© AMD
For its consumer and professional processors, as well as for its graphics chips, AMD is playing the “everything in 5 nm” card. A technology that could offer it a temporary advantage in processors against an Intel that is content with a 7 nm equivalent. And in the field of GPUs, AMD and Nvidia will be, in this respect, perfectly equal.
While accounting figures are not the cup of tea for ordinary mortals, the financial points of large tech companies like AMD allow you to take advantage of firm commitments on the roadmaps. Indeed, marketing promises to the general public are not as binding as those made to financiers. AMD closed yet another excellent fiscal quarter: from processors for the data center, through consumer chips, gaming or integrated chips, AMD is rocking the house. And make commitments.
The clearest promise is that of the launch of Ryzen 7000 processors “by the end of the third quarter” according to CEO Lisa Su. A generation of CPUs which inaugurates both the new Zen 4 cores, the AM5 platform, DDR5 and 5 nm engraving (read below). Given that there is often a lag between announcements and the arrival of products, it is likely that PCs will arrive in the fourth quarter rather – mobile chips will arrive during 2023.
But this announcement is followed by the arrival “by the end of the year”, of the new generation of RDNA 3 graphics chips. More precisely, “the high-end version of the RDNA 3 GPU”. Also engraved in 5 nm, this new generation of GPU should respect AMD’s incredible rise in power over the past 3-4 years: a 50% gain in performance per watt compared to the previous generation. A “+50%” that AMD was able to hold from GCN to RDNA 1 and from RDNA 1 to RDNA2. As with the transition from Zen 3 to Zen 4 in processors, part of the jump in performance comes from intrinsic improvements to the chip. The rest (and a beautiful piece!), of the reduction in the size of the circuits.
5 nm to (re)gain the advantage?
On the GPU side, AMD will be (almost) on par with the new generation of Nvidia’s RTX4000: the two competitors both use one of the 5 nm processes (N5 or N5P?) from the Taiwanese TSMC. In the graphics combat arena, this equality will, for once, decide which architecture (and of the software ecosystem, the drivers being crucial) will be the best. If Nvidia has delighted the hearts of gamers for years – its market shares communicated by Steam bear witness to this – the RDNA 1 and then RDNA 2 chips have enabled AMD to achieve an impressive technological “upturn” over the past three years. Especially in consoles, however: if RDNA 2 powers both the Xbox Series X, the Playstation 5 and the Steam Deck, its popularity is low among PC gamers.
The situation is different when it comes to CPU. At Intel, the current chips (12e gen, Alder Lake) as well as the next generation (13e gen, Raptor Lake) are engraved in the Intel7 node, the “Intel” version of the 10 nm which offers a density of transistors equivalent to TSMC’s 7 nm. Here, the advantage of node (the finesse of engraving transistors in the jargon) is for AMD. According to TSMC, compared to 7 nm (N7), the 5 nm node “makes it possible to reach 20% higher frequencies or a reduction of around 40% in energy consumption (at equal frequency, ndr)”. AMD seems to have more margin than Intel on this generation, in particular to contain energy consumption.
Since the launch of Zen in 2017, AMD has managed an incredible comeback in all segments, which has allowed it to take certain technological advantages. And even to afford a major takeover with that of Xilinx, which allows it to get closer to the “mass” of Intel’s r&d in certain areas. However, the American chip designer faces challenges in the availability of its products and total dependence on TSMC. A Taiwanese founder, world leader in engraving, but whose destiny is intrinsically linked to Taiwan. A small country currently under the (military) radars of an increasingly threatening China.
WCCFtech