some did say 2023 on the ihs though
The year on the IHS is not the manufacturing date of the CPU itself. Never has been.
Does this mean delay of reviews to?
Of course it will.
The X3D parts tend to be quite a bit more expensive than their vanilla counterparts, and not everything revolves around gaming. Users who are okay with the level of performance the 9700X offers can save money by going for it instead of a future X3D chip.
Must be a very specific workload that benefits from 8, instead of 6 cores but does not benefit from X3D cache. I my opinion the x900 parts where always the most pointless. Moving to dual-CCD has downsides for gaming and games dont really use much past 8 cores anyway. Meanwhile the price is such that if multi-core performance is important then moving straight to x950 makes more sense. Also x900 comprises of two CCD's that are essentially two x600 parts and thus have lower performance.
can't wait for the failed ones to somehow reach the market somehow
There were pictures of some sold 9700X boxes but if they were real and not fake, then i have not seen any proper testing on them. Only one ES sample has been put trough what i can call proper testing.
Who wants to guess TIM between the IHS and die, or is it some bad soldering/wiring on the fiberglass substrate.
Maybe IHSs that are warped?
There is not TIM between the IHS and die. I dont think any of the Zen series non-APU models have used TIM. They are soldered designs. Only the desktop APU models were/are? using TIM. IHS warping is a nonissue on AM5 as the IHS is fairly thick and does not warp like Intel LGA1700 does.
It's been almost 2 years since a cpu generation released from them and they still can't launch somthing without issues it's concerning at a minimum.
Every company has some issue at launch.
Recall all the golden samples we sent reviewers lol
There has been no documented case of any of the big RGB sending golden samples to reviewers for many generations now. And with silicon already pushed to the edge from the factory there is really no headroom to really produce a golden sample anyway. Not like back in the day where a user could get extra 1Ghz OC on air with ease. It would be only a percent faster at best but it would be quickly discovered and the PR backlash is not worth it. Besides golden sample or binned can be categorized as different things: lower than average voltages to achieve advertised frequency, higher clocks at average voltage, lower than average temperatures, better IMC that tolerates higher memory frequency etc. It's not a single metric and often a CPU that is better than average on one of these metrics may be below the average in another. That is also why XOC uses different chips for memory and frequency records.
They want to launch something around the time of Arrow Lake most likely. I wouldn't be surprised if it drops like a week or two before.
Arrow Lake is reportedly coming in October-November. Zen 5 will not wait that long. More likely AMD will introduce Zen 5 X3D by the time of Arrow Lake's launch or shortly after.
9600x vs 13600k and 9700x vs 13700k, yeah, doesn't look good for amd. With or without the intel fixes.
Based on what exactly?
There are no reviews and no official prices but you already declare that Zen 5 "doesn't look good for amd" against 13th gen (?) for some reason?
How do you do QA on units that have already shipped out? Weird.
Current units should be mostly at the hands of OEM's and system integrators. Not public or reviewers.
Maybe they don't, they just had to delay for other reasons and used this reasoning to show they are better than Intel. It's a nice marketing win without having to directly attack Intel, it's genius. Maybe the GPU department at amd should be taking notes.
It's a marketing win for sure. Not every day we can say that about a product delay.
But how would they know AFTER shipping?
OEM's do have internet in their labs i presume...
Well what competition are the 9600x and 9700x going to be beating? They are still 6 and 8 core parts. They are going to be slower than the 13600k and the 13700k,and not by a small margin
I ask again - based on what data? That these are "only" 6 and 8 core parts? They do have SMT that has always been more performant than Intel's HT.
until amd decides to move to 12core single ccds with 3d on top, or intel increases their pcore count.
What is the workload that benefits from 12 cores (or held back with 8 core) on single CCD with X3D on top?
All i can think of are latency sensitive workloads that need to have all their cores on a single die but usually these dont require 12 or more cores at once.
Sure i too would like a CCD that houses more than 8 cores but ultimately introducing more cores is pointless if those cores are bottlenecked by other thinks like memory speeds or lack of cache. If i could design Ryzen myself i would eliminate the weak iGPU (like it was pre-Zen 4). Put both the IO die and X3D (double with 128-256MB) on top of the 8 core CCD and IHS that has better heat transfer (breaking backward compatibility in terms of height) and none of those cutouts for capacitors on the sides. This wil never happen but it's fun to dream. Adding more cores is pretty low on my list.
