Apparently, AMD is producing a Zen-based APU for the Chinese market that will have HBM. It will reportedly be deployed there in desktops and then in a Chinese-market console.
https://www.extremetech.com/gaming/274919-more-details-on-the-new-amd-powered-chinese-console
The main drawback, in terms of the standard PC gaming platform, is the RAM split:
The Vega tech that helps with lower VRAM will likely help with the VRAM being 4 GB but programs might have an issue with 4 GB of system RAM.
That article explicitly states that the console uses GDDR5, not HBM. As confirmed by the pictures, and every other publication covering it.
As with other APUs in Windows, it likely uses dynamically allocated shared memory (with a fixed base amount), so VRAM allocation is likely adjusted by need. This works perfectly fine.
I heard the same things after Bitcoin. Ethereum became the next big thing, though. It doesn't seem at all certain to me that there won't be another "next big thing" in crypto. Having many existing coins also doesn't prevent that possibility. There are plenty of examples in tech where the market was saturated by players — and yet there were next big things. The NES. The PlayStation. The XBox. The greatest example is the IBM PC, which came onto a market with a lot of microcomputers already available. It had a big corporation behind it, which is why it succeeded. It wasn't its technical merits that made it sell. There were also enough well-known search engines, including metasearch engines, that plenty of people didn't predict Google.
Gambling has existed for a long time and there is a lot of money involved in it.
Was there ever a time before when people were scared of buying (and desperate to sell) Bitcoin? I sure can't remember that. Skepticism, sure, but not the "run away" attitude seen today. The situation is fundamentally different. This of course doesn't mean that a new wave of crypto won't appear - the financial "industry" doesn't like to leave potential ways of generating money alone for long, even if they're currently terrified of it. But it will likely take some time.
My point about Jaguar is the reason it was used, and especially kept for a second iteration, is because of the artificial effect of duopoly. Like monopoly, only with less severity, consumers get less product for their money. There are benefits of monopolization but the overall picture is a negative for consumers.
That's a bit of a stretch. Of course, we could speculate that if the X86 CPU market wasn't a duopoly, there might have been an established low-power CPU arch available in 2011-2012 when this console generation was designed, but that's rather meaningless speculation.
As for the mid-gen refreshes (Pro and X), they both arrived too early to implement Zen - the design wasn't yet ready for the PS4 Pro, let alone tested and known to perform outside of AMD's labs. The One X arrived later, but still too early for an implementation like that (which requires the design to be very well tested and known good). Then there's the issue of dramatically increasing CPU power in games - how do you make games scale for CPU power across such radically different designs? This makes sense for a "new generation" (which is becoming an increasingly meaningless term in the age of X86 consoles, but still makes sense in terms of software development), but not for a mid-gen refresh - you'd end up with games only working on the refresh, pissing off all the people who bought the other console 1-4 years earlier. Consoles are expected to have 5-8-year life cycles, not ~3 like a PC.
Then there's the issue of die area and cost.
The Scorpio Engine is a 359mm2 die (on TSMC 16FF), of which the 8 CPU cores make up a tiny fraction. For a console, this is
huge. In comparison, an original
Zen Zeppelin (8c) on GloFo 14nm is 213mm2. Of course that has components that could be removed in a console (such as the DDR4 controller, USB, SATA, and PCIe PHY), but adding 8 Zen cores would still balloon die size dramatically. The addition of L3 cache alone would grow the die noticeably. Even reducing it to a single CCX (
44mm2 when excluding everything else) for 4c8t would still entail a noticeable size increase, not to mention the issue of patching the OS, games and apps to account for 4 fast and 4 slow threads. Then there's the licencing cost of Zen cores vs. Jaguar cores, which would likely be in the 5-10x range given how new the arch was. And $1000 consoles don't really exist - for a good reason, as they wouldn't sell. $500 consoles usually struggle. This has little to do with a duopoly, and much more to do with the realities of chip design, chip production and fab costs. The tech simply wasn't ready in time, and while an argument can be made that a higher power/IPC/clock arch with fewer cores would have been better for gaming in the short term, that's not the direction either of the big console makers went (and thanks to the 8-core designs, consoles have a lot of cool functionality that would have been impossible otherwise). This is also likely due to them seeing single core perf flatlining and wanting to prepare their developers for the multi-core future. IMO, that's sound long-term planning.