We have commented on it many times, the future in all areas of computing not only involves having a powerful CPU architecture, but also a GPU architecture where both elements are combined in a symbiosis that allows accelerating the algorithms most used in different programs. . So far, Intel has had a huge deficit in performance of its integrated GPUs, which has allowed AMD to win several important contracts and invade spaces by taking advantage of the superiority of its graphics architectures, but this could change radically in 2022.
Intel reveals why TSMC makes its gaming GPUs
During the ASCII conference in which they spoke about their ARC Alchemists, Intel has revealed the main reason for choosing the TSMC node N6 for its first generation of gaming graphics cards: it gives them a good manufacturing capacity. Apparently, the Intel equivalent node does not seem like a good candidate for Intel ARC Alchemist GPUs because it does not give them enough manufacturing capacity (and / or will be destined for other products).
The reality is that the Intel 7 node, formerly known as 10nm SuperFin, has a much higher error rate per area than the equivalent TSMC nodes and therefore the number of good chips per wafer would be less if they were manufactured at home. Besides, we cannot forget that Intel wants all the possible manufacturing capacity to have its CPUs based on Alder Lake-S architecture ready.
For future generations of their ARC gaming graphics cards, Intel has not revealed which manufacturing node they will use, if their future nodes such as the Intel 4 or TSMC nodes like N5 and N3. It should not be forgotten that the choice of the manufacturing process affects the subsequent design of any processor, so the company in blue must already have chosen where it will manufacture its next generation: ARC Battlemage, but at the moment he has not revealed it.
Intel Drops Its Most Powerful Intel ARC Alchemist At ASCII Conference
So far we know of the existence of two chips based on Intel’s first truly gaming GPU, but it could be that we see a more powerful graphics processor. Which was something obvious since the team led by Raja Koduri had not presented a graphics card that would compete with the high-end of NVIDIA and AMD.
Given the size of the most powerful Intel ARC Alchemist revealed to date, it is clear that there was a margin and at the ASCII conference Koduri himself has confirmed that its architecture is not limited to 8 Xe Slides. Which makes us think that the company founded by Gordon Moore could be preparing a GPU with a bus of 320 or 384 bits and most likely with 10 or 12 Xe Slides, which would place it at the level of the NVIDIA RTX 3090 SUPER that is expected for the same dates.
This third GPU would not appear in the slide above and therefore would not have entered the manufacturing phase yet, so it could arrive somewhat later than the first two graphics processors based on the ARC Alchemist architecture. Is this Intel’s best kept secret to surprise NVIDIA and AMD? Surpassing an NVIDIA RTX 3090 is already a titanic task and more in its first dedicated gaming architecture …