Unlike Nvidia, which returned at length during its conference on its game partnerships and the events that marked the past year, AMD did not really take the time to mess around, as we say further north. Three and a half minutes of self-congratulation and it was over, room for new things! It must be said that the manufacturer was expected on many fronts, that of GPUs like that of CPUs, on the desktop and laptop markets.
And the first big announcement precisely concerned this last area, with the launch of the Ryzen 6000 processors, code name Rembrandt. Despite a name that implies a new generational leap, the Ryzen 6000 actually retain many of the attributes of the Ryzen 5000 chips that we found in the desktop space, starting with the Zen 3 architecture. before a certain number of optimizations to justify the new mention Zen 3+.
Improving energy efficiency has been at the heart of the concerns of the brand’s engineers, and what could be more logical for products intended for the mobile space. In this regard, the switch to a 6 nm engraving process (announced by TSMC during the last Computex) also had to pave the way for new interesting technical compromises. However, the main principles of Zen 3 remain relevant: the Zen 3+ chips therefore find a so-called structure chiplet which will mix different levels of engraving, between the CCX block which embeds the calculation cores, and the I / O block. We also keep the concept of a single L3 cache pool shared within a CCX, to reduce latency effects.
In truth, the revival of Ryzen 6000 mobile chips comes mostly from “secondary” elements. To begin with, all new generation processors will benefit from an integrated graphics circuit, which is no longer based on the Vega architecture but on that of current desktop cards, called RDNA 2. For H series processors, which will certainly be accompanied dedicated graphics circuits, the news is obviously anecdotal, but for class U chips (low or even very low consumption), this development could be very appreciable. Our past experiences had shown us that the Vega 8 and 10 circuits performed with honors in the face of the revival of Intel iGPs and their Xe-LP architecture.
And the figures delivered by AMD during its presentation suggest considerable performance gains, whether compared to the 11th generation Intel G7 circuits, or to the competition from Nvidia (MX 450, in this case). Obviously, AMD did not fail to mention that its iGP will be able to speed up calculations. ray tracing. A first in the field, of course, but in practice, it will surely be difficult to find in this functionality a favorable field of expression (perhaps on a title like the LEGO Builder Journey, and more).
The other major development concerns the platform: it will thus integrate a number of new technologies, such as support for USB 4, PCI-E interface 4, DDR5 / LPDDR5, Wi-Fi 6E, and Bluetooth 5.2. Audio and video inputs / outputs are also not left out, with the arrival of HDMI 2.1 and DisplayPort 2 on one side, and the integration of a module carried by the AI to cancel ambient noise. and the effects of echoes.
In all, no less than 13 new Ryzen 6000 processors, carrying between 6 and 8 cores, will equip laptops from all walks of life from February 2022. Note that AMD is also announcing 3 class U 5000 series chips, which will respond to previous Zen 3 / Vega architectures.
And the announcements on the mobile side do not stop there: after the CPUs, it is the GPUs that have been honored, this time not with the introduction of a new range, but the widening of the range. existing. The RX 6000 began to attack the market on the 1st day of Computex last June, and will find some reinforcements during the first quarter of 2022, with three new M references, two XT references and three S references.
In detail, we already knew the RX 6800M, 6700M and 6600M chips, in decreasing order of performance. The new RX 6650M, 6500M, and 6300M circuits will notably strengthen the entry level. By way of comparison, AMD positions the RX 6300M in its communication against the GTX 16580, while the 6500M would target more the RTX 3050. The XT chips are for their part slightly boosted versions of the models of which they share the denomination. For example, the RX 6650M XT will take the main features of the RX 6650M, but will offer 4 more Compute Units (28 versus 32).
Finally, the circuits marked S (6800S, 6700S and 6600S) should be seen as equivalents to the Max-Q products that we already knew from Nvidia: GPUs less focused on pure performance than on the best performance compromise, energy efficiency and cooling management (think S for Silence, if that makes it easier for you). As such, we should logically find them in laptops assuming a certain finesse of design and a desire to maintain good autonomy. Be careful, however: we know that between the expectations of GPU manufacturers and the vision of integration partners, there can sometimes be a gulf.
New CG, new processor
Hang in there, we’re almost at the end! Behind this large train of announcements aimed at the mobile world, the manufacturer had kept two products under the elbow to satisfy sedentary players: a new graphics card, the RX 6500 XT, and a new processor, the Ryzen 7 5800X3D (Zen 3 architecture, 7/12 nm).
For the first, the movement is not uninteresting in itself. However, this reference does not bring any new technical element compared to the current generation, and aims above all to counter the last “accessible” desktop card from Nvidia, the RTX 3050. Let us note in the process that its official price of 199 dollars seems quite enough utopian (but its competitor is in exactly the same situation).
The processor, on the other hand, gives us a glimpse of part of the future of Ryzen products. Indeed, the Ryzen 7 5800X3D integrates the famous 3D cache memory that we mentioned last October. Remember what it is: today, the functional elements that make up a processor (cores, iGPU, cache memory, modem, etc.) are positioned on the same plane. Since the surface of a die is not infinitely expandable, the manufacturer must therefore make compromises, in particular by dimensioning its cache memory as close as possible to the needs of the cores. And in this context, 3D memory appears like the Holy Grail, since it is integrated vertically, stacking layers of components linked to each other and to the CCDs by TSVs (for Through Silicon Vias, vertical channels which will allow communication between the different levels of the new die).
In practice, this allows the Ryzen 7 5800X3D to embed 96 MB of L3 cache, against 32 MB for the previous version of the 5800X. The gains, in game as in application, will certainly be variable, as AMD points out, but the brand still counts on a good 15% additional performance on average. 15% which could thus already be acquired for the following generations. The CEO Lisa Su also quickly mentioned this subject, and thus confirmed that the Ryzen 7000 products will have a new architecture (Zen 4), a new socket (AM5) and a new engraving fineness (5 nm TSMC).
And the dates? What about the prices? Quite logically, this information is not yet topical for the next generation, expected for the second half of 2022. On the other hand, AMD announces the availability of the Ryzen 7 5800X3D in the course of the second quarter of 2022.
To read also | Our pick of the best gaming laptops