NVIDIA’s fourth Ada Lovelace graphics card is now accessible in retail. That includes 5,888 FP32 cores and a bandwidth of 504 GB/s, it appears nearly an identical to its predecessor. The RTX 3070 packs 5,888 cores however pairs it with a wider 256-bit bus however slower GDDR6 reminiscence for an exterior bandwidth of 448GB/s. On paper, the bigger L2 cache and the upper reminiscence bandwidth on the RTX 4070 are the one variations between the 2 70-class graphics playing cards.
The above desk from Quasarzone highlights how potent the Ada Lovelace microarchitecture is. Regardless of unchanged core counts, NVIDIA extracted as much as 40% extra efficiency from the RTX 4070 with only a node shrink (+greater clocks) and a bigger LLC.
In real-life situations, the RTX 4070 affords roughly RTX 3080 stage of efficiency, a GPU that has practically 50% extra cores (8,704) with a 50% greater bandwidth (760GB/s). These numbers go on to show the form of generational uplift that NVIDIA was in a position to convey with out altering the shader energy.
