The RTX 4070 SKU recently received a major specification update in which the CUDA cores increase to 7680 and the memory increases to a capacity of 12 GB. Now there is also an update on the clock speeds. The memory speed has been increased to 21 Gbps, which indicates GDDR6X memory instead of GDDR6 memory we inspected earlier at 18 Gbps. With more memory and an increase in capacity, the bandwidth definitely increases as in this case – up to 504 GB/s.
The famous Kopite7kimi comes to the GPU clocks and has clarified the situation. The leaker has published full specs for the RTX 4090 flagship Ada GPU, making it easier to compare the two models in the series. The GPU clock turns out to be 70 to 90MHz higher than what the 450W model is capable of. The RTX 4070 can give a base clock of 2310 MHz with a boost at 2610 MHz. However, as we all know NVIDIA’s official boost clock is almost always lower than the real world, we are talking about a maximum boost of more than 2.8GHz.
If all this information has confused you, here it is in simpler terms. The RTX series just got a major clock speed update. Compared to this series, the RTX 3070 has 700 to 900 MHz lower clock speeds. Combining more cores with higher clock speeds here results in 40 TFLOPS of computing power, which is twice the computing power of the RTX 3070, or in other words equal to that of the RTX 3090 Ti.
The RTX 4070 is designed to deliver half the memory and half the performance of the RTX 4090n SKU.
Pretty much all the high-end cards of the RTX 40 series will be announced at the same time, but the RTX 4070 is expected to launch after the RTX 4090 and RTX 4080, meaning sometime in the middle of the fourth quarter of this year.
— kopite7kimi (@kopite7kimi) August 10, 2022