6X|
|Launch||Sep 2020||Sept 22?||Sept 22?||Q1 2023?|
The GeForce RTX 4090 Ti will be based on the full-fat AD102 die. That means an incredible 18,432 FP32 core-GPU (144 SMs) paired with 48GB of 24Gbps GDDR6X memory across a 384-bit bus. The memory will be backed by 96MB of L2 cache, resulting in a single-precision compute performance of almost 100 TFLOPs. The RTX 4090 will be a cut-down variant with 16,384 cores or 128 SMs paired with 24GB of GDDR6X memory and 72MB of L2 cache. The RTX 4090 Ti may very well come with a record-breaking TGP of 800W (but I personally expect it to top out at 600W), primarily on account of high core clocks (close to 3GHz). The RTX 4090 will draw a more reasonable 450W under load.