2023-09-18 22:05:00
Well, since everyone is disappointed with the current RX 7000 and RTX 40 series of graphics cards, rumors are already talking about the next ones. In fact, there has been talk about them for a few months, such as those who claim that AMD is going to throw in the towel and will not even compete with the RTX 5090, taking into account that right now it does not even have an alternative to the RTX 4090. A rumor has emerged from that RTX 5090 that indicates how much it would improve compared to the 4090 in various areas.
Of course, the following must be taken with a grain of salt because not even NVIDIA has finalized the design of the chips, although it should not be far from it. Blackwell would be the name of the next generation, which is said to be graphic processing units with a design like multi-chip modules. At least according to the not always reliable equisero tweeter @kopite7kimi.
He says that they will not increase the number of internal structures of the current chip, but their internal design will significantly increase. I see this as normal because the Ada architecture has not changed anything compared to the Ampère, since it has only added more cache and more operating frequency. NVIDIA will have to do something in the field of architecture, although AMD is not putting pressure on it.
On the other hand, a random Internet user has given figures, probably invented, of how much the main units of the GPU used in the future RTX 5090 would increase. It would have 50% more shaders, reaching 24,576 compared to the 16,384 of the RTX 4090 would have 52% more bandwidth (1.5 TB/s), 78% more cache, 15% more frequency, and 70% more performance overall. The memory used would be GDDR7 32 Gb/s, but this is assumed by everyone so I myself could have come out and said what he is saying.
Via:
Videocardz, Videocardz (2).
#RTX #MCM #design #higher #performance #RTX