It’s a rather unfortunate launch. It’s showing itself at the top of synthetic multi thread and single thread benchmarks and then failing to beat the 7000 series (even non x3d variants) in games seemingly due to thread scheduling going across ccds.
I suspect it’s a hiccup and easily fixed with a software update but unfortunate since the launch benchmarks tend to be the ones that stick around.
Interestingly enough, there is a bug with Windows causing bursty workloads like games to perform poorly right now on Zen 5. Only problem is, it also applies equally to Zen4, so it ends up not changing the delta between them at all.
And with some chips like the 9950X you have to do that nonsense of enabling the core parking profile with MS's Xbox Game Bar to get games to run half decent, which is a bit annoying.
In general the 9900/9950 aren't made for gaming regardless. They're for productivity.
Games are not one of my PC use cases so I am pleased to see that the 9950X is the new king of the hill for compiling software. My 14900K is glitching out so I need to replace it and the 9950X is a lot cheaper than a Threadripper or Xeon-W.
Note that you have warranty for the 14900k. You could let Intel replace it for free, after all it's glitching because of their issues, be it overvolting or oxidation.
Given the issues of those 9000 series processors I'd be wary of them, the 7950X could be the better option.
I doubt jeffbee lives in a backwater place such as mine, but Intel do not handle RMAs in many countries around the world. There's simply no such thing as RMA here for most brands you're familiar with. If my hypothetical 14900k dies and the seller I bought it from refuses to provide a replacement (for example, by claiming that I overclocked it), I'm fucked. It doesn't really matter if Intel themselves issue replacements or not.
(And no, I am not from and of the sanctioned countries or territories. It's the reality of an unimportant market.)
The Intel warranty in unfortunately not advanced replacement, so I would be without my CPU for an indeterminate amount of time while they give it a think. The issue with the 7950X is that it isn't faster on my workload than an i9-14900K, even one that has been detuned to work correctly.
The "issues" with the 9000 seem to be contained to MS Windows, an operating system I haven't used since the 1990s.
And with the stories about how RMAs are going with Intel right now, I definitely wouldn't willingly walk into that situation even if I could stomach the down time. From the outside looking in, it seems like Intel doesn't have the resources to RMA all of these CPUs and are throwing up roadblocks to either slow down or stop the bleeding. I'd expect retail purchases to get the most screwed in that situation since Intel will want to make their larger customers right first.
1. Buy new 14900 from a brick-and-mortar
2. RMA your current CPU
3. Wait until the replacement arrives
4. Return the sealed replacement unit to the store
5. ???
6. Profit
Unless your brick and mortar checks serial numbers on returned items to protect itself against exchanging a new part for a refurbished or used one. Which I've seen more than once.
I guess I'm not fully understanding their Zen4 vs Zen5 conclusion. I get that if you've already got a Zen4 machine you may not feel it is worth it to upgrade to Zen5. But if you've got some 5+ year old machine and need a new one, they aren't really saying that you might want to get a Zen4 instead of Zen5, are they?
The performance difference between Zen4 and Zen5 (which are on the same socket) does not justify the price difference one has to pay at this moment. And this just might be intentional by AMD to clear the Zen4 inventory before Zen5 gets the eventual price cut.
I'm soon to be building an updated gaming PC (while ship of theseus'ing parts like my 3090, SSDs, etc), and it looks like this Zen 5 series is...not it. My sights are likely set on the 7900X3D as the processor I'm picking.
I mean, it works properly, but a bit slower. The driver just detects if the kind of application currently running (i.e. a game) needs to be confined to only one CCD.
The core to core latency between the two CCDs seems to have doubled from the 7xxx series according to some latency benchmarks I saw yesterday.
I suspect it’s a hiccup and easily fixed with a software update but unfortunate since the launch benchmarks tend to be the ones that stick around.