AMD will hold its “Together we advance gaming” conference on its YouTube channel on November 3 at 10 p.m. The company led by Lisa Su should present its new series of consumer graphics cards there, the Radeon RX 7000. According to various rumors from usually reliable sources, this series will be the first to offer a chiplet approach. Thus, as with its Ryzen processors, AMD could offer a graphics chip with several dies: one containing the GPU, the controller and all the management blocks and others containing the graphics memory and the cache memory (Infinity Cache).
An interesting approach which will not be the only novelty. AMD intends to drastically increase its graphics performance. This will require the integration of a high number of calculation units with 10,752 Stream Processors on the Radeon RX 7900 XT and even 12,288 SP on the Radeon RX 7950 XT. More recent rumors also indicate the probability of a change in nomenclature with a return of the XTX name – abandoned since the Radeon X1900 XTX of 2006 -, at least for the most high-end model which would then take the name of Radeon RX 7950 XTX. To fit so many computing units into a GPU, AMD will use a 5 nm engraving process, just like Nvidia does on its GeForce RTX 40.
On the graphics memory side, the two high-end models would integrate 20 to 24 GB of GDDR6 type graphics memory (320 or 384-bit bus). Everything would operate on an energy envelope of 330 and 420 watts respectively. Like Nvidia, AMD would therefore have been tempted to let go of power consumption. On the other hand, the new Radeons should not integrate a PCIe 5 power connector of the 12VHPWR type. This will no doubt give the AMD marketing team a few cartridges to talk about the few setbacks encountered by Nvidia on its GeForce RTX 4090.
Appointment is made on November 3 at 10 p.m. on Digital to debrief these announcements.