They launched this chip in 2023. Where were you?
My current laptop runs an R9 7940HX, the non 3D version. It's from last year, pre 9000 series and guess what!? The 9955HX3D was launched LAST year.
Why is this article an article? The sponsor paid for it is NOT a good answer.
Please, get back to the cutting edge.
I believe that's an easy one to answer:
AMD's CCD/IOD approach gives them the flexiblity to sell the same silicon in all these different product lines and form factors.
V-Cache was originally a design targeting EPYC servers, only, where some special HPC workloads and EDA tools profited tremendously from those bigger caches, I've heard things like more than 100% faster for EDA and genomics.
That also meant giant bucks (half # CPUs = near double price to break even), because those V-cache EPYCs may not be high volume, but they can charge premium for a technology that wasn't cheap to design, but seems to turn out much cheaper to produce than it's sold. Again, rumors, but the 1st generation pure production extra pure cost was quoted at around $20.
The same may not be (have been) true for the logistics since the assembly of the V-cache costs extra steps (and time) and separate production lines, that need to be made and planned with a capacity in mind, which cannot be ramped up and down easily. Eager consumers have little patience for how slowly chips production and assembly can be shifted in any direction.
Anyhow, V-cache for consumer CPUs was a skunk-works job, accidental discovery or similar and it was only later that AMD jumped on that with a vengance, but never with a desktop/mobile-first product.
But since AMD needs to focus on profits as their primary goal, all of their CCD designs are primarily focussed on EPYC server CPU, they just happen (not entirely by accident) to be also pretty good for consumer deskops and mobile workstations.
So after producing their CCDs, the CCDs go where the margin is. After each refresh, they first go into EPYCs, anything not sold there goes into desktop and anything not sold there goes into mobile workstations,
unless some OEM picks the mobile parts at the initial high price.
So the fact that 3D CCDs now are finally assembled into mobile CPUs is simply because Zen 4 V-cache CCD demand in the other markets has dried up.
If market demand had been high enough for mobile V-cache parts before, you'd have seen them there much earlier. But every V-cache CCD that hadn't already been sold into EPYC servers yielded much better margin as a 7800X3D, or perhaps as a 7950X3D, so that's where it went.
And it's the very same with the Zen 5 CCDs: If you ordered 9955HX3D chips in numbers and at a price where selling them as EPYCs is less attractive, you'll get them. But evidently the mass market simply isn't there, ...until they sunset and prices go down.
EPYC sales likely go first, it's a very hot, but also limited market and once the next generation is up, the previous one becomes a no-go. Mobile workstations are somewhat similar, which is why small Chinese OEMs love picking up the Intel variants, which are much less flexible in terms of late server/desktop/mobile/mobile-on-desktop assembly, late in their life cycle.
Zen 4 desktops showed little sign of abating until Zen 5 supply was sufficient so it's only now that V-cache CCDs wind up in the low-budget distribution channels.
Because that is what Minisforum is specialized on: selling sunset mobile hardware in NUC or desktop form factors at budget prices.
So don't complain, it's your consumer buying patterns that rules availability. If you'd ordered a few million 9955HX3D at $2000 a pop (or whatever EPYC customers are ready to pay) last year, you'd have them!
Facebook
Twitter
Instagram