Please note that this post is tagged as a rumor.
NVIDIA RTX 4090 Ti no longer planned, according to a reputable leaker
Kopite7kimi has been relatively quiet about the latest rumors on the GeForce RTX 40 series, mainly due to the fact that most of the RTX 40 lineup has already been released. However, as we approach the one-year anniversary of the RTX 4090 announcement in September, the question arises whether NVIDIA has any plans for a mid-series refresh.
Although NVIDIA has already introduced two Ti models (4070/4060), there’s no indication of an RTX 4090Ti or RTX 4080Ti, which would be the natural choices for a refresh aka spec upgrade. Given that the RTX 4090 non-Ti doesn’t utilize the full AD102 silicon, it would be reasonable to expect a higher-end model featuring more cores and potentially faster GDDR6X or even GDDR7 memory. However, it appears that such plans are not on the table, at least not for the flagship model.
I'm afraid there won't be RTX 4090 Ti anymore. Some low-grade AD103 and AD106 chips will be another versions of RTX 4070 and 4060.
— kopite7kimi (@kopite7kimi) July 27, 2023
Kopite7kimi, known for being a highly reliable source on upcoming NVIDIA hardware, has stated that there are no plans for an RTX 4090 Ti. Additionally, it’s mentioned that NVIDIA will introduce new AD103/AD106 models for the RTX 4070/4060, but these models aren’t entirely new SKUs; rather, they are likely intended to optimize chip sales, and they may primarily be targeted at the Chinese market.
Currently, NVIDIA doesn’t have a pressing need for the RTX 4090 Ti because the RTX 4090 stands as the most powerful GPU available, and there’s limited competition from AMD in this high-performance segment. While there have been rumors about a potential refresh or the addition of 3D V-Cache to the RDNA3 series from AMD, these speculations haven’t posed any real concern for NVIDIA. Until a significant shift occurs in the market, it’s unlikely that an RTX 4090 Ti will be introduced.
GeForce RTX 5090 with 512-bit memory bus?
Combined with multiple sources, I confirm the gaming flagship of Ada-next will have a 512-bit memory interface.
— kopite7kimi (@kopite7kimi) July 27, 2023
Kopite added that a potential flagship GPU from the next generation should utilize a 512-bit memory bus. This would be a major shift in NVIDIA gaming lineup, considering such configuration has not been used for a long time, like GTX 280 or some dual-GPU models. If NVIDIA were to use GDDR7 memory with the max speed of 32 Gbps on such a potential next-gen GPU, this would result in 2TB/s bandwidth, so double what RTX 4090 offers.
However, NVIDIA is unlikely to use the fastest G7 memory at launch, and just the fact that the GPU may support a 512-bit bus does not guarantee that such configuration will be enabled for each variant. Certainly, there are more questions than answers at this point.
NVIDIA isn’t expected to unveil its next-generation consumer architecture until 2025. Given the company’s typical two-year cadence for gaming architecture releases, there might be a possibility of a mid-generation update in some form to bridge the gap. However, specific details remain uncertain at this time.
Source: kopite7kimi