NVIDIA’s next-gen RTX 40 series “Ada Lovelace” graphics are slated to launch by the end of the year. Based on TSMC’s N5 (5nm) process, the GeForce RTX 4080 and 4090 are expected to feature a core count of up to 18,432 across 144 SMs and 12 GPCs. This indicates that the AD102 die will likely be a derivative of the GH100 die, with the FP64 cores trimmed, and additional L2 cache and RT cores.
|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?|
As per a recent leak, NVIDIA’s next-gen GeForce RTX 4080/4090 GPUs are going to fall short of rival AMD’s Radeon RX 7800 XT/7900 XT. According to the source, the best NVIDIA can currently do with Ada is 60-80% faster than existing Ampere parts. While this is nothing to scoff at, it’s still a notch lower than what AMD’s chiplet based Radeon RX 7800 XT and RX 7900 XT will be capable
In addition to this, these GPUs will feature GDDR7-class memory (either GDDR7 or super-faster GDDR6/6X) with up to a 384-bit memory bus, resulting in an external bandwidth of over 1,200 GB/s. The top-end RTX 4080 and RTX 4090 GPUs are going to have a heft TBP of up to 600W on factory overclocked models, and slightly lower on Founders Edition cards.
As per the latest rumors, though, the GeForce RTX 4080/4090 launch may be delayed to the holiday season. At the moment, all sources point to an August/September release, and a delay may be the result of the recent NVIDIA hack or the excessively high power consumption.