Maybe the cold weather is affecting me, but I still don't understand why they'd have a second heatspreader, PCB, or I/O chip. I would've assumed they'd use a dummy die for single chipet products and designed the PCB in such a way that it could be used with either 1 or 2 chiplets.
Sorry I'll try to give more back ground in my thought process.
AMD are likely to have this combination of CPU's no matter the name in price.
4c8t Ryzen (maybe)
6c12t Ryzen
8c16t Ryzen
12c24t Ryzen
16c32t Ryzen
16c32t TR (no 12c version)
24c48t TR
32c64t TR
48c96t TR (Maybe)
64c128t TR (maybe but probably more unlikely)
Then EPYC with the TR configuration with no maybes
So volume works like this.
The 4c8t Matisse will be a ok seller if it exists. But for the most part I would expect it won't sell as well as the rest as interest in a 4c Ryzen will be split with the APU's.
The 6c12t Ryzen is going to be by far the best seller. It will be a single chiplet chip (they aren't going to do the whole unbalanced CCX). Maybe there are variations that will be a bit more expensive. But in the end we at worst are looking at a sub $200 6c chip that AMD will sell the most of.
8c16t Ryzen will be an interesting one AMD could sell a 2 Chiplet version or a single Chiplet version. Or Both and the one that performs better gets the premium cost. Personally I think the 4c chips will have better value in TR and EPYC where the volume of chips on a CPU make up for how cut down they are. Not doing a 2 Chiplet 8c Ryzen, reduces the number of good 8c and 6c chips they have to fuse off cores and down sell. This should be the config that comes closest to taking on the 6c12t chip. I think it has a decent chance possibly upsetting the 6c12t chip as the best selling configuration of Matisse.
This is where it is going to get dicey. The 4c and 6c chips are almost certainly 1 Chiplet version. All of the three above have at least 2 if not more configurations for performance and power. Specially the single chip ones. They could have 35-45w versions, 65w versions and +90w versions. So there will be ample opportunity to take up many price points. A 12c Ryzen and a 16c Ryzen have two options. 1 is AMD realizing how far ahead of Intel they are, price the 12c to compete against Coffelake-R and CometLake and leave the 16c8t Ryzen significantly more expensive for margin and halo purposes. 2. Is price the 12c and 16c CPU's and in fact compress the whole lineup price wise to make the question not "Should I get a CFL-R/CometLake or a Ryzen 3k" it will make it "Which Ryzen 3k am I going to get".
So I will go over the affect of both theories. Starting with the keeping up with Intel cost. AMD has never been one to shy away from super high priced CPU's. But when they did it with the FX versions of the A64 and A64x2. These were all the same chips. If the FX's didn't sell, they just didn't alot as much dies to it. All the margin was free margin with little to no other constraints. They also didn't have a product range like TR act as that bridge to super expensive consumer CPU's. When AMD priced their FX's to 1k, those were HEDT CPU's before we had a name for it. So now to go into the impact of doing an expensive 16c32t Ryzen (this applies as well to the 12c but a little less straight forward). Lets call the 16c32t CPU a $700 CPU ($50 less than a 2950). This won't sell very well. Game wise it will be slower than a single chiplet 8c CPU. It's over twice as expensive as the general consumer socket CPU. Nearly 50% more than either AMD's or Intel's highest cost consumer chip in recent history. There are no if ans or buts about them if they space out the price on the CPU's and keep 6c and 8c chips roughly the same price as now a $500 12 and $700 16 core chips won't sell really well. In some worlds that is okay. But its how they affect the sales of the 6c and 8c chips that I am worried about. Maintaining current costs If Ryzen 3k sells well the 6c and 8c CPU's will take the brunt of the sales. Meaning a significant amount of of those 6c chips will be cut down 8c Chiplets. Even now those chips sell for ~$200.
So now that we have established the affect on how the wide and at the top end expensive range of pricing lets look at the Ryzen 1k pricing window. Keeping the 6c at ~$200, the 8c ~$300, the 12c ~$400, 16c ~500. As the prices lowers on the top end the sales exponentially grows. A $500 16c CPU won't set the world on fire but it will sell huge amount more than a $700 one. The more of them that AMD sells they might actually be getting a higher ROC (return on cost). So a $500 is under the double price of a 8C Ryzen at $300. But if AMD sells 2x 8c Ryzen, they are building 2 full chips. 2 More IO Dies. 2 More heatspreaders. 2 More sets of pins. 2 more PCB boards with the needed work on the substrate. In the end they might be making more profit on one $500 16c CPU then 2 8c Ryzen's do to the increased manufacturing costs of two CPU's versus 1 with just an additional chiplet. Now we start getting into my original point. Down binned 8c chiplets fused into 6c CPUs. It is in AMD's best interest to sell as many 8c chiplets as 8c Chiplets as possible. AMD can't just limit 6c chips to 6c capable chiplets. Its going to be a massive seller. Might not be many at a launch and their will probably be a lot more legitimate 6c chiplets in these CPU's then there were for the 6c cpu's in Ryzen 1k and 2k. But I still see a significant amount of them being 8c chiplets down binned and fused. So anything that AMD does price wise to get people to purchase 8c Ryzen 3k and higher, the better. For every 16c they don't sell that AMD has to take stock that would go to that chip and bin it for a 6c Ryzen, the more lost money.
Sure at $700 a chip they could be getting $350 a Chiplet. But for every one of those they don't sell and the wider the array of prices in works it's way down the line and those chiplets just became 2 $200 CPU's with twice the manufacturing cost. Where if they lower prices, squeeze the prices together. The more initialized people are to buying a 8c Ryzen, 12c Ryzen and 16c Ryzen. Increasing sales on items with more profit with reduced manufacturing costs. This doesn't even touch on the economics of increased profits even if it means lower margin on increasing volume of sales through price reductions (or in this case lower initial pricing). Considering where AMD still is in the market. Their eye should be as much on volume as it is on Margin and Profits.