Sizzling Chips 32, the (digital) convention that delves into silicon layouts from across the damage, is in complete swing this 7 days and AMD is a single of the corporations teeing up to chat about its hardware. The Xbox Sequence X has been the issue of a great deal speculation — now, thanks to Microsoft’s presentation, we know more about how the useful blocks of the SoC are arranged.
AMD/Microsoft opted for two quad-core CPU clusters rather than a unified 8-core chip. It would be pretty exciting to know what the penalty is for a cross-cluster L3 look for, but we didn’t get a ton of new information and facts off this slide.
The total dimensions of the Xbox SoC is 360mm2, down a little bit from the 375mm2 of the Xbox One and the 367mm2 of the Xbox One X. In the earlier, approach node shrinks resulted in a web charge reduction. Today, which is no more time the case. The density leap from 16nm to 7nm is equivalent to a complete node reduction, per AMD’s remark.
This is the substantial-stage all round aspect chart for the SoC. Blue blocks refer to hardware processing blocks, and the very last things referenced are protection hardware blocks. The reference to device learning acceleration is exciting, but there is no signal of a devoted AI processing block on the die shot Microsoft unveiled. This may refer to GPU-based processing — or Microsoft could have one more shock or two, I suppose.
As for the GPU:
The remarks on “Ray-box” and “ray-tri” are very likely references to ray-box intersection tests and ray-triangle intersection tests. The two of these are very important actions in ray tracing, nevertheless simply realizing these two numbers does not tell us nearly anything about what the alternative will be capable of or how it compares with Nvidia hardware. The chat of a slight region charge for a 3x – 10x acceleration is exciting. One issue we really don’t know nevertheless about RDNA2 is whether or not or not it has any specialized hardware blocks for these sorts of calculations. Investing more die dimensions to increase performance is not a undesirable tradeoff, nevertheless definitely it’s more outstanding when the speedup is toward the 10x conclude of the array.
The RDNA2 diagram Microsoft offered appears like RDNA with the addition of the ray acceleration blocks. Every single CU is capable of undertaking either 4 texture functions per clock or 4 ray “ops”, but not the two. We really don’t know nearly anything about what a ray “op” consists of, but the simple fact that it’s an either/or circumstance may make clear why Microsoft’s illustration precisely states ray tracing will be an “economical upgrade” rather than a comprehensive substitute.
Total, Microsoft’s description of its ray tracing sounds very similar to what the Personal computer has realized to-day with Nvidia’s Turing, nevertheless definitely we really don’t nevertheless know the particulars of AMD’s technologies. It is solely probable that RDNA2 will be speedier than Turing at ray tracing — but what that may well make it possible for AMD to do is pack a more Personal computer-like performance stage into a console as opposed to deploying some kind of RT-redefining performance beast.
Even now heaps we really don’t know, but exciting details below. The Xbox Sequence X is going to pack Personal computer-like performance and Personal computer-like abilities, from backwards cross-generational functionality to at the very least some video games delivered in a “purchase the moment, play on both” fashion. The hardware within this technique wouldn’t be out of position in a midrange gaming Personal computer. The a single significant issue we really don’t know? The cost. Microsoft is holding mum on that so far.