AMD's Ryzen 9000X3D CPUs with 3D V-Cache reportedly delayed to CES 2025

[Redaktion]

AMD's highly anticipated Ryzen 9000X3D CPUs with 3D V-Cache technology have reportedly been delayed to CES 2025, a significant delay from the earlier rumored date of a late 2024 launch. However, the delay may allow AMD to unveil its full lineup of new AM5 platform products at the prestigious CES event, including the B850 and B840 chipsets.

https://www.notebookcheck.net/AMD-s-Ryzen-9000X3D-CPUs-with-3D-V-Cache-reportedly-delayed-to-CES-2025.878514.0.html

[Mr Majestyk]

Too late AMD. Clearly X3D also sucked as badly as just released Zen 5% models. The new motherboards are also a joke and make no sense at all to buy unless you simply must have USB4. B840 is ultra trash tier, even worse than the already pathetic A620.

Intel just needs to deliver on the Lunar Lake and Arrow lake platforms, and they will have AMD sweating for the first time in a decade.

[OP12RX10VI]

Intel just needs to deliver

Yea, I'm abit worried about that considering their history / track record (e.g. letting CPUs crash at math for over 1.5+ years). They seem equally as incompetent.

This year for desktop might be the year of just getting cheap X3D zen 4 deals like 7600x3d. As the lack of avx-512 on arrow lake is a big no for many.

Regarding lunar lake, by the time it becomes widely available it'll have to deal with discounted Z2 extremes and lower binned / CU enabled strix point. There's also kracken point and strix halo coming soon early next year CES, which AMD may fix some of zen 5's initial windows issues by then.

So yea, think I'm gonna wait till atleast CES next year before making a decision on lunar lake, as don't wish be massively outdated within a period of few months.

[NikoB-+]

X3D and the same increased L3 cache are a pointless crutch on x86 due to the shamefully low RAM bandwidth on all x86 in the consumer segment, but not on servers, where the bandwidth has long exceeded 1TB/s with HBM3+ memory.

It will be interesting to see when the delayed Zen5 Strix Halo with a 256-bit controller finally comes out - what will be the efficiency of the memory controller. I bet that it will hardly be more than 60% there. Apple, to its shame, even in top SoCs with a 512-bit controller actually provided only 30-33% of the theoretical declared bandwidth of 400 GB/s. Hence their epic failure with igpu - the lack of bandwidth for shared system memory is obvious.

At the prices that Apple demands for laptops with Max chips, they could very well install soldered 16-32GB of HBM3 memory directly in the chiplet as independent VRAM, like AMD in their SoCs, but something holds them back even in the HEDT segment from an obvious solution that will speed up igpu several times and provide easy support for many of the latest standards like DP2.0+/UHBR20 for 8k monitors, which we have been sadly waiting for at least 11 years in the mass segment...

[NikoB-+]

Sales of new AMD desktop processors are bad precisely because they cannot offer the buyer anything fundamentally new in technology or performance.

The HEDT segment has long been required to have only 256+ bit memory controllers with a real (not theoretical) throughput of 250 GB/s+. This is a disgrace to x86 as a whole.

Intel also has nothing to offer. And even against the backdrop of wild price increases for memory and SSD since the summer of 2023. Therefore, sales of everything are falling vertically, like a jack, and workers are being fired en masse.