I didn't know about the rigidity of this pipeline. What's up with that? And why must it always be 3:1?
Any links where I can read more about this?
Well, I've read it from
here (some other write-ups on the wafers for the systems are
here as well, though this source seems to estimate more Xs than Ss at least for the time period the article was written), and it sounds like they know a good bit about the production process. The way MS have their wafer setup is allocating space for the S and X on one wafer to save costs, rather than having one wafer for S chips and another for X chips. Plus, each wafer has to be provided by TSMC, and they're in limited quantities varying depending on the node.
So basically, the wafer is taped out for 3x as many S chips as X chips; if Microsoft wanted to change that, they'd have to get a new wafer (whether a replacement or just an additional one) and have it taped out for a different ratio. Unfortunately for them, chiplets aren't a thing for RDNA2, otherwise I'm guessing they could have just arranged the wafer for CPUs (since the two systems basically use the same CPU) and GPU chiplet tiles where say Series X has 3 or 4 and the S just has one. There would still be logistical production issues to contend with, just not as early in the pipeline, so they'd of had more flexibility.
Sony doesn't have this problem because all PS5s use the same APU, so it's just a matter of making sure they have enough wafers (and I'm pretty sure Sony paid for more wafers than Microsoft did). MS's setup favors the S but it also means it's pretty inflexible; now that it seems S demand has been tapped out, they would have to increase Series X volume but it doesn't make sense to only use 25% of a wafer, now does it? So then they either eat the costs on not using 75% of a wafer, or they eat the costs putting that 75% in systems that don't have a lot of demand so they have to cut the prices on and lose out in revenue anyway. It's a lose-lose situation.
Or, they buy more wafer and use it solely for X chips, but now that'd be an extra cost in production. Do they then shut down the old wafers that have mostly S chips? They'd still probably have to make something to replace the Series S; even if they don't, they'd still have to maintain support for it through the rest of the generation for the people who purchased one. The logistical production problem for Xbox at the wafer level is even worst than I ever gave thought to before maybe a week ago.
Excellent analysis post btw, and i'm glad to see some thought put into deriving some estimates and forecasts on here.
Based on the numbers you and
Heisenberg007
are discussing, it seems that the bigger pubs are even more bear-ish on MS' ability to sell HW this year than even we are.
Your last 2 paragraphs citing the production line issues MS has with having to produce wafers at a 3:1 split between S and X chipsets could be why: if demand for the S is tanking, and MS has no choice but to produce X chipsets this way, then that means they are virtually handcuffed on their ability to move units.
That actually makes MS' leadership look EVEN WORSE.
Yep, more or less. Always had doubts about this X & S approach right off the bat; even aside the production issues it's now causing, it just really confused the messaging that was more or less (IMO) pretty on-point from when they revealed Series X up to that May "gameplay" showcase. Which was the first big blunder for their messaging, but at the time to me it seemed like a footnote.
There's also the potential development problems the approach could have on more ambitious 1P & 3P AAA games in the next few years; not everything scales down easily the way resolution or textures do. Game scope could be pared back due to Series S; there are non-graphics uses (general purpose tasks, physics, logic, AI etc.) for the GPUs that we may never get to see fully utilized on Series X because the Series S exists, and Microsoft still mandates all games support them both with native builds.
They really need to figure something out with Series S; if demand remains low as-is, could be in their best interest to just use the chip for the cancelled Project Keystone. Bring out that Game Pass xCloud streaming box, give it like 4 GB RAM, some cheap eMMC internal storage, downclock the chip as necessary, market it as a Roku & Apple TV competitor with "custom" Game Pass xCloud functionality. Throw in a controller & universal remote, sell it for $150, get some deal with Disney+, HBOMax, Netflix whoever with 6 months free bundled in the package. Maybe that would actually work.
The current strategy they have is leading to a death spiral for Series S. So they either have to scrap it altogether, hope to generate some big demand for it, or spin the chip into a product that can have the costs cut down a lot & maybe actually see more demand. They could maybe also put some of those S chips into Surface laptops but, Surface sales dropped a ton last fiscal quarter too. Hardware in general just isn't looking very good for Microsoft.
Yeah, because this doesn't make logical sense to me. No hardware launch ever had a weak version that sold better than the "real" version. Do you guys remember the gimped PS3 and Xbox 360 Arcade options?
Those indeed didn't sell better than the "genuine" versions, but they had a big difference: the performance specs even in those systems were the same as the "premium" versions. Same CPU, same GPU, same RAM, same Southbridge, same cooling. They had different storage capacities and I/O, but those affected more so QoL features for the end user, and didn't necessarily impact game performance between models.
Series S has a completely different GPU to Series X, different RAM capacity, different CPU clock (everything else CPU-wise is probably the same though), alongside different storage capacities and some difference in I/O. The GPUs being so vastly different, along with the RAM being so different (both in capacity & speed) have very real impacts on games across the two systems. Stuff like GPGPU is going to be negatively impacted if software has to run natively on both systems at acceptable levels.
MS & Sony both knew that production costs reduction was going to slow down this gen compared to previous ones, but Microsoft once again lacked patience, they rushed to a solution that looked great on paper in the short-term but is actually an albatross holding them back long-term. Sony designed the PS5 so that they could still find ways to lower production costs after an initial amount of time, including moving to 6nm (are the Series X & S even on 6nm yet? Can they move their wafer design to 6nm without needing to do significant re-designs of the wafer tapeout?) and other stuff.
So I guess once again, Sony made the right call. The funny part is how most of last year and in 2021, some people were shouting at the top of their lungs the complete opposite. They spoke too soon.