If this morning we were talking about the prices of graphics cards and their drop, now it’s time to talk about what is to come and of which we really know little beyond rumors and conjectures. Now a new leak specifies the models that AMD and NVIDIA will launch on the initial market and, curiously, the bet is much higher than in other series, where there are several curiosities since there will be two lithographic processes depending on model.
There is no doubt, or at least according to the leaker, that TSMC has acquired the services of both leading companies in the gaming GPU industry. If everything goes as it should, the Taiwanese will have the responsibility of making NVIDIA and AMD compete at the top and middle of the table with their models, but what about the low range?
Up to 8 models of RTX 40 and RX 7000 GPUs
Here are my guesses on the chips and models for the next gen.
AD102 (RTX 4080-90 series) 5nm
AD103 (RTX 4070series) 5nm
AD104 (RTX 4070series) 5nm
AD106 (RTX 4060series) 5nm
N31 (RX 7900series) 5nm
N32 (RX 7800series) 5nm
N33 (RX 7700series) 6nm
N22 (RX 7600series) 6nm
— Greymon55 (@greymon55) February 13, 2022
Why are NVIDIA and AMD going to put so much meat on the grill? There are two key reasons:
- The first is undoubtedly pure competition and gaining ground (or not losing it, depending on how you look at it) in each market stratum. Normally between two and three models have been presented for each series and architecture, but this time they go further with 4 each.
- intel squeeze, ARC Alchemist It is not a threat in terms of performance, but it is in terms of price, so an attempt will be made to give a blow to the mid-range to the high-end of the blue team, making them have to lower the price of their “recently” released architecture so much that they will not be as beneficial as it should be.
To all this we must add the factor of competitiveness of orders within TSMC. The agreements are signed for an initial production and from there they will not move until 3 months have passed from that date, which will obviously be different for each one. Once this happens, the volume factor plays an important role that each of the two will adjust for TSMC and this will influence prices to a lesser extent.
4 different chips for each one, two lithographic processes
As can be seen, NVIDIA will bet on those discussed in previous articles. AD102, AD103, AD104 and AD106, which will give life to the RTX 4080, RTX 4070 Ti? RTX4070 and RTX4060.
Logically, there is controversy here because there are two different chips for the same GPU model, which in the words of the leaker himself seems to be that the AD103 will be reserved for the RTX 4070 Ti if NVIDIA sees it appropriate for AMD’s competition, which is expected .
And it is that numerous rumors indicate that AMD would be squeezing NVIDIA to the maximum with its bets. These are based on four chips for 2022: Navi 31 for the RX 7900, Navi 32 for the RX 7800, Navi 33 for the RX 7700 and lastly and surprisingly Navi 22 for the RX 7600.
Furthermore, Navi 31 and Navi 32 will be manufactured in 5nm while Navi 33 and Navi 22 will be manufactured in 6nm. The former are not surprising because they try to compete in density, frequency and performance per watt, but the bets on 6nm are surprising.
First of all, N6 is a mixed EUV lithographic process, that is, it will mix DUV layers with this new technology, which gives a quite interesting result of the +18% logic density which will also allow not to change the patterns and blank masks if we come from the N7 or N7+.
For Navi 33 this is not a problem because it will be a new design from scratch, but Navi 22 is currently manufactured at 7nm for the RX6800M and RX6700 in different variants, so what is expected for the mid-range of the RX 7000 is precisely a chip with a performance similar to a current desktop RX 6800, which is not bad at all.