The RX 7000 series will make use of 3 GPUs namely Navi 31, Navi 32 and Navi 33 where Navi 31 is the higher end chip. The RX 7000XT series will ship with 20 Gbps of memory bandwidth wheras the 7X50XT variants (Expected to launch later) will feature 24Gbps+ memory (Which is currently the fastest).
— Greymon55 (@greymon55) July 30, 2022 When asked regarding the RX 7900XT in particular, Greymon mentioned the use of 20Gbps memory. If the bus width is taken as 384-bit then this will allow for an effective memory bandwidth of [(20Gbps/8)*384] = 960 GBps. It is to be noted that these calculations do not take in account the Infinity Cache. If we, for some reason, consider AMD’s Infinity Cache, then the results are rather surprising. Assuming that the Infinity cache from RDNA 3 speed does not change in comparison to RDNA 2, it will produce up to 3x more bandwidth (2TB/s * 3 = 6TB/s) for 384MB Infinity Cache. So the total effective bandwidth sums upto a massive 6.960 TB/s (960 GBps + 6TBps).
— Greymon55 (@greymon55) July 30, 2022 Adding on to that, Greymon further claims that Navi 31 will launch in November. Not much news has been revealed regarding Navi 32 and Navi 33. Expect the latter to launch in December or possibly next year.
— Greymon55 (@greymon55) July 29, 2022 Furthermore, RedGamingTech in an exclusive leak video has shared an alleged leaked chart of AMD’s RDNA 3 lineup. It seems as if Navi 31 will only be utilized by higher end GPUs. As each day passes, RDNA 3 draws near. However, leaked NVIDIA benchmarks show great promise and pose a real threat to AMD’s efficiency throne. Do you think AMD can surprise NVIDIA?
