The CPU duel for next-gen desktop CPUs between the AMD Ryzen 9000 (“Granite Ridge”) and Intel Core Ultra 200 (“Arrow Lake”) could begin as early as Q3 2024, according to usually informed reports. circles in Majlis Channel Forum Regulated Chinese OEM partner can be heard. PCGH's editorial team is once again collecting all the information and rumors that have already leaked and categorizing them accordingly. The big question will be whether Intel will be able to compete with Zen 5 with 3D V-Cache with its upcoming hybrid processor architecture.
Ryzen 9000 vs Core Ultra 200
It has been clear for a long time regarding the infrastructure, which in the case of the Ryzen 9000 will remain the AM5 platform with the LGA-1718 socket, and at Intel, the new LGA-1851 socket will debut next year with the Core Ultra 200. When it comes to the new chipsets, AMD will build on further development of the existing A620 and B650(E) chipsets and the upcoming chipsets, called Platform Controller Hubs (“PCHs”), are said to have these new specifications, standards, and feature sets:
Ryzen 9000 + Ultra Core 200**
- 256GB DDR5-5600 to DDR5-6400
- Wi-Fi 7 and Bluetooth 5.3 or 5.4
- LAN speed of 5Gbps to 10Gbps
- USB4 at speeds of up to 40 Gbps
The architectures themselves will also look significantly different from the current AMD Ryzen 7000 and Intel Core-i-14000.
Ryzen 9000**:
- Zain 5
- RNA 3.5 (tRNA 2)*
- 4, 6, 8 and 16 processor cores
- 8, 12, 16 and 32 threads
- 65 to 170W TDP
- N4/N5 from TSMC
- 3D Cache V
*) Grantie Ridge should take over responsibility for Raphael's I/O template.
Core Ultra 200**
- Efficiency cores (“Skymont”)
- Performance Core (“Lion Cove”)
- 6, 10, 12, 14, 16, 20, 24 processor cores
- Up to 8 P cores + 16 E-cores
- 35 to 181 W (241 W)
- Intel Xe-HPG+ with XMX
- From 12 to 32 threads
- Intel 20A
**) Not officially confirmed yet.
There will also be a significantly larger number of 5th generation PCIe lanes and a significantly expanded I/O area with more internal and external connections for components and peripherals. At least at Intel this is already guaranteed.
Source: AMD
Source: AMD
While the Intel Core Ultra 200 (“Arrow Lake”) will be based on the Core Ultra 100 (“Meteor Lake”) and will adopt modern I/O tiles, the I/O chip of the upcoming AMD Ryzen 9000 processor (“Granite Ridge”) is a question mark Larger. The rumors here aren't really clear yet.
Details about the AMD Ryzen 9000 processor
As already speculated, the number, so-called core count, of processor cores based on Zen 5 and at 4 nm (“N4”) will continue to roll off the production line at the world's largest contract manufacturer TSMC.
Source: Moore's Law is dead
Source: Moore's Law is dead
The original document, which is available to the author of this report but cannot be published for source protection reasons, shows that the AMD Ryzen 9000 on the desktop will contain the following basic data.
- Zen 5 CCD devices (“Eldora”)
- Zen 5 CPU cores (“Nirvana”)
- 6 to 16 Zen 5 processor cores
- Power loss from 65 to 170 watts (“TDP”)
- Up to 64 MiByte L3 + 16 MiByte L2 Cache
- 4nm (“N4”) manufacturing at TSMC
- Release in the second half of 2024
As with Zen 4, where AMD switched from N5 to N4 later during the architecture's life, a subsequent switch from N4 to N3, N3E, or N3P could also be envisioned for Zen 5. However, several sources indicate that the launch should take place at N4 .
There are still a lot of question marks
It is not yet known what improvements AMD will make to the I/O die, for example regarding higher memory clock frequencies. If the I/O chipset of the Ryzen 7000 (“Raphael”) and Ryzen 7000X3D (“Raphael-X”) remains 1:1, at least 28 5th Gen PCIe lanes will be assigned.
