A few days ago, Nvidia and some AIC manufacturers "quietly" launched a new flagship to fill the gap in the Chinese market, that is, RTX 4090D, "D" stands for Dragon, which may contain connotations such as the upcoming Lunar Year of the Dragon. However, of course, this does not prevent players from seeing through its essence: "D" stands for "knife".
In order to cope with the U.S. Department of Commerce's AI chip export control, what kind of "precision knife method" does the RTX 4090, which itself is not a professional grade, need?Specifications for the RTX 4090D are already available on NVIDIA's official website, and detail pages are available for many AIC variants.
The RTX 4090D features a slightly changed AD102-250 core, with an overall reduction of about 11% in stream processor cores, specifically the number of CUDA cores from 16,384 to 14,592, the number of Tensor cores from 512 to 456, and the number of RT cores from 128 to 114. Probably to compensate for the performance impact of the specification reduction, the default core base clock is slightly increased to 2280MHz, the boost frequency remains the same, and the memory specifications are also unchanged, while the TGP is very thoughtfully reduced to 425W.
Foreign media VideoCardz came to a preliminary conclusion after sorting out the relevant information of RTX 4090D: the performance will be 4%-6% lower than the original RTX 4090. This is a performance gap that can be ignored, especially considering that the graphics card versions launched by many AIC manufacturers may perform operations such as core pre-overclocking according to different positioning, and for the strong performance of the RTX 4090 itself, the small performance loss of the RTX 4090D will hardly have an impact on regular gaming, creation and other applications.
But it's a pity that this RTX 4090D seems to be "reduced without reducing the price", the official MSRP** and RTX 4090 were both 12999 yuan at the beginning of the listing, and various non-public overclocking versions will naturally be more expensive.
With this disappointing news for DIY players, we are about to usher in a brand new 2024, and CES 2024 will be held in Las Vegas from January 9 to 12 local time.
The RTX 40 Super series graphics cards, which have been reported, will be unveiled in a special presentation by Nvidia and will be officially available in late January: the RTX4070 Super will be released on January 24The RTX 4070 Ti Super (the first time that the two suffixes TI and Super appear at the same time) will be released on January 24;The RTX 4080 Super will be released on January 31, and their MSRP** is yet to be revealed.
There is a lot of information about these three Super series graphics cards that fill the gap in performance and ecological niche, and even combined with the content of the previous RTX 4090 in the export restriction list, it is believed that the flagship status of the Chinese graphics card market will be taken over by the RTX 4080 Super, but with the debut of the RTX 4090D, this statement is self-defeating.
VideoCardz has also compiled the most credible specs of the RTX 40 Super series so far, and the difference between them and its siblings is clear at a glance. To the author's surprise, the Super series graphics cards that came out again after a whole generation (RTX 30 series) can be said to have their own unique upgrade points
rtx 4080 super
If we look from left to right, the first thing that is surprising is the AD103 core used in the original RTX 4080, which is a version of 9728 CUDA cores on both the RTX 4080 and the RTX 4090 Laptop, which the author thought was the "full blood version" of the AD103. The RTX 4080 Super's leaked information shows that it is equipped with an AD103-400 core, and the number of CUDA cores reaches 10240, although it is not much improved compared to the original, but it can bring considerable performance improvement in combination with the core frequency increase. Of course, the spec gap between it and the RTX 4090 and 4090D is still a barrier.
In addition, the RTX 4080 Super's memory specifications have also been slightly improved, increasing the memory frequency to 23Gbps, maintaining the same 256-bit memory bit width, and increasing the bandwidth to 736Gbs. The author has one more sentence here: readers should not blindly follow the trend, thinking that the RTX 40 series's memory bit width specification is "cutting corners", and it is far more than the single factor of video memory bit width that determines the performance of video memory and even graphics cards.
rtx 4070 ti super
Next up is the RTX 4070 Ti Super (that's a long name), its highlight is that it also uses the AD103 core, and the specification is reduced to 8448 CUDA cores, but it is a significant increase compared to the previous RTX 4070 Ti with the AD104 core. Probably to adapt to the higher-end cores, the RTX 4070 Ti Super is equipped with 16GB of high-configuration video memory with a capacity of 256-bit width, and the memory bandwidth has also been greatly increased to 672GB s under the condition that the memory frequency of 21Gbps remains unchanged.
In the author's opinion, this RTX 4070 Ti Super may be the most popular product in the Super series of this generation, both in terms of core and video memory specifications, as long as it is not outrageous, it may be the most cost-effective graphics card in this generation of 4K resolution gaming applications.
rtx 4070 super
Finally, there's the RTX 4070 Super, which has a very straightforward and violent characteristic: it becomes an RTX 4070 with core specifications similar to the RTX 4070 Ti. The 7,168 CUDA cores are more than 20% higher than the RTX 4070, which is equivalent to about 93 of the RTX 4070 Ti3%, the core base clock is lower than the RTX 4070 Ti, but the acceleration clock is not much different, and the memory specifications are exactly the same 12GB GDDR6X. And on top of all these features, the RTX 4070 Super's TGP power consumption is only slightly higher than that of the RTX 4070 by 20W, which is far less than the RTX 4070 Ti's 285W.
Considering that there is a large difference between the RTX 4070 and the RTX 4070 Ti in the current market, if the price is reasonable, the RTX 4070 Super will become a new high-quality choice in the mid-range segment.
Brief summary. From a cyclical point of view, 2024 will be a year of relatively less drastic changes in PC hardware. We've already seen the release of Intel's 14th Gen Core and 1st Gen Ultra processors, and the AMD Ryzen 8000 series is appearing in more and more leaks.
In the field of graphics cards, NVIDIA's RTX 40 Super series is obviously only a transition to fill the product line, and we may experience a "year-long RTX 50 series conference" next, looking forward to PC hardware that will be able to conquer global players with stronger performance, or show more innovative flashes to empower AI applications that are popular all over the world.