Scorching Chips 32, the (digital) conference that delves into silicon types from across the harm, is in complete swing this 7 days and AMD is one particular of the businesses teeing up to converse about its components. The Xbox Collection X has been the issue of substantially speculation — now, many thanks to Microsoft’s presentation, we know much more about how the practical blocks of the SoC are organized.
AMD/Microsoft opted for two quad-core CPU clusters fairly than a unified eight-core chip. It would be very attention-grabbing to know what the penalty is for a cross-cluster L3 look for, but we didn’t get a ton of new facts off this slide.
The whole dimension of the Xbox SoC is 360mm2, down a bit from the 375mm2 of the Xbox One particular and the 367mm2 of the Xbox One particular X. In the earlier, procedure node shrinks resulted in a internet charge reduction. These days, which is no lengthier the situation. The density leap from 16nm to 7nm is equivalent to a complete node reduction, for each AMD’s remark.
This is the high-stage over-all aspect chart for the SoC. Blue blocks refer to components processing blocks, and the past factors referenced are safety components blocks. The reference to machine understanding acceleration is attention-grabbing, but there is no sign of a dedicated AI processing block on the die shot Microsoft introduced. This could refer to GPU-primarily based processing — or Microsoft could have another shock or two, I suppose.
As for the GPU:
The remarks on “Ray-box” and “ray-tri” are most likely references to ray-box intersection checks and ray-triangle intersection checks. Both of these are crucial steps in ray tracing, while merely figuring out these two quantities doesn’t notify us anything at all about what the answer will be capable of or how it compares with Nvidia components. The converse of a slight location charge for a 3x – 10x acceleration is attention-grabbing. One particular point we do not know however about RDNA2 is no matter whether or not it has any specialized components blocks for these styles of calculations. Investing much more die dimension to improve performance is not a negative tradeoff, while obviously it’s much more remarkable when the speedup is to the 10x finish of the selection.
The RDNA2 diagram Microsoft offered looks like RDNA with the addition of the ray acceleration blocks. Just about every CU is capable of undertaking both 4 texture operations for each clock or 4 ray “ops”, but not each. We do not know anything at all about what a ray “op” is composed of, but the simple fact that it’s an both/or problem could explain why Microsoft’s example especially states ray tracing will be an “economical upgrade” fairly than a total substitution.
Overall, Microsoft’s description of its ray tracing sounds identical to what the Pc has achieved to-date with Nvidia’s Turing, while obviously we do not however know the particulars of AMD’s technologies. It is entirely doable that RDNA2 will be more rapidly than Turing at ray tracing — but what that may possibly allow for AMD to do is pack a much more Pc-like performance stage into a console as opposed to deploying some sort of RT-redefining performance beast.
Still heaps we do not know, but attention-grabbing specifics in this article. The Xbox Collection X is going to pack Pc-like performance and Pc-like capabilities, from backwards cross-generational ability to at the very least some games shipped in a “purchase when, perform on both” vogue. The components inside this technique wouldn’t be out of position in a midrange gaming Pc. The one particular massive point we do not know? The rate. Microsoft is holding mum on that so significantly.