When will it start?
It can currently be assumed that the Zen 5 architecture will celebrate its debut in the first quarter of 2024 in server processors of the Epyc 9005 series (“Turin”) and from the second half of the year in desktop processors (“Granite Ridge”) and the first APUs are used ( “Strix Point”). We'll probably have to wait until early/mid 2025 for “Strix Halo” and “Fire Range.”
All previous information and rumors should still be taken with the utmost caution, as AMD itself has yet to comment on its upcoming processors and APUs with Zen 5.
It is also still unclear which nomenclature will eventually be used. In the mobile segment, AMD now wants to officially link the first digit of the model name to the year of release, but does this also apply to desktop?
Question after question, but the PCGH editorial team will of course stay on top of the topic and report accordingly.
Intel Core Ultra 200 details
Following Intel's soon-to-be-released hybrid series While the Core 100 (“Meteor Lake”), which was officially presented in New York on December 14, will only be used in the mobile segment, the Core 200 (“Arrow Lake”) will also address classic desktop PCs again in 2024. ..
As it currently stands, the code for the latest patch for Intel Graphics Compiler (“IGC”) As it turns out, the Core (Ultra) 200 series processors are based on Arc Xe-LPG+, an enhanced Alchemist+-based graphics unit with dedicated XMX drivers.
Core 200 with Arc Xe-LPG+ and XMX engines
As can be seen from the IGC patch, when the generation changes from the current Core 100 (“Meteor Lake”) to the Core 200 (“Arrow Lake”), Intel will upgrade its iGPU from Arc Xe-LPG to Arc to first use matrix accelerators (” “XMX” engines are already known from desktop graphics cards and are dedicated internally to the acceleration. In Meteor Lake, these tasks are still solved with DP4a instructions.
Source: Intel
XMX drivers accelerate Intel XeSS
XMX Engines, of which there are between 96 and 512 units on Arc (“Alchemist”) desktop graphics cards, accelerate not only classic AI workflows but also Intel Xe Super Sampling (“XeSS”) AI upscaling, Compared to Nvidia also practices it with DLSS and its Tensor cores.
Source: Intel
Source: Intel
Maximum of 8 P cores and 16 E cores
There is also widespread agreement on the number of processor cores: Intel should still rely on a maximum of 8 P-Cores (“Lion Cove”) and 16 E-Cores (“Skymont”), which would likely offer a slightly lower clock speed than This was the case with Raptor Lake Refresh, which is due to the new manufacturing process (“Intel 20A”). This clock penalty is often observed when switching to a finer node.
New socket with new chipset
Intel's Core 200 and Core Ultra 200 series desktop CPUs will use Intel Foundry's new 20A in-house manufacturing process, which Intel expects will provide significant efficiency gains.
Source: Intel
The new node is a pure NMOS process that aims to improve the performance and efficiency of the 10nm ESF node. The new LGA1851 socket is used as the platform.
As Intel CEO Pat Gelsinger emphasized, Meteor Lake is well positioned for Q3 2023 and Arrow Lake and Lunar Lake for 2024, and the company is also very satisfied with the development of Emerald Rapids and Sierra Forest in the enterprise division.
The Intel President also focused on the positive development of Intel Foundry Services and pointed to the technological development and leaps in the transition from Intel 7 to Intel 4 and Intel 3.
Your opinion is needed!
what's your opinion about this subject? The PCGH editorial team would be happy to hear your well-founded opinion in the comments on this report. To comment, you must log in to PCGH.de or the Extreme forum. If you don't have an account yet, you can register here Registration without obligation. When commenting, please note the applicable regulations Forum rules.
“Subtly charming coffee scholar. General zombie junkie. Introvert. Alcohol nerd. Travel lover. Twitter specialist. Freelance student.”
More Stories
Deeper videos offer many improvements
Data Leakage: Android TV can expose user's emails and files
How did life begin on Earth? Munich researchers find important clues