GeForce RTX 4000 “Ada”: Other key data, TDP and weekly dates


Rumours about AMD's and Nvidia's next-gen graphics cards based on RDNA 3 and Ada are currently booming. In view of the sheer mass of “leaks”, which are incomplete, often contradictory and often only short-lived, the question arises: How credible are they anyway?

At the start of the week, it is once again the leaker Kopite7kimi, who is active on Twitter, and who recently caused the “certainties” he had spread to shake – both in terms of the deadlines and in terms of the key data of the first graphics cards of the GeForce RTX 4000 series.

RTX 4000 already “mid-July”

It all started on Sunday, when Kopite7kimi dug up a tweet from August 2021, in which he had talked about Ada Lovelace appearing “early” at the time, and now in May 2022 adds: “Q3 early”. When asked what exactly he meant by that, he said on Monday: “Mid-July”.

In view of the latest new products, that would actually be earlier: Nvidia presented the GeForce RTX 3000 to the public on September 1, 2020, GeForce RTX 2000 even on August 20, 2018 (as part of Gamescom). In both cases, however, the embargo on tests fell in mid-September – Turing had allowed a good month between the presentation and the market launch.

Series architecture presentation 1st test RTX 4000 Ada “early in the 3rd quarter” – RTX 3000 Ampere September 1, 2020 (own event) September 16, 2020 RTX 2000 Turing August 20, 2018 (for Gamescom) September 19, 2018 GTX 1000 Pascal May 7, 2016 (own event) May 17, 2016

But that's not all. On Monday, Kopite7kimi presumably followed up with new key data for the first Ada graphics cards – once again.

As a result, the GeForce RTX 4090, as the largest expansion stage for the time being, “only” comes with 126 active streaming multiprocessors out of the rumored 144 available in the AD102 chip and obviously therefore with only 450 instead of the last rumored 600 watts. He was “disappointed with RDNA 3”, according to the leaker, and apparently wanted to imply that Nvidia will obviously not bring out the 600-watt crowbar at the start because it is not necessary against AMD's new flagship.

In fact, the rumor mill recently spoke of 20 percent fewer execution units for AMD's RNDA-3 graphics cards – that was also a turnaround after almost a year of reports to the contrary.

How credible are rumors currently?

Should Nvidia actually adjust the specifications of the first Ada graphics cards in this form again, it could indeed be be reason for it. But is that really the case?

“Leaks” to the GeForce RTX 4000 based on Ada were recently characterized by three properties: They were plentiful, they never gave a complete picture and often changed: If the rumors of the last few weeks were correct, Nvidia would not have been only exchanged the GPU countless times in the respective classes, but also tweaked their configuration and TDP again and again. It is doubtful that things are so chaotic behind the scenes.

Of course, both opponents stalk each other as far as possible and react to each other, but with a probability bordering on certainty not like this and certainly not if the starting signal should be given in two months – quite apart from the fact that RDNA 3 was on Monday from more than months pursued status of the “Ada Lovelace killer” would have changed to a clearly inferior opponent.

This cannot be ruled out either. But this realization should not have matured behind the scenes in such a short time – many of the leaks, possibly also the completely new ones, were simply wrong.