Bofferbrauer2 said:
Captain_Yuri said:
https://www.tomshardware.com/news/microsoft-xbox-series-x-architecture-deep-dive
" That GPU section is, not surprisingly, massive. The full chip is 360.4mm square, with 15.3 billion transistors. Doing some quick image analysis, the GPU takes up roughly half of the die (47.5% if you want a more precise estimate)."
"A Zen 2 CPU chiplet measures 74mm square (with four times the L3 cache compared to the Xbox Series X APU), and then tack on a GPU that has more features and shader cores than Navi 10 (RX 5700 XT), which measures 251mm square. That's 325mm square without the enhanced Navi 2x cores and 12 additional CUs."
|
Is it just me or do those sizes don't add up?
360mm2, out of which the GPU takes less than 50%, so only ~175mm2? I find that die size hard to believe since Navi is already produced in 7nm, has less CU and no hardware Raytracing yet was already 251mm2. If that is truly the size of the GPU part, then they would have needed to severely cut down on the cache sizes inside Navi to make it this small, I don't see what else they even could have cut.
CPU looks like cut out right of Renoir, complete with the cache size.
|
I/O, Shared Caches and so forth take up a massive chunk, there is allot of consolidation of various pieces of logic when building an APU.
JEMC said:
NAND prices going down is good for all. hopefully that fall arrives to us, customers.
Captain_Yuri said:
Also Microsoft is going to have Xbox Series X's AMD Architecture Deep Dive at Hot Chips 2020 today at 6:30PM or 7 PM PST. It could give some insight into RDNA 2 and AMD's next gen APUs.
They have already gave out some slides on what to expect from the presentation.
https://www.tomshardware.com/news/microsoft-xbox-series-x-architecture-deep-dive
" That GPU section is, not surprisingly, massive. The full chip is 360.4mm square, with 15.3 billion transistors. Doing some quick image analysis, the GPU takes up roughly half of the die (47.5% if you want a more precise estimate)."
"A Zen 2 CPU chiplet measures 74mm square (with four times the L3 cache compared to the Xbox Series X APU), and then tack on a GPU that has more features and shader cores than Navi 10 (RX 5700 XT), which measures 251mm square. That's 325mm square without the enhanced Navi 2x cores and 12 additional CUs."
"While the chip size of the Xbox Series X is in line with previous console hardware (375mm square for the Xbox One in 2013, 367mm square for the Xbox One X in 2017), and transistor counts have more than doubled relative to the Xbox One X (6.6 billion to 15.4 billion), the die cost is higher. Microsoft doesn't specify how much higher, but lists "$" as the cost on the Xbox One and Xbox One S, "$+" for the Xbox One X, and "$++" for the Xbox Series X. As we've noted elsewhere, while TSMC's 7nm lithography is proving potent, the cost per wafer is substantially higher than at 12nm."
|
(This one is shorter to quote than the other one.)
I'll wait until someone does a simplified summary so I can understand it. But I've looked something: during the Turing architecture deep dive they did at Anandtech, they mention that the 2080Ti is capable of 10 GigaRays/second (page 7). That slide says that the custom XBX solution can do up to 380 G/sec ray-bos peak or 95G/sec ray-tri peak. I don't know how comparable are both metrics, but this does seem to confirm that Turing won't age well for ray tracing tasks.
|
Left my bath salts somewhere... Only need a pinch of it.
In seriousness, I wouldn't be surprised, AMD and nVidia can invest more into RT at 7nm than at 12nm. (Aka refined 16nm which is a refined 20nm.)
But like all things tech, wait for the real-world benchmarks to see where the cards fall, nVidia may have employed better techniques that drive up efficiency.
