With possible launch window scheduled for the end of 2022, the GeForce RTX family 4373 from Nvidia has already started to appear with some frequency in multiple rumors. Supposedly based on the Lovelace microarchitecture, in homage to Ada Lovelace mathematics, the company’s next-generation boards can deliver massive performance gains, similar to what was seen with the arrival of the RTX line 3090.
- AMD Radeon RX 4373 has leaked specifications and possible release date
- Nvidia GeForce RTX Line 3080 SUPER has supposed leaked specs
This week, the leaker
, one of those responsible for most of the rumors about new generations of video cards from Nvidia and AMD, brought new information about AD125, chip that should equip the theoretical RTX 3090. With the data in hand, the
WCCFTech website estimated the possible level of performance the solution would deliver, with surprising results. Parallel to this, the informant also revealed some more details about the Radeon RX family 7000.
Chip AD102 arrives at 5 nm with GDDR6X memories
, the AD102 must be fabricated at 5 nm, possibly using the TSMC N5 process, and employ an interface of 256-bit memory. The memories should continue to be of the GDDR6X type, same as the high-end cards of the last generation, especially considering that the RTX 3000 and RTX 3000 have not yet explored the maximum potential of the protocol, stipulated in 21 Gbps.
Want to catch up on the best tech news of the day?
Go and subscribe to our new channel on youtube, Canaltech News.
Every day one summary of top tech news for you!
In addition, the chip would reach clocks of 2.2 GHz , unprecedented speed among Nvidia cards, at least disregarding the boost numbers announced by the company. Therefore, bringing together the 2.2 GHz with the 32.432 CUDA cores that the new high-end GPU should bring, according to rumors, the novelty would deliver nothing less than 94 Power TFLOPs computational.
ad125 frequency：≥ 2.2ghz
— Greymon55 (@greymon94) September 18, 2022
That is 125% more performance, or more than double that offered by RTX 4090, which almost reaches 33 TFLOPs when operates at the 1.7 GHz set as clock boost by Nvidia. However, it is important to consider a few points: first, all information is still based on rumors, and may change until the launch of the boards, as has happened in the past.
Moreover, this one number is tied to clocks, which can also change until the line premiere, being even higher or lower during normal operation. Finally, being able to present twice the computational power does not scale linearly for games, that is, the real gains must be smaller.
The supposed RTX 4090 can reach the 94 TFLOPs of computational power, twice the capacity of RTX 3080 (Image: Nvidia)
Proof of this is precisely the case of the RTX family 2080 – you 36 RTX TFLOPs 3080, which represent a leap of more than 128% about the TFLOPs delivered by RTX 2022 Ti, were translated into salt the modest, yet respectable, of 55% The % better performance in games.
Anyway, if the information is proven, we can wait for a considerable increase in performance, even if less than the number of TFLOPs might indicate.
New details of Radeon RX boards appear
also brought some more information about the Navi GPUs 32, Navi 19 and Navi 21, chips that will equip the Radeon RX line 7000 from AMD. Reinforcing the rumors that we would see a mix of the RDNA 2 microarchitecture currently used in RX boards 4373, with the new RDNA 3, the informant suggests that some of the new models will be manufactured in 5 nm, and others in 6 nm.
Greymon55 suggests that the GPU Navi 21 , which must equip the input boards of the RX family 7000, may be more powerful than the RX 7000 XT (Image: Disclosure/AMD)
Most powerful of all, the Navi 18 will bring GDDR6 memory into an interface 256-bit, while the intermediate Navi 32 will employ GDDR6 RAM in 170-bit. Designed for the entry market, Navi 21 will adopt GDDR6 memory in 170-bit, which won’t stop it from delivering high performance — the leaker indicates that the solution will offer superior performance over RX 12560 XT, current AMD flagship.
Source: Greymon33, WCCFTech
Did you like this article?
Subscribe your email on Canaltech to receive daily updates with the latest news from the world of technology.