Intel is launching new CPUs for the LGA 1954 socket, refreshed “Plus” models of the Core Ultra 200S “Arrow Lake” lineup. These focus on offering great bang for buck, adding more cores. As a result, Intel abandoned the pricier Core Ultra 9 flagship and gave full core count to the lower-tier Core Ultra 7 270K Plus. In the end, it will even be faster than the previous Core Ultra 9 285K—in practice, ironically running at higher clock speeds.
The Core Ultra 9 285K has a maximum boost of 5.7 GHz and will likely remain faster in single-threaded applications. However, that will be the only advantage the older model keeps over the new Core Ultra 7 270K Plus. Well-known leaker Jaykihn has now published detailed boost data for the new model, indicating that in multi-threaded workloads—and likely also in games—the Core Ultra 7 270K will have the advantage or at worst match it in raw performance, while also benefiting from higher uncore and chiplet interconnect speeds.
For the Core Ultra 9 285K, maximum clock speeds under full-core load are 4.6 GHz for E-cores and 5.4 GHz for P-cores. This 5.4 GHz maximum applies when 3–8 P-cores out of the total eight are active. The processor features factory-designated preferred cores that can reach higher clock speeds than the others—these can run at 5.6 GHz using so-called Intel’s Turbo Boost Max 3.0 technology (Intel’s long-standing term for preferred cores). If temperatures remain below 70 °C, the clock speed can increase slightly further to 5.7 GHz, which Intel refers to as Thermal Velocity Boost. However, both of these enhanced “turbo” modes can only be applied to one or two cores; with more cores active, the maximum is 5.4 GHz.
For the new Core Ultra 7 270K Plus, these details are now also known, and it turns out that under all-core load (when 3–8 P-cores are active), the processor runs at the same 5.4 GHz clock speed. So the 285K’s edge only appears when just one or two cores are active. The Core Ultra 7 270K Plus does not have Thermal Velocity Boost and, although it offers preferred cores (Turbo Boost Max 3.0), these only increase clock speed to 5.5 GHz.
However, when running a multi-threaded program that utilizes all cores, the Core Ultra 7 270K Plus will actually be faster than the Core Ultra 9 285K. While the all-core boost for P-cores is 5.4 GHz on both processors, the new Core Ultra 7 270K Plus increases the maximum clock speed of its 16 E-cores. In the Plus model, they can run at 4.7 GHz instead of 4.6 GHz—and this still with all threads active.
In the end, the impact may be relatively small, but higher raw performance will likely also add up to gains from increased D2D interconnect clock speeds in the Plus models. We therefore expect that, except in single-threaded applications, the Core Ultra 7 270K Plus will almost always outperform the older 285K high-end SKU. This will likely apply to games as well, which typically use more than one or two threads and therefore their performance will be dependent on the all-core P-core boost clock.
The fact that multi-threaded performance will compete with the Core Ultra 9 model likely led Intel to cancel the planned Core Ultra 9 290K Plus. In benchmarks, it might have offered very little additional performance compared to the 270K Plus.
Meanwhile, for the second new processor, the Core Ultra 5 250K Plus, the situation with clock speeds compared to the previous Core Ultra 5 245K somewhat differs. Its E-cores do not get faster and still run at a maximum of 4.6 GHz, as in the 245K. However, the number of E-cores increases from 8 to 12, so multi-threaded performance will significantly improve anyway. On the other hand, the maximum boost of the Core Ultra 250K Plus on P-cores increases compared to the 245K—from 5.2 GHz to 5.3 GHz for 1–2 cores, and from 5.0 GHz to 5.1 GHz for 3–6 active cores.
Maximum turbo power consumption remains unchanged: it is set at 250 W for the Core Ultra 7 270K Plus and 159 W for the Core Ultra 5 250K Plus. These limits may prevent reaching the stated all-core boost clock speeds in practice if power consumption at those clocks exceeds the power limit. However, power limits can optionally be unlocked in motherboard settings.
| Model | P-Core/E-Core | 1-2 P-Core TVB | 1-2 P-Core TBM 3.0 | 3+ P-Core | 1-8 E-Core | 9+ E-Core |
| Core Ultra 9 285K | 8+16 | 5.7 GHz | 5.6 GHz | 5.4 GHz | 4.6 GHz | 4.6 GHz |
| Core Ultra 7 270K Plus | 8+16 | – | 5.5 GHz | 5.4 GHz | 4.7 GHz | 4.7 GHz |
| Core Ultra 7 265K/KF | 8+16 | – | 5.5 GHz | 5.2 GHz | 4.6 GHz | 4.6 GHz |
| Core Ultra 5 250K/KF Plus | 6+12 | – | 5.3 GHz | 5.1 GHz | 4.6 GHz | 4.6 GHz |
| Core Ultra 5 245K/KF | 6+8 | – | 5.2 GHz | 5.0 GHz | 4.6 GHz | – |
Faster NGU
Intel has also increased the uncore clock speed (NGU) of the processor—the default value for the Plus SKUs is 2.8 GHz, compared to 2.6 GHz on previous processors. In this case, the increase is more modest than the 900 MHz rise in the clock speed of the D2D interconnect linking the chiplets. In the refreshed models, D2D interface should run at 3.0 GHz, whereas in the original models it was 2.1 GHz. It should be noted that users of the original models can increase NGU and D2D clock speeds via overclocking to match the Plus model advances (though a Z890 motherboard is required).
Cheaper “KF” versions of processors? Only the 18-core 250KF Plus is getting one, no 24-core model
In the official presentation of the refreshed Plus models last week, Intel mentioned only two models: the 270K Plus and the 250K Plus. Previously, we stated that slightly cheaper KF models without integrated GPUs should also be released alongside them, providing a more cost-effective option for gaming PCs. However, it now appears this will only apply to the Core Ultra 5 250K with 6+12 cores, which will be offered in such a cheaper Core Ultra 5 250KF Plus variant. As usual, the only difference should be the removal of the integrated GPU, which could reduce the price by a small sum.
Intel is not planning a 270KF Plus model, so it will not be possible to save money in this way with the full 8+16-core configuration. This mirrors the approach taken with the original processor lineup, where Intel also did not release a KF version of the 285K. On one hand, Intel loses a more affordable option, but on the other hand, it forces buyers of the 8+16-core configuration to pay a bit more.
Skymont at 4.7 GHz
The mentioned 4.7 GHz clock speed of the E-cores is, incidentally, the maximum that AMD officially achieved with the Zen 2 architecture in Ryzen 3000X processors (the Ryzen 9 3950X model), which may serve as an interesting comparison. In Zen 2’s case, however, this was achieved at relatively high voltages leading to high power consumption, whereas here it is an all-core boost achievable across all E-cores (and the real maximum frequency of the cores is likely higher). It should also be noted that Zen 2 silicon was manufactured on a 7nm process, while Arrow Lake relies on a two-generations-newer 3nm TSMC process, so this is not a fair comparison.
Even so, it shows that the Skymont architecture is not as weak in terms of clock speeds as is often assumed based on its origin in the “Atom” lineage.
It remains to be seen whether Intel can achieve a similar jump in clock speeds as AMD did with Zen 4. That would help the E-core architecture eventually replace today’s large cores in the role of “P-cores,” which is reportedly what Intel plans to do somewhere down the line.
Sources: Jaykihn (1, 2, 3, 4, 5)
English translation and edit by Jozef Dudáš
⠀
