In the early morning of January 4th, the NVIDIA Ada RTX 40 series family ushered in the third member “RTX 4070 Ti”.
There are rumors that the card was originally planned to be named RTX 4070 Ti, and RTX 4070 was also considered, but in order to increase its value, it was forcibly renamed RTX 4080 12GB before release.
In this way, the same model has two video memory versions for the first time. The key is that the core specifications are different. This so-called RTX 4080 only has a mere 192-bit video memory interface width, and the result was unanimously criticized.
In the end, NVIDIA rarely withdrew the RTX 4080 12GB, and after a long time of deliberation, it changed back to the original name and reappeared in front of everyone.
Back to product specifications,RTX 4070 Ti is based on AD104 core and integrates 35.8 billion transistorsmore than the previous flagship core GA102, but294.5 square metersOn the contrary, the area is more than half smaller, the transistor density exceeds 120 million per square millimeter, and TSMC’s 4N process is quite powerful.
CUDA cores are enabled for all 7680which is 25% more than the previous generation RTX 3070 Ti, but not as good as the RTX 3080 12GB. The core frequency standard is 2310-2610MHz, which is also the highest frequency of the N card so far.
At the same time there are240 fourth-generation Tensor cores, 60 RT ray tracing cores, and 240 texture units are also 25% more than the previous generation. The secondary cache is expanded to a full 48MB, but the ROP raster unit is reduced from 96. To 80.
However, the RTX 4070 Ti is not really a full-blooded AD104 core, becauseSix NVENC codecs are only half enabledPresumably it is reserved for professional graphics cards of RTX A series workstations.
AD104 core
RTX 4070 Ti hides three NVENC
192-bit video memory interface width, this is the first time in the history of the x80 series, not even the x70 series.
Even with 12GB GDDR6X video memory with 21GHz ultra-high frequency, the bandwidth is only 504GB/s, which is a full 17% reduction compared to the previous generation, and only a little higher than RTX 3070 and RTX 3060 Ti.
It is no problem to run 2K high-definition games with such specifications, which is also the main resolution of the card, but it will obviously struggle at 4K resolution.
The power consumption of the whole card is TGP 285W, which is similar to the RTX 3070 Ti 290W, and continues to use the 12VHPWR 16-pin power supply interface, and usually comes with a dual 8-pin transfer power supply cable.
The standby power consumption is generally 12W, the video playback power consumption is 20W, and the average game power consumption is 226W, which are far lower than the previous generation.
In terms of performance, NVIDIA claims,Combined with the brand-new DLSS 3 technology, the performance of RTX 4070 Ti can be up to 3 times that of the previous generation RTX 3090 Ti, while the power consumption is almost half. Compared with the legendary GTX 1080 Ti, the performance improvement can reach An astonishing 12x.
RTX 4070 Ti has no FE public version this time,The non-public version of the AIC brand will be officially opened for sale at 22:00 on January 5th, and the price starts at 6499 yuanCompared with the price conscience of RTX 4080 12GB 7199 yuan.
