Nvidia’s GPU Technology Conference will take place from April 12th to 16th, GTC for short, instead of. In keeping with standing, company co-founder and boss Jensen Huang kicked off with a keynote last night. The most important new technologies were presented.
As usual at the GTC, the focus was not on player graphics cards, but above all on dedicated chips for AI calculations and this time even the first own central processing unit, CPU, based on ARM.
However, we players do not go completely empty. A roadmap shows at least when we can expect new generations of graphics cards:
RTX 4000 and RTX 5000
How the graphics cards are ultimately called is actually speculation at this point in time. However, it is very likely that Nvidia will remain true to its previous naming scheme, which is why we – as long as we don’t know better – assume the RTX 4000 and RTX 5000.
RTX 5000 not “Hopper” after all?
Only the terms “Ampere Next” and “Ampere Next Next” emerge from the roadmap. Until now it was generally assumed that the code names for this were »Lovelace« (from Ada Lovelace) and “Hopper” (from Grace Hopper) would be. However, since Nvidia’s first generation of CPUs goes by the name of “Grace”, it seems likely that the name of the famous computer scientist and computer pioneer will not also be the godfather of a generation of GPUs.
We are therefore currently assuming that the next generation of graphics cards for gamers will be codenamed Lovelace and that one name will be known for the following generation.
RTX 4000 should again be a significant increase
There are already some rough key data about Lovelace and RTX 4000, which, however, can still be assigned to the area of rumors. The full expansion, which may then be called RTX 4090 in the form of a graphics card, is to be equipped with 18,432 computing cores, manufactured in a 5-nanometer process and thus clearly surpassing the current RTX-3000 cards.
The current top model RTX 3090 it has 10,496 arithmetic units and is manufactured at Samsung using an 8 nanometer process.
more on the subject
Lovelace is said to double RTX performance
RTX 5000 supposedly in chiplet design
For the cards, which are presumably named RTX 5000, there is only the assumption that they are supposed to be the first GPU generation from Nvidia, which is manufactured in the so-called MCM design. MCM stands for Multi-Chip-Module and describes a chip that is composed of several smaller chiplets – in this context it is also referred to as a chiplet design.
The chiplet design offers advantages due to the smaller chips that can be manufactured in production, at the same time GPUs can be made larger overall, keyword: scalability. However, there are still technical hurdles to overcome. For example, the memory management and the connection between the individual chips.
With desktop processors you are already further. AMD already uses such a process for its Ryzen processors and has recently celebrated great success with it.
The real reasons for the hardware shortage
Hardware shortage: In the video embedded above, we will try to clarify whether something will have changed in the currently difficult delivery situation by the possible release of the next generation of graphics cards from Nvidia.