The presentation of the Ampere architecture allowed us to see all the keys of the first GeForce RTX 30 graphics cards, a generation that marks an important change compared to Turing, both at the level of silicon and gross power, since it doubles the number of shaders per each SM unit and introduces new generation tensor cores and RT cores.
After analyzing the RTX 3080 I was clear that NVIDIA has made a huge leap in terms of performance, although it is true that this is noticeable, especially when we move in 4K resolution, and also when we activate settings that push the GPU to the limit, like ray tracing for example. The biggest differences are achieved precisely with the most demanding quality settings, something that makes me think that there is a clear bottleneck at the CPU level due to the poor level of optimization that current games have, in this sense.
There is no doubt, the new GeForce RTX 30 were promising, but unfortunately NVIDIA has not been able to offer a good level of supply, and in the end we have found a very limited launch that has been aggravated by the action of the speculators, who have managed to get almost the entire stock of the new GeForce RTX 30.
NVIDIA promised that it would do everything possible to improve the situation, but the truth is that today the stock is still just as low, so much that it is practically impossible to buy an RTX 3080 or an RTX 3090 . We’ll see what happens to the GeForce RTX 3070, but early reports suggest that the stock will also be very limited.
Why is there such low availability of the GeForce RTX 30?
We don’t have an official answer from NVIDIA, but numerous sources, including the prolific DigiTimes, have helped us understand the current situation a bit better.
The company directed by Jen-Hsun Huang had initially planned to bet on the TSMC manufacturing process in 7 nm with its GeForce RTX 30. However, AMD had contracted a significant part of the production in this process, and therefore NVIDIA had no choice but to turn to Samsung’s 8nm process. This explains why the NVIDIA A100 uses the 7nm process, and why the RTX 30 series uses GPUs made in the 8nm process.
Those in green had no choice but to split production between Samsung and TSMC in order to arrive on time and launch before the end of 2021. I am convinced that AMD’s plans to launch an “RTX 2080 Ti killer” in the fourth quarter of 2020 have been key for NVIDIA to decide to accelerate the debut of its GeForce RTX 30.
It is easy to understand, if NVIDIA had waited, AMD would have put on the market the Radeon RX 6900 XT, a more powerful and cheaper graphics card than the RTX 2080 Ti. Ahead of the launch of the GeForce RTX 30, those in green have been able to send a devastating message to the user, and to AMD itself: “The RTX 3070 will allow you to enjoy performance at the level of the RTX 2080 Ti for only 519 euros.”
All this helps us understand part of the problem that has led to the shortage of stock of the GeForce RTX 30 series, but to complete the puzzle we must add a piece: NVIDIA said there would be stock problems until 2021, something that is very interesting and that finally makes all its sense, especially now that we know that those in green plan to jump completely into the 7nm process of TSMC by 2021.
Those of you who have known how to read between the lines will already be clear about what is happening, for those who do not, I leave you a summary: it is most likely that NVIDIA has contracted a limited production of the GeForce RTX 30 under the Samsung 8 nm process so as not to to leave AMD alone with its RX 6000, but its goal would have been, from the beginning, to carry out mass production of those under TSMC’s 7nm process, something that it could not achieve before 2021.