"Leaked" prices I have seen for the 9800X3D  low end $484.00 to a little over $500.00.

I will wait with my 7600 non X for a while longer to see how higher end X3D cpu's pan out.

I would expect $500 given the price the 7800X3D is currently selling at, and the fact that Intel still has nothing to compete with the 7800X3D, much less the 9800X3D.

7600 is still a great CPU, and you will likely have some high value upgrade options in the future if when they inevitably release a 9700X3D, and maybe a 9600X3D as well.

So they packed the L3 cache much tighter to save die space while also shrinking the TSVs for the 3d cache ridiculously small that they are hard to spot on a die shot.
Supposedly, they were able to make the TSVs smaller because the 3d cache isn't getting stacked on top but on the bottom with the CCD on top.

That solution seems so obvious now with all the advantages that it would provide like better thermals so it should clock much closer to the non 3d CCD and supposedly allows overclocking.
Really the only downside would be the change to packaging/assembly but who knows if that really changes anything for them that would impact the complexity of the process, i.e. more expensive or time consuming.

Video investigating the die shot of Zen5


WCCFtech article about the delidded Zen5 3d part

9800X3D coming Nov 7th.  Great timing, too, since my 13900K shit the bed.

Doh

While the extra 8mb of cache, higher boost and cpu direct overclock are nice, I don't trust manufacturer performance numbers. Especially after the 9000 non 3d launch.

Got a 7800x3d (with SM2 & UA for free) for $360 the day the 9000 non x3d series reviews came out and paired it with an asrock x670e taichi that can 'overclock' the 7800x3d through a seperate eclk.

Finally built it last weekend, only to find out one of my sticks of ram is bad, new pair should be here today, so I've got that to finish up this weekend.

Meh. An 8% average uplift isn't that impressive anyway. If they were going to lie, I'd think they'd embellish more. 8% isn't bad either, but I still don't play on buying into this generation, unless the 9950X3D really knocks socks off.

Yeah, 8% is pretty much just "fine" and doesn't seem like it would be worth shaking customer confidence over, especially after the recent launch numbers for the non X3D variants.

I won't be able to make it to Microcenter until 11/10 at the earliest.  Hoping they'll still have a decent selection of mobos.

If AMD can fix the cross chiplet latency issue AND put vcache on both 8 core chiplets (thus eliminating core parking), that would be fucking killer. They know this, but can/will they do it?

That's the current endgame, though, a few generations away. The actual implementation will likely change before we see it in action, as more advanced packaging techniques and interconnects get developed.
With currently available techniques, the IOD gets turned into an active interposer/bridge chip so the CCDs with Vcache can be stacked directly on top with a small <120mm2 GPU chiplet with Vache also stacked on top.
Active interposer means you could throw HBM on it or use edge connects for a 256/384bit LPDDR bus. That sort of implementation would save a bunch of power and latency by removing or reducing the amount of external interconnects.

Edit- It mainly is a cost/capacity issues right now. Advanced packaging for a high volume market without crazy margins is coming but probably not right now with the current bubble.
Most of the lower cost implementations don't have the same performance or higher bit/j costs, though there are some promising concepts on that front.

Gamers Nexus talks about some of the changes with the new Vcache setup

Looks like a pretty good improvement, putting the cache under the processor tile is a great idea for cooling.