At the end of July, AMD will release two series of processors – desktop Ryzen 9000 (Granite Ridge) and mobile Ryzen AI 300 (Strix Point) – based on the latest Zen 5 architecture. The HardwareLuxx.de portal found out the missing details about these processors. In particular, the sizes of the crystals of these chips and the number of transistors used in them became known.
Image source: AMD
The Ryzen AI 300 (Strix Point) processors are built on a monolithic crystal, which is manufactured using TSMC’s 4-nm N4P process technology. This is a slightly improved version of the N4 process technology, which is used to manufacture AMD’s Phoenix and Hawk Point processors based on the Zen 4 architecture. The Strix Point crystal area is 232.5 mm2. Thus, it is significantly larger than the Hawk Point and Phoenix crystals with an area of 178 mm2.
Image source: HardwareLuxx.de
The increased area of the Strix Point crystal is explained by the increased number of execution units of the integrated Radeon 800M graphics on the new RDNA 3.5 architecture from 8 to 12 for the Ryzen AI 9 365 model and from 12 to 16 for the Ryzen AI 9 HX 370. The area of the new crystal has also become larger due to the increased L3 cache memory volume to 24 MB and, in general, due to larger Zen 5 and Zen 5c cores.
The Ryzen 9000 desktop processors (Granite Ridge) use the same chiplet layout as the Ryzen 7000 (Raphael). AMD confirmed that the new processors use the I/O die (cIOD) from Raphael, manufactured using the same 6 nm process technology. The area of this crystal has not changed and is 122 mm2. It contains 3.4 billion transistors. For comparison, the cIOD of the Ryzen 5000 (Vermeer) and Ryzen 3000 (Matisse) processors were manufactured using the 12 nm process technology from Global Foundries and had an area of 125 mm2, but contained significantly fewer transistors – 2.09 billion. The key factor in the increase in the area of the cIOD crystal was the integrated iGPU block with two execution units.
The Ryzen 9000 processors use a new CCD die with eight computing cores, called Eldora. According to HardwareLuxx.de, it is manufactured on the same 4nm N4P process as the Strix Point processors. However, according to other sources, these chips may be manufactured using an even more advanced N4X process, which works better with high clock rates.
Image source: HardwareLuxx.de
The Ryzen 9000 CCD itself contains 8.315 billion transistors, which is a significant increase compared to the 6.5 billion transistors in the eight-core Durango CCD of the Ryzen 7000 chips based on the Zen 4 architecture. Notably, despite the 28% increase in transistors, the area of the Eldora CCD based on Zen 5 is 0.5% smaller than the area of the Durango CCD on Zen 4 – 70.6 mm2 versus 71 mm2. As a reminder, the Durango CCD on Zen 4 is manufactured using the 5 nm TSMC N5 process technology.
Image source: HardwareLuxx.de
Thus, the total number of transistors in the flagship 16-core Ryzen 9 9950X processor with two CCD chiplets is 20.03 billion. In turn, the Ryzen 7 9700X with one CCD contains 11.715 billion transistors.
If you notice an error, select it with your mouse and press CTRL+ENTER.
Source: 3dnews.ru
